Automated computer analysis of the electrocardiogram

Interpretation of the electrocardiogram (ECG) has become one of the larger tasks now being placed on the cardiologist. Besides the magnitude of the task, variability among cardiologists and aid for the rural population are two other reasons for looking toward automation in an effort to reduce the burden placed on the cardiologist. This thesis presents the work done in developing computer programs to automate the analysis of the ECGo. Three phases of ECG analysis must be considered. The first is analysis of the contour of the ECG complex in order to detect abnormalities in the electrical conduction system of the heart muscle. The second phase is analysis of the heart rhythms, i.e., the relationship between ventricular depolarizations and atrial depolarizations. The final phase is an analysis of the changes which take place in the ECG over a period of time. Programs to tackle each of these problems have been developed and are described in the thesis. Two problems exist in developing computer ECG analysis programs. The first is electrical "noise" which is superimposed on the signa1. The other problem is development of precise logic for proper interpretation of the ECG. Each of these problems are discussed with solutions given to a portion of the difficulties. The contour and rhythm programs were tested against 202 ECGls which had been interpreted by one or more cardiologists. Detailed results of the study are included in the thesis. These results showed 8% false positives and 2% false negatives. The sequential analysis program was evaluated using 50 pairs of ECG's. It was found that the program detected 22 ECG's which showed minor changes and 6 ECG's which showed major changes. The future of such programs are considered with the type of research which is necessary to overcome the present difficulties in the program.

AUTOMATED COMPUTER ANALYSIS OF THE ELECTROCARDIOGRAM by Thomas Allan Pryor A dissertation submitted to the faculty of the University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Electrical Engineering University of Utah June 1972 This Dissertation for the Doctor of Philosophy Degree by Thomas Allan Pryor has been April approved 1972 Supervisory Committee Supervisory Committee Supervisory Committee Chairman, Major Department UN\VERS\1Y OF Ul J.H LIBRARIES ACKNOWLEDGEMENTS I would like to first of all Department of Biophysics research. In particular, effort: England Dr. who the support nd the acknowledge the entire staff Bioengineering for their support in this following people aided greatly in this Reed Gardner and his staff of hardware helped in the given by receive credit for her Finally, I would clinical effort in programming. specialists, Willard I would also like to Mrs. E. Stockham. patience in typing and re-typing this like to acknowledge Dr. Alan E. evaluating the work, and whose faith and continual Dr. acknowledge Marion Nordfors should Thomas Oro of the work. Lindsay for his critical Homer R. Warner, without encouragement nothing would have been completed. iii TABLE OF CONTENTS Page iii ACKNOWLEDGMENTS vi LIST OF TABLES vii LIST OF FIGURES , ABSTRACT CHAPTER I 1. 2. 3. Introducti on . . 1 3 6 . . . History of Computerized ECGls Design Criteria for this Research . . CHAPTER II 8 10 13 Definition of the Electrocardiogram Noise Sources on the Electrocardiogram Fi 1 teri ng of the ECG 1. 2. 3. CHAPTER III 1. Introducti 2. Samp 1 i ng the ECG Waveform Recognition 3. 4. 5. on 23 23 27 31 33 . . . . Measurement of the Parameters Logic for Classification . CHAPTER IV 1. 2. 3. 4. 5. Introduction Determination of the R-R Intervals P Wave Recogni ti on Recognition of Premature Depolarizations Classification of the Rhythm. ... . . . . . . . . 35 36 32 40 42 CHAPTER V 1. 2. 3. Introduction Criteria for Comparative Statements Calculation of Limits for Parameter . iv . . Changes. 44 45 48 CHAPTER VI 1. 2. Evaluation of Contour and Rhythm Programs Evaluation of the Sequential Analysis Program. 52 58 . CHAPTER VII Significance of this Research 60 62 Future of this Research BIBLIOGRAPHY 64 . APPENDIX A Mathematical Definitions of Parameters Used in Contour Analysis 68 APPENDIX B Criteria for Interpretations for Morphology Findings. 72 Rhythm Determination 79 APPENDIX C Criteria for APPENDIX D Criteria for ECG Diagnostic Statements 81 APPENDIX E Criteria for ECG Serial Statements. 85 APPENDIX F. 93 VITA. 98 v LIST OF TABLES TABLE Page 2.1 COMPARISON OF QRS DISTORTION 2.2 FILTER WEIGHTS FOR SAMPLING RATE OF 200 3.1 CLASSIFICATIONS FROM MORPHOLOGY PROGRAM 4.1 LOGIC FOR RHYTHM 5.1 LIMITS 5.2 RESULTS OF REPEATED ANALYSIS OF SAME ECG 49 5.3 RESULTS OF REPEATED ANALYSIS FROM SAME PATIENT 50 6.1 ABNORMAL ECG'S INTERPRETED BY COMPUTER AS NORMAL 53 6.2 ABNORMALITIES REPORTED IN NORMAL ECG's 54 6.3 QRS INTERPRETATIONS 55 6.4 ST-T INTERPRETATIONS 56 6.5 P WAVES 56 6.6 RHYTHM 57 6.7 CHANGES DETECTED IN SIX PATIENTS WHICH WERE OF MAJOR VALUE 59 F.1 WEIGHTS FOR P WAVE FILTER 93 F.2 ADDITIONAL RHYTHM INTERPRETATION IN NEW ARRHYTHMIA PROGRAM 95 22 SAMPLES/SECOND INTERPRETATION (IN VOLTS) FOR PARAMETERS vi 22 24 35 IN SEQUENTIAL ANALYSIS 48 LIST OF FIGURES FIGURE Page 2.1 TYPICAL ECG COMPLEX 2.2 AN ECG WITH BASELINE DRIFT 10 2.3 ECG WITH 60 CYCLE NOISE SUPERIMPOSED 11 2.4 ECG WITH RANDOM MUSCLE NOrSE SUPERIMPOSED 12 2.5 ECG WITH SINUSOIDAL RESPIRATORY VARIATION 13 2.6 ECG SHOWING PATIENT MOVEMENT 13 2.7 SPECTRAL DENSITY OF A TYPICAL ECG 14 2.8 MAGNITUDE OF FILTER B 16 2.9 MAGNITUDE OF FILTER C 17 2.10 MAGNITUDE OF FILTER D 18 2.11 RESULT OF FILTERS BAND C ON AN ECG WITH MUSCLE NOISE 19 2.12 RESULTS OF FILTERS BAND C ON AN ECG WITH 60 CYCLE NOISE 20 2.13 RESULT OF FILTER BAND C ON AN ECG WITH NOTCHED 20 2.14 RESULTS OF FILTER D ON ECG WITH NOTCHED QRS 21 BLOCK DIAGRAM OF THE AUTOMATED ECG SYSTEM AT LATTERDAY SAINTS HOSPITAL 26 AN EXAMPLE OF AN ECG SHOWING MARKS LOCATING START AND END OF QRS 28 3.3 AN ECG WITH NON-LINEAR BASELINE DRIFT 31 4.1 LOGIC FOR PREMATURE DEPOLARIZATION DETECTION 41 5.1 EXAMPLE OF A SERIAL ECG COMPARISON REPORT 46 F.l RELATIONSHIP BETWEEN P WAVE AND P WAVE MATCHED FILTER 95 F.2 FREQUENCY OF CHARACTERISTICS OF P WAVE FILTER 96 F.3 RESULTS OF P-WAVE LOCATION ALGORITHM USING MATCHED FILTER 97 3.1 3.2 9 vii QRS ABSTRACT Interpretation larger tasks lation the on cardiologist. variability among cardiologists two other reasons are to reduce being placed now of the task, electrocardiogram (ECG) has become of the the burden placed for on looking the of the one Besides the magnitude and aid for the rural toward automation in popu­ effort an This thesis presents cardiologist. the work done in developing computer programs to automate the analysis of the ECGo of the contour of the ECG analysis in the electrical is analysis changes in order to detect abnormalities complex The second phase rhythms, i.e., the relationship between ventricular and atrial which take tackle each of these to The first is conduction system of the heart muscle. of the heart depolarizations of the must be considered. phases of ECG analysis Three depolarizations. in the ECG place problems have been The final over phase is period a developed and an of timeo are analysis Programs described in the thesis. Two fi rs tis problem ECG. problems e is 1 ectri exist in ca 1 "noi development Each of these se of developing computer II whi ch is ECG superi mposed analysis programs on the signa 1 0 precise logic for proper interpretation problems are discussed with solutions given to The 0 The other of the a portion of the difficultieso The contour and had been the study rhythm programs interpreted by are one or more were tested against 202 ECGls cardiologists. included in the thesis. which Detailed results of These results showed 8% false viii The positives and 2% false negatives. sequential analysis program found that the program de­ evaluated using tected 22 ECG's which showed minor changes and 6 ECG's which showed 50 pairs of ECG's. It was was major changes. The future of such programs research which ;s necessary to are considered with the type of overcome program. ix the present difficulties in the CHAPTER I 1. Introducti on With the advent of of the pretation consider we for all admissions, The usual (assuming a 600 bed one lith that all spend 7\ hospital have to measurements are made of variability of cardiologists for even on all a ECG"s) a current manual computerized the ECG. Not but also patients, logic after time mean vary, the cardiologist by so do their as fatigue has giving only exactly reproducible reading precision to 5.00 ECG same two on personal problems or x be solved can it will the of the ECG be followed interpretation. for left ventricular that the square of the is greater than The interpretations. same and training may by the the latest criteria be made available to can been established more interpretation interpretation. interpretation "QRS voltage QRS reading, Because the The need for standardization of ECG criteria the day. major goal for any computer­ single cardiologist reading the influence him, and thus, his Once the program Reduction of the time spent cardiologists. Outside factors such occasions. all screening If cardiologist interpret all of thus becomes problem associated with the is the experience is true ECG an big business. analysis program. Another methods one reading the ECG's. cardiologist reviewing ECG's ized ECG with inter­ programs, average of three minutes for each an hours become has would expect about 150 ECG's to be taken per practice would be these ECG's. would electrocardiogram (ECG) typical a well-patient health screening magnitude 106 mv2. can be expected. exactly time For example, hypertrophy" will of the maximum vector Hence, once during the criteria have been 2 established, exact stood who by all Health There care is meaning the use is areas becoming the United States presently within are 5,000 and 10,000 people without physician, a Hope for solving this problem lies in the automated central and it is under­ of areas population between let alone use of ECG a skilled cardiologist. paramedical personnel and lines for phone over increasing problem. ever now available, it is interpretation an at a computer facility and have the interpretation returned in real time. The above of development cally an With the communication systems techniques. to transmit the possible interpretation system. rural to the to the given a have set the tone for this program for automated ECG research, namely the interpretation which three separate programs need to be To do this, useful. The first is problems is clini­ developed. analysis of the ECG morphology (shape of the waveform) which results from abnormal conduction pathways through the heart. Abnormal pathways result primarily from the presence of diseased heart muscle which is no longer able enlargement of or one development of such - the be more gram. are an the program. A second of the depolarization from or disproportionate Chapter III discusses the analysis detects the heart rhythm depolarization regularity of these possible. The final in signal another heart chamber. events. of the ventricles and This analysis has proved magnitude of the signal of the atria and the large number of rhythms Chapter IV develops the logic used in the rhythm pro­ analysis compares different times and reports any useful electrical difficult both because of the small generated by which a temporal relationship the atria and the to conduct to evaluating the ECG's taken on the significant changes. progress of a patient who has same patient This is at particularly recently had a heart attack. comparison is also important in tnterpreting The the first two programs since "abnormalities" may have different clinical meaning for that individual a have occurred 2. the first two programs by seen over no changes History of Computerized ECG's 1960. still Two centers of (P, QRS, investigation forefront of at the the major were computerized initiated at this time and ECG At both centers interpretation. One center located at the U.S. from the ECG. Washington, O.Co analyze a single lead at a meters measured, pattern used by the well as the time from each of the 12 standard cardiologists diagnostic logic, an interpretation, i.e., data from each Frank in clinical 15 had developed a in closely the It vectorcardiography. studies, he claimed that the same clinical at the Veteran's Adminis­ simultaneous sampling leads calculated from the 3 leads. was Based on his information found in the 12 lead system could be derived from the 3 lead system. of the 3 orthogonal with this lead was 00c.11,12,18,24,25,26,44. Washington, amount of data of the 12 leads. three lead system that H. V. Pipberger developed his system Hospital approach cardiologist. Prior to this time E. use This since both the para- followed disadvantage of the Caceres program has been the necessary for Public Health Caceres whose approach cardiologists.5,6,7,42. lends itself to easy acceptance by the as A. C. headed by Dr. was ECG leads used almost universally by tration are T) and system for begun about has been extraction of the various waveform components has been to One was emphasis Service in was if long period of time. a Early work in the analysis of the ECG by computer own the results of done and the In this approach spatial velocity (SV) 4 Search algorithms the SV. Working cardiologists, were with now we defined to locate the waveforms of interest from a derived originally however, proved often were ECG. As a to correlate them with autopsy because findings. correlations This, by the time of autopsy other lesions present causing difficulties in correlating with result all most evaluating this classification technique In unsuccessful by parameters and statistical techniques for new classification of the ECG. he tried lead set different from that used are now made with the a previous interpretation given by skilled cardiologists. A third major development was started by Dr. This system Mayo Clinic in cooperation with the IBM Company. the 3 lead system, in but samples It is worthwhile to note here some L. of the primarily Cady early statistical methods and M. technique using selected Fourier coefficients of set of linear discriminant functions. achieved of uses certain of the standard leads in order to aid used in the classification process. a of the interpretations. some for Smith37 Ralph success some ventricle enlarged in separating using only normal Woodbury4 the ECG With this as described a the variables technique they ECGls from those with a pattern six of the Fourier coefficients from the vector leads. Another As the new early attempt by ECG was already measured. type was Lo was involved entered into this system it was "adaptive averaged new ECG was a interpreted with those patterns. filters". correlated with patterns If the correlation coefficient with sufficiently high, this that class and Stark39 particular proto­ as a member of If the correlation coefficient for all as the classes beginnings tried to increase the interpretation In made a class new ratios that exist when presence of certain Cox8,g the programs when is due This P largely works with P waves on other the not to rely on information from the P investigators such the of an wave. rhythm interpretation. of the shape on the same histogram Gersch17 was made rhythm problem. use use found in P-wave waveform. Simborn36 as (two as was for each a as as the found QRS). generalized such histogram. information, but results overcome only knowledge of depolarizations and R-R signal a rhythms based some waves this work involved using only the Bonner1,2 studied the Unfortunately this P To and used this of the Markov process Again and have been tried which Separation of possible. between ventricular problems effort to this timeo W. D. be true for others, e.g., 2:1 A-V block relationship to the made a histogram resulting from various rhythms Other work by W. attack signal/noise components of the ECG in the This has caused rhythms. techniques has (atrial depolarization) waves problem, several interesting statistical methods starting point on to the to search for criteria of data reduction which result in interval R- compared to the small where any deviation from the baseline is considered do not ECG techniques and investigators have programs the success of computer one super-imposition of this started with this was cardiologists.34,3S,40,42,43,44. by impressive. as many other performance of developing rhythm not been and low, only member. From these the was are not was subject We have since complete at 6 3. Criteria for this Research Design As noted above, satisfy the need for research has been to programs The first and Lanager19 the was have the 50-100 cps, Hence, most clinical Based spectrum. at 200 need for out the samples in the initial decisions made are per second per lead. interpretations. as true the logic More but in each instance this as response sampling was in use only chosen to frequency was chosen order of detail same defined here a in the digital found data means that each of the values of the parameters on in the process based were tried, degradation of the cardiologists. closely occurrance. interpre­ Boolean statement, which is The logic of the correlate with the decision on Only selected rhythms have been chosen of frequency logic for making morphological usually resulted a rate general statistical methods interpretation process of the clinician, the basis of high frequency interpretations. Boolean use (or false) depending correlation with the FrankeI4 data only in this on Thus the by evaluation of measured from the ECG. Boolean statements to was Boolean be made of frequency analog recording should be present A second decision evaluated analysis flat a these considerations, on Although recommended by the American Heart Association. as used for the automation of similar tation will set of recording equipment (including that most Latter-day Saints Hospital) has out to ECG. rate of the sampling pointed components of the ECG, by clinically useful a philosophy influenced many basic decisionso This 0 major emphasis in the development of this the more than 60 years for measurement, In this way experienceo primarily on relatively simple logic has been which developed permits diagnosis Thus, the program will adequately leaving only the rarer, A three lead system amount of data of a cases chosen since it was one for the Finally, ECG problem use the most complex), morphology matched filtering). of the easily For the wide using matched example, required filtering. well. a With research is directed. This was on done in the on standard signal (e.g., slope which may not have been would have as algorithms for locating various points general features detection methods. complex (e.g., overwhelmingly those parameters must be determined. empirically designed rather than being based were detection methods to on that much of the the search is considers not only the transmission system, parameters to be measured and the Boolean decision statements based It is to this cases. requires only one-fourth the difficulties, but the memory requirements in the computer the selection of the three lead 95% of over cardiologistc Clearly the argument 12 lead system. in favor of this decision when rhythms in the needs for most routine ECG's serve difficult more of an pulse attempt of the QRS translatable into standard variety of shapes of the QRS series of models for accurate detection CHAPTER II 1. Definition of the The or Electrocardiogram electrocardiogram leads bipolar placed on can be defined the body surface) which results from electrical recording (by unipolar a of the In membrane of the heart muscle cells. polarized (charged) with is said to be cell membrane of about 80 been proposed a cell is neutralized and the cell cell rises from the -80 requires about mv. to .01 seconds. outflux of Na+ ions and the is these action potentials originating electrocardiogram and their beat the process the have across the the charge within the potential begins the cell with the (return to resting superimposition of all in the heart muscle cells. in this thesis T has been wave to deSigned of the electro­ Fig. 2.1 illustrates typical examples of these waveforms to a the to depolarization of separate physiological following sequence of the heart is then results from the analysis described relationship corresponds across theories This rise in mv. analyze the P-wave, QRS complex, ST segment and cardiogram. The the across negative charge within repolarization of potential). Several depolarized. reverse The state the cell resting potential difference the about +20 The The The computer the K+ ions for the transport of ions causes With the influx of Na+ ions membrane. Na+ and (inside negative). mv. explain the to potential difference activity within the heart muscle. activity is generated by the flow of electrical cell as events event. takes a single cell. During place: spreads through node of the atrium. the atrial muscle, causing This normal wave wave heart 1) Depolarization initiated by spontaneous depolarization of cells in the sino-atrial a Each of special pacemaker of depolarization the surface electrodes to record 9 the P wave. 2) As the atrial ventricular junction depolarization delay of a wave .08 seconds about reaches the atriooccurs allowing the atria to empty into the ventricles. R P"R iterv, I I I : segment rS-T rtJ ,. ....__. I I I I I I I I I I I I I I I I I I I QS I complex\ LQ_ QRS interval Figure 2.1 Typical ECG complex This delay is called the PR interval depolarization wave then spreads through contraction of the ventricles. complex of the the R wave; the The electrocardiogram. Q wave, and is if present, normally resulting The first is a iso-electrico the ventricular muscle electrical signal positive going negative wave prior wave 3) The causing is the QRS is called to the R wave 10 and the S on is wave the first occasions there may be (negative 4) the R The final wave repolarization plays vital a and Also, an SI considered in this re­ of the ventricles. QRS complex and the is called the ST segment and wave. (second positive wave) which represents wave following wave wave between the end of the signal wave R' an following R'). wave search is the T The negative start of the T role in proper inter- pretation of the electrocardiogram. 2. Noise Sources In the the on considered. This Electrocardiogram of any electrical analysis is especially signal, superimposed true whn noise must be working with electrocardiograms since noise may be generated, not only from the electronics involved, but also from the 1) baseline drift, 5) noise, and Fig. caused Five patient. 2) 60 of noise cycle noise, 3) noise due to 2.2. sources illustrates patient an were studied in this random muscle project: noise, 4) respiratory movement. ECG with baseline drift. by the instability of the electrodes as they come The slow drift is in contact with the skin. Ii II iii I iT j" . In _ i II Iii I I'll IHI [. i: 1 iUl :r:i :i! I :1 I Figure :;11 I:: Hill W: UI' 2.2 An ECG with baseline drift \. . , ;::1 ii:i 11 After 10-15 minutes the electrode contact will problem. Practically, however, the routine clinical be treated in Fig. 60 Chapter 2.3 shows cycle pickup is electrical devices 60 situation. an electrocardiogram in the modern presently in noise electrical resulting energy spectrum is a with 60 cycle noise Filtering a superimposed. great number of of muscle noise and next section. 203 cycle noise superimposed of muscle cells other than those of the heart muscle also generate electrical the total is not acceptable in delay hospital where use. Figure ECG with 60 Depolarization this Compensation for baseline drift will cycle noise will be dealt with in the will reducng III when the calibration is discussed. common are this time stabilize, potentials. Since the electrodes signal this activity will be from asyncronous throughout depolarization the entire typical example of such a frequency sensed by them. is random, with band of the ECGo contaminated signal. measure a The uniform Figo 2.4 12 - __ I L-- I __ __ :":"::. ---_I-::-: J -- __ .--- --- ---- - .:::t:: -= ::-:'=.:::..-=- Jr::::: == .t== -:- .- =::-_:-:..::._ =-:=:::-:"':__:-=--:::- .. =.::::.:. ::::- ... _. _. l' ":--==.-. =:_: :..:.:. - -.- i. _. ':::= :-:-_ __ . o_ .• - .... - __ o..____. __ . ... - _. .. •• _ • .- --- o_o_ _'·_ _::j _-u\; n::-n;.; ... ::::::---=-:.-.- Figure 2.4 ECG with random muscle noise The final the patient. In both face. two sources Fig. cases 2.5 and movement Respiratory of noise studied result from 2.6 Fig. causes movement is a is manifested wave rate. The noise in sinusoidal movement of examples of this type of noise. regular as a are body interference with the electrode-skin inter- and out with each breath and, as superimposed a movement as the chest moves in Fig. 2.5, the noise result as with frequency equal a seen Fig. 2.6 is generated by random to in respiratory movement of the patient. No attempt is made in any of the research to filter this noise from the signal. In Chapter IV, however, algorithms of movement artifacts to insure accurate elimination of movement artifact is instructs the program. patient discussed for detection rhythm diagnosis. assigned to remain still are during to the the In practice, ECG technician who sampling phase of the 13 1/ Figure ECG with sinusoidal 2.5 respiratory variation i} .......... - H - fTf rl fit ,:;::- 1ft k jitl 4t I =::r - ,-_ 1,;:j:I --j-- ;i""-.-r"".---c.r'--''-''Tr ,:t J i} -t-t It ,_ 4: : 1li "0.# - "-++,t-HI;{"ffiH-H'·TT.,THl V :T- 3 t-': ; rUt ,---- T ,_ +-1"- r-r...-:-rtF f:r-·;r=-=-l::-_:.:_:=r;':::-T:;: j \ I_;t._mtt:,_ L; 'N" ;; f _W 1 ,-;_; ;:;:rjEW 'tj j±H. :.r; ::' :;ii:"''-'+H++-1t-+++-++t-H--H-1; I LERi It-+'b-H-tt-H+',t--H-I--HT>THH-t:i Figure 2.6 ECG 3. Filtering showing patient movement of the ECG As noted, two types of noise filtering techniques. These are can be eliminated by standard digital random muscle noise and 60 cycle noiseo However, before des i qn i ng such fi 1 ters, knowl edge of tr., spectra 1 content of the electrocardiogram spectral density of a is required. Figure 207 depicts typical electrocardioqram. a plot of the 14 Figure 2.7 Spectral density This plot was generated from for 4 minutes. at 500 cps complexes (350 points for each complex per an etc. Since This data complex). which had been sampled then sectioned into individual The raw periodogram was calculated complexes computed. Average periodograms were The generated electrocardiograms with varying heart rates, QRS configurations, the 60 designed ECG was and the average for 50 such The power in all energy to typical a electrocardiogram figure shows the averaged periodogram. for several of cycle cases at 50 cps was down 40 db from the peak power. noise and much of the random muscle noise contribute frequencies higher to attenuate than that, a low pass digital filter that noise with little distortion to the was signal. 15 Design of the low pass filter appropriate SIN(X)/X low pass filter function. expression for the family of SIN(X)/X truncation of the implemented by was curves Equation 2.1 gives the used in the designo Equation 2.1 Y = k 2fc fs • _SI_N (21Tfck) fs , k=o , (21Tfck) fs 1, ... ,N cut off sampling frequency fs These Yk values were Several levels of samples per second. in detail. (filter C) fc The first fc=30 Figures 2.8, 2.9, cps then smoothed using the Hanning smoothing function. studied with and N were After preliminary analysis three filters (filter B) had fc=30 N=17 and the third and 2.10 show the db of each of these filters. frequency fs constant at 200 being cps and N=255. (filter D) plots had for the were studied For the second fc=40 and N=17. frequency spectrum 16 db -120 -140----IO----2O----30----40----50---6O--7O----80----90---I-OOFREQUENCY Figure 2.8 Magnitude of filter B 17 db -IOO--------------------+------------------- o 10 20 40 50 60 FREQUENCY Figure 2.9 Magnitude of filter C 70 80 90 100 18 0,+--------- -20 db -40 -60 -80 -100 -1201..,___--+-----+------+--+----+----+-------i---+----+------+-o 10 20 30 40 SO 60 FREQUENCY Figure Magnitude 2.10 of filter D 70 80 90 100 19 Figures 2.11, 2.12 on varying and 2.13 show the effects of the two 30 cps filters sets of-data. signal with waveform filter C. the noise As as degradation seen B the figures-waveform output of filter in these filter B. of the In each of the as original output of effective in attenuating investigation showed significant QRS complex after having been filtered by filter B, i.e., apparent widening and attenuation of the QRS. in waveform B of the B and waveform C the figures, filter C is However, further A is Fig. 2.13). (See the loss of the notch For this reason, the 40 cps filter Figure 2.11 Result of filters Band C on an ECG with muscle noise was studied. 20 Figure 2.12 Results of filters Band C on Figure Result of filter Band C on an ECG with 60 cycle noise 2.13 an ECG with notched QRS 21 Note in Fig. 3.14 the return of the notch in the R smoothed in Fig. 2.13 Figure Results of filter 0 Again affect clinical decisions. implemented were scaled up by is used in the 106 Following analysis this programs. and truncated for use Table 2.2 gives the two sets of study QRS each of the three distortion which would not the two 17 In each case with fixed The 40 cps filter is used in the the filters has been capable of implemented using filtering in real time. the point filters the filter weights point logic in the weights used. only for rhythm studies where QRS distortion is problem. is resulting from the 40 cps filter shows minimal were programs. 2.14 ECG with notched on Table 2.1 shows the QRS distortion filters. which had been wave by the 30 cps filter. morphological The 30 cps filter not a significant program. Each of digital convolution equation and 22 TABLE I MAX-MIN QRS DURATION Unfiltered 27 87 Filter 31 85 Filter 2 28 86 Filter 3 28 87 COMPARISON OF QRS DISTORTION TABLE I I lQ CYCLE F I L TER 1141 CYCLE FILTER -705 1644 3127 -7796 12614 -26304 -27448 -44413 24591 -46794 133659 82606 249753 293602 300000 400000 249753 293602 133659 82606 24591 -46774 -27448 -44413 -26304 -7796 12614 1644 3127 1141 -705 FILTER WEIGHTS FOR SAMPLING RATE OF 200 SAMPLES/SECOND CHAPTER I I I 1. Introducti The various analysis of ECG morphology involves classifying the shape of the components (P, QRS, ST, and T) into wave significant wi 11 on classes. Clearly, set of a occur. Table 3.1 made by P wave, these gives QRS complex, ST segment, and findings terms of the are C wave T then used to effect gives (findings) list of the classifications a system described in this thesis. the Appendix wave a diagnostic These criteria were Sample 2. I so 1 ate the waveforms 3. Measure selected parameters 4. Perform classification Solutions to each of these to this or leads has flat a a To arrive at the necessary: discussed in this chapter. are Marquette Corporation 3-channel ECG recorder amplifier (or through computer facility similar 3-channel are con­ logic problems interfaced directly CDC 3300 a formulated in in the ECG output of been has are statements the ECG 1. The considered "diagnosis". classification statements the, following steps Sampling are Combinations of independently. the criteria used for the classifications. which The classifications formity with the latest standards used by cardiologists. 2. clinically within anyone class considerable variation at the equipment, a 3-channel voice grade data Latter-day Saints Hospital. the modified available for simultaneous sampling. frequency response from 0-100 cps. (Z inverted) The set) Using Frank X,Y,Z Marquette equipment A similar cutoff frequency 24 Table 3.1 Classifications from Morphology Program QRS Interpretations 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. ST-T Left bund 1 e branc h bloc k QRS criteria for left ventricular hypertrophy Right ventricular hypertrophy QRS criteria for anterior infarction QRS criteria for inferior infarction Right axis deviation Left axis deviation Indetermi nant axi s Right bundle branch block Intraventricular conduction defect, unspecified Questionable Q wave in Y, consider inferior infarction QRS criteria for lateral wall infarction QRS criteria for true posterior infarction Incomplete right bundle branch block Poor R wave progression Interpretation 1. 2. 3. 4. 5. 6. Wave 1. 2. 3. or normal T wave in X, Y, Flat T wave in X, Y, Z Inverted T wave in X, Y, Z Elevated ST in X, Y, Z Digitalis effects ST depression in X, Y, Z Upright Interpretation Broad P wave Abnormal P axis Tall P wave Z 25 is included in the the computer plexor. telephone equipment. through REDCOR 10-bit, a Figure 3.1 is both from patient per lead 20 sec, (i .e., was giving a time of 60 sec total simultaneous sampling a sampling of the 3 take leads). The approximately sample from one signal dynamic range of the the optimize to For the rate of 200 SPS The conversion time between leads is chosen. about 5,000 times to to 10 volt AID converter and multi­ given in Chapter I and Chapter II, reasons interfaced to Latter-day Saints Hospital. at the phasic Screening Facility are Latter-day Saints Hospital and the Multi­ at the rooms signals diagram of the system used for recording block a ± These is signal each lead, amplified at the input AID converter. Since the calibration routine requires MEDLAB (Medical Laboratory Time-Sharing allows only 2048 words of devised which would is done by setting tolerances sample. into a are in set during this assumed to have occurred samples core a the ECG which on lead and The tolerance is difference found the user, sampling strategy had Initiation of the can be used equal a be sampling trigger for two second during at any to one-half the absolute maximum a QRS. this difference is period, and During this operation only that at the end of the be found at the begins sampling The 40 words per lead will sampling phase sufficient can two time, ioe. the present sample and the previous circular buffer of 40 words per lead. QRS as to taking the difference between successive Given this reference difference the program the start of the operating system This tolerance is found in the first two seconds of by sampling the points. System)28,29 satisfy these constraints. subsequent sampling. data to core QRS complexes, and the two data will beginning insure be present such that of the buffer. Once the 26 ECG from the ECG floor from screenino Figure 3.1 Block diagram of the automated ECG system at L.DS. circular buffers are full for the first time the program Hospital looks for two whose absolute difference is greater than the reference difference. as this buffers occurs are rearranged rates down phology so that there to 46 beats are on each lead. 300 continuous soon The circular samples signal required that on each of the start of QRS complexes be in the buffer, analysis of ECG1s with heart per minute is filter described in the buffers are Since calibration of the the three leads. two consecutive samples 260 points the program points As Chapter written to disc. sampled using the same logic guaranteed. II are The weights for the then used to filter the Following this, and moved to disc. two more mor- signal and complexes are 27 On a completion of the data millivolt squarewave one This computer. has signal 200 milliseconds and is sampling rise time of about a repeated during the first seconds. site when the cording 3. Waveform Three by depressing are The are meters measured control precise location on the complex in the at the three critical of these three At this of the other two pOints) re­ are point he may To points. assure (with appropriate marks oscilloscope the memory or feature of the program is sample representative to the technician. QRS's. If they are for that one patient. The vote is made within a more to QRS. one data. incorporated by in which the program decides which the most displayed a subsequent para­ accept the analysis and proceed, view complexes previously sampled, "voting" technique because all the waveforms on start of the points becomes the basis be relative to these displayed namely, verification of the accuracy of location of the quality control of the first two will interpretation, technicians for visual which is sampling switch at the QRS, the end of the first QRS, and the accurate classification of the waveforms is of both searched for in each set of data, for complexes the as the line. on QRS. A a Recognition points of importance (Fig. 3.2) at sampled the last four and stored interrupt an duration of a points searched for averaged second quality and msec operator initiates The appropriate signals the start of the first an is transmitted to the again during are consisting of Each lead is three over four seconds and calibration value for that lead. one per second. once For each lead the differences the ECG and the calibration calibration signal each of the three leads on 50 sps for 8 seconds and the difference maximum a means of the three It is this of a sampled complex by comparing the widths fixed tolerance (± 25 milliseconds), 28 the first finally complex is used. If not, the second and third the first and third. the earlier of the two is If two of the always used. complexes of analysis is aborted. an This procedure isolated premature ventricular example of an ECG showing and meet the criterion If no two should insure complexes compare, against analysis depolarization. Figure An compared This is done to minimize any effects which might be due to baseline drift. the are 3.2 marks locating start and end of QRS 29 The search algorithm proceeds by generating a modified spatial volocity (SV). SV(I) =(X(I) where X, X(I+l))2 - Y, and Z values for the X, Y, and Z leads sampled are (Y(I)-Y(I+l))2+ (Z(I)-Z(I+1))2I=1,2, + The SV array is searched until SV(J) This = 1/2 Max SV (I),I point is assumed is searched in the QRS. representing (; above .e., search of the where point on the ·SV(I)<30 point SV and the same The end is defined next successive four too small Use of the SV not as search only the on optimizes algorithms successful algorithm an apparent late Note that absolute, the end of the QRS using the not QRS. or QRS, logic If the width large ( .22 seconds) tried which searched were to mark the onset onset too same the information available for single one lead could be early and the end late. since the SIN ratio the lead with the patterns Given the end of the first (.05 seconds) on largest signal-to-noise ratio could misleading since it might be isoelectric would result in on QRS, search backward for the start). Previous QRS. quite small, causing start and QRS, especially the start of the second find next leads, but proved be at this Starting ventricular conduction defects. QRS found is either a that as segments may be found analysis aborted. Also, the QRS. on points just preceding that point. search is made to find the ,299. The need for successive failures of the SV results from the fact that short flat of the ... respectively. is found such that point where SV(I) 50 and is also 50 for the that as be to The onset is defined points. a 1,2, = ,299 forward and backward direction for the onset and end of a is true for the three as point SV(J) a ... at either and/or an early the start of end. end of the This QRS complex. relative, values have been used to define the Clearly, these are dependent on sampling rate, 30 the number of bits of the AID converter and Previous limits signal but tried which were did not prove relative to the magnitude of the were The accuracy achieved successful. as amplification of the signal. using absolute limits is the justification for this tactico No a attempt is made search interval is defined based selected parameters will will be described in At this the start and end of the T waves; to find point the signal is calibrated and isoelectric Y(I) M = (X(I) - linear a just prior occurs to is done the onset of straight line between a M, where • 1000 volts lAID VAD is the difference in AID units for the VAD ' 1 millivolt calibration X -X b This interpolation. following equations specify the calibration: The points. b(I)) wave correction introduced to a QRS and assuming that the baseline drift is = which IV. Chapter by measuring the isoelectric point which these two over Search for the onset of the P be measured. compensate for baseline drift using each (see Appendix A) heart rate on instead (I) = X + o signal. _1_0 ·11=0, 1, N-l ... ,N-1 isoelectric value for the first QRS start of the QRS). (average of 4 points just preceding isoelectric values for the second QRS here b (I) represents It should be line drift. Error Fig. that this is only emphasized 3.3 shows in proper baseline of ST and T the linear baseline a case compensations parameters and if these cation of the T wave where and lor ST compensation. an approximation of the drift is causes errors segment will are errors obviously primarily the base- not linear. in measurements sufficiently great, misclassifi­ occur. 31 r l!1 'ill 11111 'I:f ;l l I fir n I I :1 ,i i : H"I n.i ! ,- i : i I II I I 1 r-' i I I : i:j :: ;-:r ! i lJII!i'1 ,i I : i_,_,_ I :: I I I! II r-r-r , ' f: : I::U ! I oJ I i'l! Ii :1 I ,, T'T I II i I:: q I I III i fl 111 IJH fll r Hil iH j lli 1 i !i ii 1 I 11 II Ii ( : i l1 I i il i: 'i II!: r It I 11 i ! i i J ! I I J I ,I "-:r""---:" i ! !Itil -- I ri II] iTI:j: ilF11 , ii : __ i II r i 1"I:_c iW":- iri I]i W Hi! ,: ;"] 11:1 I: i! I I I ; : i !1 ,t : 1\ d-1 1:li i: I :iJJ nl I i 1_ \11i IWI iJI Ilt i i ill S III i n l il ,I I T II H+ r 3.3 Figure An ECG with non-linear baseline drift To prevent these errors a second compensation is made for those in which the heart rate is greater than 75 beats per minute. line is constructed from the beginning of the milliseconds beyond the end of the QRS. T considered relative to that line as as well QRS complex wave Here another point 480 to a deviations will to the ST cases then be segment. (See Appendix B) 4. Measurement of the Parameters Appendix in detail the A the set of gives algorithms have been chosen which clinical decisions. used in are For parameters which on the computer subroutines. would measure and yet example, parameters as a measured and measuring these parameters. easy to vectocardiography require knowledge of the ECG vector forces are of the give rise explains Parameters to the needed used in standard clinical magnitude and angular direction function of time. To calculate these require floating point hardware as well as accurately trigonometric 32 Since goal of this project a produce to was is transferable to many smaller systems not search for produced these having set of of the "V4" The most of this search striking example These parameters. are called In reality it is the X and positive Z axis, again for simplicity The need for done in a vector in .707 lead serves defining parameters. as an (no 1 eads with the but V2, an leads "poor as and in Z, wave R V3 increase a wave and or V4 show a at 450 ECG between the (X+Z). However, not used. example of the type of research ". X), Q in small Tha tis, decrease reappearance in Z lead or the X lead generation This that is, of the "V4" interpretation example a many infarction classi­ local V6. or the R infarcts wave were followed by R wave This would result in the X and Z these deviations from normal can With be detected and made. searching for parameters, study of the clinical logic used in making (or better) being is present in 1 eads disappearance of the indicates the process followed in measurement of was missed is read from the 12 standard being on only the adequate for were large progress i standard showing insufficient change for interpretation of infarction. the proper same R these leads The type of infarction missed. VI Although enough. fications was the looking for criteria for anterior infarction In it became evident that information from not plane V4=sin 45°·X+cos 45°·Z=.707 i.e. the scale factor V4 Z the X, a was parameters since IV4" they approximate information which might be taken from V4. a parameters requiring only fixed point addition, multi­ and division. generation lead capabilities, equally good parameters requiring only simple computer operations a plication real-time system which a a information. parameter which would, an interpretation hopefully, contain the 33 5. Logic for Classification Having selected those parameters formulate logic in all Parameters used: R c. Ratio of the magnitude of the R the magnitude of the S wave in "+" means histograms error (-50 rate of both (high false electrical coincide with standard ECG practices. With these classifications made, output of such and have a to V4 and. example) Levels have been chosen which were positive and false negatives. (high tend to over-read false This will negative rate) the ECG. false exceptions (wide QRS, II in above of the parameters. the results of the program This final II. and or, underread made. in wave V4. (AO)+(B<-50)+(B<-3·(B100XC))) positive rate) some in the Z lead. Ratio of the magnitude of S wave in the Z lead to the magnitude of the R wave in the Z lead. the cutoff levels will as Appendix Infarction" wave Clearly, changing be evident (See given here. b. the cutoff levels or for each interpretationo Magnitude of where would minimize the be to was a. Boolean statements: case developed using statements). the Criteria for Anterior "QRS determined from one of these statements will one was The strategy used classify the waveforms. to used Criteria for In each parameters, logic series of Boolean statements, a An example of B for series of a discussed are axis) the the levels in Chapter In VIc have been chosen diagnoses of Appendix C can to be step is necessary since the clinicians who view the program are accustomed given specific meaning to them. to well For established example, interpretations the modifier "acu te " 34 and "old" have traditional meaning when The in the interpretation patient's file subsequent ECG's on to this and the applied parameters are to stored permit interpretation of patient. an infarction. on serial magnetic disc changes in CHAPTER IV 1. Introduction Computer analysis of In developing included a all possible manageability reasonable. The usefulness of such the of majority relatively common patients small on program be can rhythms. useful a clinical tool. Criteria for Interpretation Rhythm Determinations: 1. Normal 2. Sinus 3. Atrial sinus mechanism arrhythmia fibrillation P not fbund 4. Regular rhythm, 5. Ventricular bigeminy 6. Atrial 7. Three to two A-V Wenckebach waves bigeminy degree AV block 8. First 9. Supraventricular rhythm tachycardia 10. Premature ventricular 11. Supraventricular premature depolarization depolarization on a those For these reasons, rhythms. Table 4.1 for Rhythm great, since be used exhibit list of the rhythms included in the programs. Logic program is By concentrating limited set of a a limited program is still a number of different rhythms the of such which the program will this research has been directed to a rhythms would be impossible. computer rhythm analysis program, limitations must be that both the size and so ECG Table 4.1 is 36 To diagnose these rhythms following decisions the 1. Are there P 2. Are there ventricular waves must be made: preceding each QRS? (point of origin in the ventricles) premature depolarizations? 3. Are there 4. Are the R-R intervals supraventricular (point of orlgln ventricles) premature depolarizations? (intervals not in between QRS the complexes) regular? To these answer questions data is gathered The data from each interval (onset of P wave QRS), to onset of values of these data become the basis of questions and, secondly, 2. on 16 consecutive heart cycle include the R-R interval, QRS duration, to and logic to first the PR of configuration cycles. The QRS. the above answer classify the rhythm. Determination of the R-R intervals Sampling of the ECG is from a single lead Marquette equipment described earlier, by appropriate selection since it generally on the cart. exhibits the subsequent measurements. be used in an intensive Since care any of 15 Normally P waves, largest only unit on at 200 sps. one a lead is possible leads a the Using may be used standard lead 2 is thus the accuracy of aiding required the monitoring ECG lead sampled logic same attached to can the patient. With only one lead being sampled, recognition of the QRS complex is accomplished by measuring difference will difference be a the difference between successive maximum during the first during the QRS. two seconds of Having found sampling, a have a even practically all ventricular premature much smaller first difference due to the a This maximum absolute tolerance one-half this maximum difference is sufficient to flag all complexes, points. equal to subsequent QRS depolarizations which relatively slower rate of 37 depolarization of the ventricles in such instances. Samples conserve from the ECG gathered (to insure are complete QRS). to a Once storage space. a in stored are a 120 word circular buffer to a QRS is detected, 45 additional samples P search sufficiently long The data is then reordered temporary work buffer. region chronologically complete before a the program of While this exists up is monitoring of the 17 seconds to 20 of precise for (18 patient especially an true intensive sampling excess of 170 flagged, completed. is indeed simple, one sampling procedure re­ a heart rate of being superimposed unit where the on the following noise logic is included: unaware sampled points greater than Acceptance of the artifact results in etc. 60), the patient is In many instances difference between successive QRS widths, the if the routine is used for automatic care occurs. rate for which have been patient with movement artifact in a a a studies have shown the heart rates in Because the seconds for patients times when on intervals) R-R signal. is RR intervals, time Timing finding QRS complexes the QRS recognition tolerance. erroneous the on This puts Since the program analysis. classification This great. of noise movement artifact shows the in real subse­ parameter searches must be QRS, the maximum heart a complexes (16 algorithm - possibility signal recognizing is discontinued and final sampling the buffer is ready for operating Once beats/minute. That is, all operate is 266 beats/minute. can capable program quires new 45 values after samples pitfall the program. a Meanwhile, sampling continues at the 200 sps rate within the circular buffer. on as and transferred It is from that work buffer that all quent searches and parameters will be measured. time constraint well as To eliminate these two "QRS complexes" 38 1. The maximum that signal (Sm) is set the maximum value of to equal QRS minus the minimum value of that QRS. (Sm=QRSmax-QRSmin) 2. at the onset of that Starting for .25 seconds minima set Ni = as SM/Ni I successive maxima and (N) 10 for 3 too being then are minima, Imini-maxil· or or more then the Ni complex is noisy. The ratio must be less than may be found within the N. searched backward The values of the noise mini)' [maxi-mini If the value rejected data is to the absolute difference of the maxima and equal i.e., 3. (50 points) locating and (max QRS the 10 at least three times since search such that the ratio is less region which could give than a P wave at least two 10. 1 is that the A second criterion a noise found. Finally, are informed that the Once a if three reached the is too signal QRS complex is found a as in contour the difference is The end is less noisy complexes analysis are two If nonnoisy found before is terminated and the noisy. QRS is not as important for rhythm analysis, the search is made using 1/8 the maximum difference previously found. must be .055 seconds. search is made for the onset and end of Because exact width of the QRS. analysis are RR intervals "clean" operator the must exceed complex is found the parameter search is delayed until QRS complexes 16 QRS width The onset is a tolerance of that point where than that of the tolerance for three successive points. found with similar criteria, except that successive differences below the tolerance for five values. 39 At this point another parameter is extracted from the QRS This parameter is of that designed complex. give information concerning the configuration QRS complex: Config.1 I(QRS up -QRS m d up ) = 1+I(QRSd n -QRS m dd n )1 QRSmdup=median value of the QRSUpls QRSmddn=median value of the QRSdnlS QRSmax=maximum value of the QRS value of the QRS at its onset QRSonset=the value of the QRS QRSmin=minimum parameter will be useful in detecting premature ventricular depolarizations. This 3. to P wave Recogniti on Search for onset of the a P wave The QRS. algorithm assumes or that inverted) (.3 seconds) just preceding a P wave of the is the always present. lip wavell is the point which differs most from the average value of the P search interval. P /max I Pi Pave I - P = ave where the 60 points peak (upright The location of the (Pj) covers 60 60 2:: i=l .L P60=value From this of the P wave. of , i = 1, 2, ... ,60 p. 1 point just preceding point the P wave is the onset of the QRS. searched backward to find the onset (Ps) The onset is located at that point where the P value differs 40 from the average of the first four of the P search interval points than the maximum difference between successive points. This also must be true for the three This will prevent premature termination of found the PR interval Using this algorithm is calculated. a are present" is based the QRS) lip wave" on and regular 4. waves are Recognition of the of P is defined of Premature well as lip the decision Ps)' waves ght or peaks do more (±2 points) P waves to not are Two types of The influenced by outliers than the average. set to equal depolarization is based of the If zero. premature depolarizations flow a the on 1) prematurity depolarization, The former is characterized Fig. 4.1 is prematurity. 16 PR intervals. Depolarizations of premature supraventricular. and QRS onset-index of the median of the as depolarization, 2) configuration waves. If ei than ±.01 seconds more present the PR interval is Recognition of Hith the Ps waveo consistency of the location (with respect median is used since it is less P P biphasic always found, but is the PSG points just preceding present. The PR interval no points during the first four (PR=index peaks of those lip waves". of the differ from each other by assumed the a by less flagged, are its by and 3) presence ventricular configuration as diagram of the logic used in this decision. As can seen be found in the flow only when P diagram supraventricular premature depolarizations waves during rhythms with variable complex is configuration to present. RR intervals designated premature premature (compared in are such as prevents false diagnosis atrial when its RR interval the median from the normal This RR interval). fibrillation. A is at least 28 percent If the complex (QRS duration complex does not not differ greater than .02 41 .02 seconds of median QRS duration complex not the or configuration parameter for that greater than 7/2 of the median configuration) the complex is judged supraventricular. R-R<W% NO prematu,. ecm configuration diff...m from a normal beat R-R<28% pr ... Cltur. ORS duration NO 2: 1I01N P YES wove VES NO NO preeent Figure 4.1 Logic for premature depolarization detection As noted from the flow is flagged primarily on diagram the normal complex prematurity is relaxed (14%), ectopic beats. ventricular premature the basis of abnormal where being a has thus a wide QRS depolarization configuration, the (.11 seconds). allowing recognition of late one exception Note also that or end-diastolic 42 If will seven or check the If the types premature depolarizations more of possibility a recognized,the are bigeminal rhythm (every bigeminy, included: 3) and sequence). 1) 3:3 AV Wenckebach If there 2) ventricular bigeminy, is not the number of premature a (PR atrial depolarizations Three possible (supraventricular) intervals follow bigeminal rhythm present premature)o other beat rhythm is bigeminal then the type will be reportedo are program a short-long the program reports and continues to classify the basic rhythm. 5. Classification of the Rhythm The classification algorithm needs continuing: Is histogram of the RR intervals mature the RR interval complexes). regular? (excluding RR interval or less are added non-zero on The in width. more equal is assessed from about the median RR interval to 16% of the median RR interval is classified rhythm 100 fast. each classification. this regular if there rhythm is then modified by the heart rateo cation of sinus 2) P waves It is example as 1) greater rules for morpholoqy classification logic is the classifi­ of this follows: rate a RR intervals irreqular, present. As will has indeed arrhythmia. An and set of decision Note that in contrast to the mutually exclusive. set is 14% of the median three gives the complete D Appendix 0 are as A rate less than 60 beats per minute is considered slow, than a pre and post intervals from pre­ each side, each of which is Each basic bins. bit of information before Regularity histogram is centered The with the first bin added of width Subsequent bins one be seen in Chapter VI, this simple approach proved useful, but is, of course, subject to rhythm analysis to much criticism. 43 The is major problem is the definitive location of primarily probabilistic probability only one (two P sinus P of wave waves rhythm. algorithms attained. regular is P in nature, waves is complex) It is to these must be directed heart cycle. more are present the problem is that Clearly, a one and 2:1 AV block would be falsely recognized as normal that continued research on wave problems before or approach here The waves. A second great. possible in each for each QRS i.e., if 8 P more precise rhythm analysis can be CHAPTER V 1. Introduction With the contour of the patientls ECG over In many instances serial period of time. a one instant in time, the true meaning of the ECG. these of value, both of time. In particular over such healthy patient a ECGls may be state at For example, T Assessing recent heart attack serial in developed detecting changes taken on a given patient, inversion not present in wave patientls stay serve Computer System. active storage are As patients (C.D.C. such that data for subsequent the program will are 854 disk on the Coronary Care Unit. the confirming no diagnosis as first step a the last two ECGls analysis of rhythm and As disk. on only, (i now .e., imposed by con- used, during a the limited Latter-day Saints Hospital MEDLAB discharged, their files readmission patient with a phases of the disease. only limitation is drives) comparisons. be in on In serves short term nature This capacity available be previous both to confirm the It compares the output of the hospital). in the on-line storage problems ECGls. a can inform the doctor of part of this research changes of assesses changes comparisons ECGls having been previously stored tour for both the program as more long periods (years) changes may in serial us can and to evaluate the state of the heart in the various The program changes, and to 2 early sign of ischemic heart disease. an tell a days) periods (1 short potential problems. a analyzing both the rhythm and ECG, it should next be possible to detect changes in rather than the concerning of programs for development and placed on are removed from magnetic tape. attempt is made to recover Logistic previous With this constraint, the major service of the diagnosis and following the patient in 45 In reporting serial changes diagnostic statements such of anterior infarction". cardiologist "changes as in ECG show These evaluations statements are derived both from the A previous diagnostic a search To avoid the through previous ECGls. compared. comparison data used to morphology contour data previous made only that the label comparisons of the 2. in in that ensure area. comparisons complete E. Fig. This is may have been Comparative of completion medical The record the ECG's on taken within minutes of each more than six hours. from the notes 5.1 shows the date of the have been madeo Criteria for Appendix statements of placed same is present, the program terminates. indicates previous ECG's The change The rhythm been recorded within 15 minutes of the present, the program proceeds but data is appropriate­ rhythm analysis). A time constraint is comparison of ECG's must have on patient's computer other, the ,two ECG's must differ in time by used for the of selected parameters from analysis (contour analysis precedes for results of were show the to analysis. program initiates to be can sequential analysis is automatically requested the rhythm evolution continuing Criteria for these rhythm and/or morphology and the values the contour to make additional presently left are further correlation studies of such statements in this program. ness the until attempt is made no a on If comparisons sequence label. previous analysis with which important no only morphology the report that typical If recording. Note the for correlation since not all processed by the computer. Statements set of criteria for each comparative statement is As with the other programs of this research, been made to set criteria for all possible comparisons which no reported attempt has could be made. 46 NO. NAME 101289 Changes since 11/16/71 Upright T wave present in lead Y Atrial fibrillation no longer present Heart rate has decreased from 107 to 85 Figure 5.1 of Example a serial The statements listed in Appendix E most commonly seen used in lying logic ECG comparison report only those conditions which cover and have the greatest clinical forming those criteria will are The under­ significance. be discussed in this section. With only one exception (heart rate) all comparative regarding rhythm changes are made exclusively statements generated by the rhythm tion about an ECG shows abnormal normal a program. rhythm is reported. is rhythms reported. the basis of the on In each instance, For have been abnormal, then the present". now (60 limits In tions if this beats reporting changes are magnitude reported of the change crossed the per minute for slow and in a in morphology, manner similar to parameters within a In a case state of both abnormalities This convention has been selected in reported only informa­ example, if the previous an the changes in the most useful way to the cliniciano heart rate is diagnostic rhythm and the present ECG shows atrial fibrilla- sinus tion, the message would be "atrial fibrillation where both statements 100 beats attempt to report The actual change in boundaries of normal per minute for fast). the presence and absence of condi­ rhythm changes but the given diagnostic category is change also in 47 mentioned. normal are in both ECG's but theneis (magnitudes wave change. in both both within normal Thus, considered a comparison of parameters is no previous ECG is abnormal in that parameter. or a However, we attempt This is ECG's, magnitude of the T waves reporting or presence (present will reported be decrease) state of the T wave is T as be can inversion is present compared absence of waves to determine if has occurred. myocardial condition the state of a reported, except for T reported (upright, flat is or . morphology statements in reporting serial comparisons, situations. For example, -90 microvolts other criteria on the that criteria based on Q because initial cases are the not used there may appear primarily have to conflicting on another. Given morphology program might have reported in Y, comparing morphology attempts proved criteria in borderline consider inferior infarctiontl• really changed significantly. comparisons as in lead Y had been measured wave change whereas comparing parameters would let the statement parameters This results "questionable Q has wave some suppose the (see Appendix B) Q and occasion and -110 microvolts one the second occasion, Obviously, to wave present) not or statements which should be made. on if T e.g., the progress of present Since both as not be assessing condition for this will in When inverted) limits) or the abnormal where the change in the magnitude of the a (increase particularly important present if the T eample, in its value infarction. waves made unless the report only those changes which to clinically significant, significant change For The paradox statements would indicate not. The precedence. approach This be confusing to the is a taken here is approach was taken cardiologist when 48 reports of "no changes when (based in fact the actual been met 3. II parametri c on di fferences) were interpretations reported different pri nted, criteria had two occasions. on Calculation of Limits for Parameter Changes The selection of limits (Table 5.1) was made after reviewing the reports of several cardiologists. Table 5.1 (in volts) Limits RZ STX ±200 ±100 ±100 ±100 Since their reports as as recorded change in the change of ments by based on the T wave magnitude is two millimeters would be of the parameters first study the program and used normal ±2000 ±200 a was For the limit example, on lead placement, the corresponding A easily noted and reported by the cardiol­ on stored analoq tape on magnetic some to reproduce was initiated. repeatedly processed disk. on measure­ were The second study different days by changes result from finite digitization, and improper calibration. the results of these studies. use (200 microvolts). two millimeters signal, physiologic changes (from respiration, noise decided to the ECG without changes, two studies Clearly ±2000 ±2000 clearly noticeable change in and recorded the results taken different technicians. the in the results patients Sequential Analysis morphology program's ability ECG recorded an in comparisons, it visual the strip chart. on To evaluate the ogist, In ±200 ±200 limits values which would represent ECG to are for Parameters for example), Tables 5.2 and 5.3 show some of 49 Table 5.2 Results of RZ -310 -305 -310 -325 -355 -330 -325 -345 -325 -355 -315 -320 -360 -350 -325 -365 -355 -360 -315 -325 STX STy STZ 22 23 23 31 38 24 30 31 29 35 19 31 35 32 22 39 31 29 23 42 51 51 53 57 43 43 40 43 37 37 31 48 44 43 40 52 33 42 46 43 32 34 42 45 30 53 50 41 53 46 52 46 46 46 51 38 47 43 52 54 Repeated Analysis of Same ECG TX Ty TZ TAX TAy TAZ 195 195 195 185 195 190 185 195 185 200 195 190 205 275 195 195 200 195 195 205 120 120 185 185 185 195 195 190 195 190 200 210 185 190 210 265 175 195 200 210 175 210 2795 2745 2760 2675 2695 2820 2625 2835 2630 2935 2780 2665 2995 3145 2745 2710 2940 2815 2860 3035 1675 1670 1780 1615 1640 1750 1515 1545 1540 1610 1715 1575 1645 1415 1755 1620 1675 1625 1670 1435 2650 2705 2675 2895 2770 2820 2895 2665 2895 3055 2655 2820 2995 3000 2650 2790 2950 3020 2540 3045 125 110 115 125 115 110 110 110 120 110 110 125 125 110 115 115 115 105 Mean Absolute Difference: 20.36 7.47 5.74 7.84 13.68 6.05 18.16 156.32 113.68 190.79 6.45 21. 57 5.52 22.08 105012 89.62 143.59 Standard Deviation: 13.62 The 5.64 means 4.36 and standard deviations of the absolute differences between successive measurements As seen from Table 5.2, limits for that however, there to reduce noise no parameter was reported in are need to the bottom lines of the tablesc difference in any of the parameters exceeds as analog superimposed in Table 5.1. given on In processing filter the data before the signal the the data, sampling from the tape recorder. in order This 50 Table 5.3 Results of Repeated Analysis from Same Patient RZ STX -338 27 43 123 33 55 18 37* 27 66 39 -319 -430 -275 -302 -296 -828* -369 -413 -582 STy STZ 24 15 12 10 30 6 82* 37 26 27 52 52 9 29 20 25 52* 27 42 53 TX Ty TZ TAX TAy TAZ 200 176 282 160 236 196 262* 205 291 243 181 172 202 175 245 159 2780 2376 4089 2096 3170 2638 3606* 2819 3974 3178 2226 2202 2456 2207 3293 1971 5910* 2847 2571 2787 2160 2342 1679 1889 1960 1923 3366* 1818 2640 2963 1046.89 1181.00 610.33 167 124 139 147 136 288* 137 187 219 142 445* 224 194 218 Mean Absolute Difference: 21. 22 169. 11 37. 78 17.22 29.44 90.78 52.22 12.41 29.77 97.29 54.98 Standard Deviation: 23.43 183.87 26.82 resulted in an 490.62 552.07 1357.86 essentially noise free" signal, but allowed for measurement of changes in the parameters primarily resulting from finite samplingo is, the magnitude of the digitized. No attempt will waves was made to vary depending digitize the on same what beat, point thus That is actually some inter­ beat variance is also reflected in the data of Table 5.3 The results for one patient (Table 5.3) show instances in which the that absolute difference between successive ECG's exceed the limit for parameter. In all Those instances parameters, however, falls well within its limit. significant changes occur the on mean both sides of the asterisked valueso absolute difference for that parameter Note that there before and after some are always particular two instances of measurement (in all 51 probability the measurement is in discussed in Chapter VI with the clinical example of Table 5.3 it was indicating that one error set of measurements, parameters. results of the program. In the occurred in improper calibration. that indeed the program has measurement of phenomenom will be further felt after reviewing both the tracings of data and other parameters that the Excluding This error}. all other ECG's compare well, sufficient reproducibility in the CHAPTER VI 1. Evaluation of Contour and Rhythm Program Evaluation of the contour and rhythm analysis programs was conducted by comparing the interpretations reported by the programs with those made by one or several here. cardiologists. more comparative evaluations Additional at the in-patients were was the development of the made; only the last be conducted the to hospital (150 ECGls). No The the last 150 ECGls taken at the time of the area active in the admission area) for five reported improvements were are made. taken both on (52 ECGls), and selection of these special 150 taken in-patients on study, while those from consisted of the first patients each day the admission programs one Latter-day Saints Hospital made for inclusion in the study. period as The ECGls used in this evaluation patients being admitted ECGls were evaluations will (see Chapter VII) on During (the days before the least start of the evaluation. Of the 202 ECGis, 72 preted as a "norma 1 ECG II were if left axis deviation, set: abnormally tall R in a normal the contour fi nd i ngs 11 right The ECG is and 130 abnormal. axis deviation, are inter- from the fo 11 owi ng indeterminate axis, Z, flat T in Z, and the rhythm is a "normal sinus mechani sm". Six normals givng an emphasize abnormal 8% false the were reported positive and a abnormal 2% false and three abnormals negative rate. ECGls for further evaluation. Even as normal 9 These results screening aspect of the program in attempting this study, other studies also rates. as to flag all though these percentages reflect give similar false positive and false negative 53 Table 6.1 lists the abnormal these three cases relatively were possible lead problems as the ECG's reported minor in cause cardiologist these were present by reported 40 ECGls were and of the which where the computer. abnormal the (20%) and either cardiologist was or error in two showed Table 6.2 lists the error. ECGls. discrepancy existed between the some the nine considered above, 31 of Excluding some The degree and the last abnormalities falsely reported in the six normal There normal. as computer finding he felt that some not considered was finding should have been not. Table 6.1 Abnormal ECG's Interpreted by Computer questionable Q T ST wave wave analysiso In found that 14 of 226 the cardiologist. width and the error seen in not depression statements comparison of possible interpretations (6%) The major problems appear 4 QQls, Since the to with 17 analysis of wave 176 cases disagreement with Similar accuracy is disagreements P in evidenced by the as IVCD's. out of 223 morphology Table 6.5 reports only the recognition of abnormal out of the were be the measurement of QRS in lead Y, 2 RBBB's and 2 interpretations only 2 instances (1%) the various components reviewing the results of the QRS interpretations, it is reporting (6%)0 in Y in lead X significance of small Qls in the ST-T Normal inversion in lead Tables 6.3 to 6.6 show the detailed of the as vs. where the is normal possible quite limited, P In waves. computer detected P waves 54 did disacreement a rhythm analysis. (5%) was when not signal they Table 6.6 gives the results of the The computer failed to recognize the P were saved, at the Finally, occur. one Since the present. analog only speculate about the can time of analysis. Because of the in waves 11 cases data from these patients amount of noise frequency on of this the occurrence the program will henceforth attach the message "probable sinus mechanism" to those ECGls with "regular rhythm, other than was P made in this (rhythm or to evaluate the accuracy of study recognition since not found" when wave lIabsent" P-wave is detected the actual no abnormality contour). No attempt premature depolarization analog recording during the sampling was un­ available; thus it would have been impossible to know if, in fact, there had been one that even all the 17 beats used for during though the 202 ECGls fell rhythm analysis. rhythm program has into one of those limited set of a It should be noted interpretations, categories capable of being recognized by the computero Table 6.2 ABNORMALITIES REPORTED IN NORMAL ECGIS One QRS criteria for anterior infarction Two non-specific ST non-specific T Two One abnormal In waveo evaluating these statistics, first is that tracing. P This no ECG was was changes. changes. two points need to be remembered. excluded from the study because of a poor The quality felt necessary because the decision to accept the computer interpretation had already been made by the technician and the report had 55 Table 6.3 QRS INTERPRETATIONS Computer N L R V H A 0 L A 0 P A I Q Q 0 0 R L B R B I V B B B B C 0 T U 0 A R A L L R , I C r d i 0 1 1 1 RAO 1 LAD 1 1 1 00 4 1 14 12 II 0 s 127 LVH a g ; N 6 AI 18 f OORR 4 2 LBBB t s RBBB 2 IVCO 2 1 5 4 12 UOA 6 - ALLR N LVH RAO LAD QQ II AI POORR LBBB RBBB IV CD UOA TALLR - - - - - - - - - - - - - Normal Left ventricular hypertrophy Right axis deviation Left axis deviation Questionable Q wave in lead Y Inferior infarction Anterior infarction Poor R wave progression Left bundle branch block Right bundle branch block Intraventricular conduction defect Indeterminate axis unspecified Abnormal R wave in Z 56 Table' 6.4 ST-T INTERPRETATIONS Computer N C ISC N-T 4 N 113 1 r ISC 1 22 d i N-T 3 N-ST 3 N-ST 2 DIG 1 a 3.3 o 1 31 1 o DIG g i 7 1 s t Legend N ISH N-T N-ST DIG Normal Ischemia - - Non-specific T wave changes Non-specific ST changes Digitalis effect - - - Table 6.5 P WAVES Computer N A C A 4 N 2 r d 170 i 0 1 Legend 0 g i Norma 1 N P wave s t A - Abnormal P wave 57 Table 6.6 RHYTHM Computer SM AFIB SA RR SUPRA C a r d i SM AFIB 173 11 10 o 186 1 o g i s t 2 14 3 SA RR S---+-----4-----4-----*-------------------- IsuPRAv SM AFI§ Nl 10 SA RR SUPRAV - - - - - - - Sinus mechanism Atrial fibrillation Not first degree block First degree AV block Sinus arrhythmia P wave not recognized Regular rhythm Supra-ventricular tachycardia - 58 gone to the patient's chart. obvious that some of the However, in reviewing the tracings, it errors reported resulted signals. The second deviating standard with which the computer The point is the difficulty in finding a performed at the of two Latter-day cardiologists, ment between Since the entirely it was Hospital found that there performance for evaluation purposes to in the sequence had been of the was analysis, error. use 0 study interpretation area of disagree­ it only pairs have in either errors some to was those cases, the decided tracings It should by the computer. that this could result in 40% of the a program relies of ECG's in which both rhythm or con­ sequential reports being analysis of the data study of the frequency of the type of changes occurring in rated as follows: The changes L No 2. Minor 3. Major changes (change in diagnostic statement) were change changes (no change in diagnostic statement) pairs of ECG's, and change, 22 showed minor changes lists the changes which primarily major a contour programs, properly interpreted The test group consisted of 50 no was compared interpre­ a rhythm and By limiting the study limited to the ECG. compare the the in example, sequential analysis be noted that if 20% of the ECG's tour were be Sequential Analysis Program the results of the on to For can un­ them in 8% of the ECG's. Evaluation of the 2. Saints exact and an interpretations different group of cardiologists. "noisyll of analysis percentage reported here could vary if the standard tation of in from was in the ST-T were 6 showed Within this group 22 showed major changes. considered major changeso segment with no real change in the Minor Table 6.7 changes diagnostic were state 59 of the patient. Table 607 CHANGES DETECTED IN SIX PATIENTS WHICH WERE OF MAJOR VALUE One patient changed from normal fibrillation. Two patients developed first degree AV block. One patient One sinus mechanism to atrial changed from normal ECG to abnormal patient changed from atrial fibrillation 0 to normal sinus mechanism. One heart rate decreased to within normal limits and premature ventricular depolarizations were no longer patient's present. The program could from the suffers has been tour program writes from a inaccuracy of those programs implemented On analysiso a to aid in the quality as well. control message on indicating the scope on that patiento quality signal, action to the is taken computer for to a improve new the but contour, Thus, the program of the if any rhythm and the changes con­ sequential are present If so, the technician is verify the quality of the present signalo again or completion of the ECG analysis programs, previous ECG recorded asked to mitted readily reflect changes in rhythm If he notes quality interpretation. and the a poor ECG is trans­ CHAPTER VIr Significance In of this Research the assessing significance this, the most important question that task for which it has been in Chapter VI indicate the of goal of this the can continued evaluation algorithms interaction with in actual factors Completeness in this parameters" used by "parameters" It is ized program reading ECGls as a high level a perfect and still requires criticism holds for any point, however, changes in order that there be can or distinguish this research from other context the are are means best defined that the cardiologist has rhythm, expected to aid to incorporated changes environment. in into the programs. cardiologist that over a a period computer­ of the task of non-cardiologist physician seeking Anyone of these programs by itself For programs "decision in the ECG set of decisions relieve the the remote set of complete been contour and only with this complete be Future "completeness". as solution to part of the problem, with limited the clinical no discussed in the next section. The first is specialized consultation. only cautiously accomplish to degradation of the performance rather than improvement. important of time. results described "unrelated" components of the program which might result automated ECG analysis. These same At this analysis. must be made improvements in the programs Two This as does it accomplish the program has reached by cardiologists. ECG project such present procrams be said that it is far from other program for automated in the basic How well The clinical of the Although research applied an designed? ability research. reliability, it of must be: example, the contour or information about the ECG. by themselves only partial can practical be viewed value in rhythm analysis give Preliminary attempts 61 to relieve the all it cardiologist Latter-day Saints Hospital of reviewing ECGls have shown the importance of sequential was both found that morphology patient had though even and previous ECG which had a significance the programs and parameters for contour filtering specific most Emphasis frequently the final analysis left been is based few placed favorably with those puterized ECG to only the The algorithms Digital point logic. weights optimized recognizing those on possible. simplistic approach in the research. for this conditions Where specific on its results of rhythm com- analysis and contour or reported by analysis develop something which a Regional in the United can in reported to all in Medical States.21 Chapter Program compare survey of Because of this applied research since modularity of the program. VI community at The programs a com­ simplicity, sizes of computer system is aid the medical has Although simplicity practical implementation. particular advantage the Finally, the logic for cardiologist. transferability of these programs feature is A possible conditionso built into the program, the results This has programs would be parameters, and by reducing the storage requirements for this increases analysis, has been hospital difficult, only abnormal parameters have been reported and are relatively ECGls taken in the "simplicity". minimum set of a reported for It also showed that analysis require only fixed rather than all seen, diagnoses paring ECGls developed accomplished using task. is study interpreted by the computer, complete evaluation of the The second factor of is not been were saved little effort if the of these programs for all would be necessary before throughout interpretations was In this analysis. compared with the present ECG. complete implementation seen accurate rhythm the cardiologist and could thus not be is at the practical. major goal is large. developed Another on the to 62 MEDLAB clinical 2048 words of system at Latter-day Saints Hospital core area inherent modular In view of the systems, the modules pendently the as significant result implemented clinically and This attests to the of transfer of was used between to other logic separately and implemented inde­ of this research is that all in are daily operation programs have at three hospitals. and usefulness of the research, but also practicality that continuroevaluation and assures to the program repository for communication desirability be considered can design of each other. A final been an magnetic discs serving the modules. allowed only memory at any instant of time for both data and programs. To fit within this with the were improvement of the programs will be accomplished. Future of this Research As mentioned for all design criteria This concept, as well the as The second is the decision the electrical coupled should with noise is more greatly one emphasis reduce this logic. solution to the on the source error. refinements should be considered, however. quality a become too strict, computerized the time ECG might become quality required impractical. set of pro­ of the transmitted More accurate measurement of quality problem. IIvotingll technique of present Two factors influence the of research. The first is the that the possible interpretations. rate inherent in the error course rate of the program. signals. not intended was of the programs include all grams, dictate the future error it in this thesis, previously A approach, described in Chapter III practical If the This limitation of these requirements for for the technician to signal perform 63 Improving the criteria for decision-making for parameters which further enhance the This search will is that measured investigate the ST segment in either all or city in recognizing minor ST abnormalities. and Z lead to the standard the Q new rhythm analysis is algorithms F for such more specifi­ type of parameter is approximate a parameter from AVF). or an algorithm) rhythm continuous The While these then be can wave recognized. be used in recognition. wave. New (See Appendix an With increased intensive care reliability environment for monitoring of the ECG. comparative programs will, of criteria for as P optimize P As confidence is increased in this area, other Block) program could analysis programs. well AV accurate detection of the more need to be tried to rhythms (e.g., 2:1 as allowing A second in leads 3 wave the slope of as parameters should aid in the morphological analysis the' key to improve­ ment in the (e.g., Z leads such of those leads, thus of the program. One type of parameter parameters. or that generated from the X, Y, 12 leads, involve searching diagnostic ability two types of from the X, Y, directly will to The diagnostic logical course statements to course, of investigation here is to develop reflect the progress of the patient confirm the interpretation. needed into the set of improve in parallel with the Also, further investigation is comparisons which should be made. 64 BIBLIOGRAPHY 1. Bonner, R. E. and Schwetman, H. D.: Computer Diagnoses of Electro­ cardiograms. Ir. A Computer Program for EKG Measurements. Comput. Bomed. 2. Bonner, 3. Bootama, B. K., lloe l E. R. 1:366 (1968). and Schwetman, H. D.: Computer Diagnoses of Electro­ R: cardlograms. III. A Computer Program for Arrhythmia Diagnosis. Bomed. R. 1:387 (1968). Comput. en A. J., Strackes, J. and Mejder, F. L.: of R-R Intervals in Patients with Atrial Fibrillation at Rest and During Exercise. C. 41:783 (1970). , Analysis 4. Cady, L. 5. Caceres, C. A., Steinberg, C. Ao, Abraham, S., Carberg, W. J., McBride, J. M., Tolles, W. Eo, and Rikle, A. E.: Computer D. Jr., Woodbury, M. A., Tick, L. Jo and Gaitler, M. M.: A Method for Electrocardiogram Wavepattern Recognition. Example: Left Ventricular Hypertrophy. C R. 9:1078 (1961). Extraction of 6. E1ectrocardiographic C. Performance of the Caceres, C. A., and Hochberg, H. M.: Physician in the Analysis 79:439 (1970). Parameters. of the Computer and Am. Electrocardiogram. Hecur;t J. 7. Ctinial Eleodiogaphy and Caceres, C. A., Dreifus, L. S.: Comput. Academic Press, New York 1970. 8. Cox, J. R., Nolle, Fe M., Fozzard, H. A., and Oliver, G. Co: Preprocessing Program for Real-Time TMM. Bomedo big. 15: 128· (1968). a 9. 10. R., Fozzard, H. A., Nolle, F. M., and Oliver, G. Data Transformation Useful in Electrocardiography. Bomed. R., (R. [IJ. Stacy and B. 'D. Waxman, Eds.), 1969. Academic Press, ;ew York C.: IEEE Some Comput. and Vol. III, Debrow, R. J., Fieldman, A., Clason, Wo Po C., Gorman, P. A., Reinfrank, Transmission of Electrocardiograms from R. F., Caceres, C. A.: Draper, H. Am. Community Hospital for Remote Computer Analysis. Cdiot. 21: 687 06 W., Peffer, C. Jo, Stallmann, Fa W., Littman, D., Pipberger, H. V.: Eddleman, J. (1968). The Corrected Orthogonal Electrocardiogram cardiogram in 519 Normal 12. AZTEC, rhythm Analysis. Cox, J. a 11. ECG (1962). 25:356 E. Men. Cine E., and Pipberger, H. V.: 30:853 and Vector­ (1964). Computer Analysis of the in 1,002 Orthogonal Electrocardiogram and Vectorcardiogram Am. f-Ie.cur;t J. 81 :608 (1971). Patients with Myocardial Infarction. 65 13. Feldman, C. L., Amazeen, P. G., Klein, M. D., and Town, B: Computer Detection of Ventricular Ectopic Beats. Comput. B1omed. R. 3:666 (1970). 14. Franke, E. K., 8raunstein, J. F., and Zillner, D. C.: Study of High Frequency Components in Electrocardiograms by Power Spectrum Analysis. CVz_c.. R. 10:870 (1962). 15. Frank _E.: An Accurate, Clinically Practical System for Spatial Vectocardiography. 16. Geddes, J. S., and Warner, H. B1omed. 17. CVz_c..13:737 R. R.: (1956). A PVC Detection Program. Gersch, W., Eddy, D. M., and Dong, E.: A Heart-Beat Interval R. 4:J85 (1970). - Cardiac Arrhythmia Classification: Markov Chain Approach. Comput. B1omed. Klingeman, J. and Pipberger, H. V.: Computer Classifications Electrocardiograms. Comput. B1omed. R. 1:1 (1967). 19. of Langner, P. H. Jr., Geselowitz, D. V., and Mansure, F. T.: High Frequency Components in the Electrocardiograms of Normal Subjects Am. He J. 62: and of Patients with Coronary Heart Disease. 746 20. Comput. (1971). 4:493 (1961). McConahay, D. R., McCallister, B. G., Hallermann, F. J., and Smith, R. Eo: Comparative Quantitative Analysis of the Electrocardiogram and the Correlations with the Coronary Arteriogram. Vectorcardiogram: CVz_c.. 42:245 (1970). 21. Computer National Advony Coun on Reg10nal Med1c.al Pnogn: Report Assisted ECG Analysis and the Regional Medical Programs. August 3, 1972. 22. Nolle, F. M., olff, G. A., Cox, J. R., and Ambox, H. Oliver, G. C., Detection of Premature Ventricular Contractions with a Clinical System for Monitoring Electrocardiographic Rhythms. B1omed. 23. R. 4:523 D.: Comput. (1972). C. K., Fallowes, L., and Pordy, L., Jaffe, H., Chesky, K., Friedberg, IV. Computer Diagnosis of Electrocardiograms. Bonner, R. E.: A Computer Program for Contour Analysis with Clinical Results Compute B10mede R. of Rhythm and Contour Interpretation. 1 :408 (1968). V., Bialk S. M., Perl?ff, ,Jo x., and Schnaper, H. Wo: Correlation of Cllnlcal Informatlon ln the Standard 12-Lead ECG Am. He J. 61:34 (1961). and in a Connected Orthogonal 3-Lead ECGo 24. Pipberger, H. 25. Pipberger, Use of Computers H. V.: ReA. 11: 555 (1962). co«. in Electrocardiogram Interpretationo 66 26. Pipberger, H. V., and Stallucamp, F. W.: Computation of Differential Diagnosis in Electrocardiography. Ann. 06 N.Y. AQad. S. 115: 1115 (1961). 27. Pryor, T. A.: R. 28. Pryor, 29. Pryor, A Note On 4:542 (1971). Filtering Electrocardiograms. T. A., and Warner, H. R.: Datamation 54:23 (1968). T. A., Gardner, R. G., and Day, C. W.: Research and Clinical 33:809 (1968). 30. Time-Sharing Pryor, T. A., Medical and Application Comput. Bomed. in Biomedical Research. Computer System for AFIPS Conn. PnOQ. to Medicine. Warner, H. R.: Diagnosis. PnOQ. Some Approaches to Computerized (In PubliQon). 06 DagnotiQ PnoQ 31. Pryor, T. A., Russell, R., Budkin, A., and Price, W.G.: Electrocardio­ graphic Interpretation by Computer. Comput. Biomed. R. 2:537 (1969). 32. Rey, W., Laird, J.D., and Hugenholtz, P. G.: Digital Computer. Comput. Bomed. R. 33. Stallman, F .t1., Pipberger, H. V : Automatic Recognition of Electro­ cardiographic Waves by Digital Computers. CQ. R. 9:1138 (1961). 34. Specht, 35. Specht, 36. Simborg, Classification J. E., and Toole, J. G.: D. F., 'tanqe l sdor-f of Electrocardiograms Using the Polynomial Discriminant Method of Bomed. S. Itnum. 4:132 (1968). Pattern Classification. , D. F.: Vectocardiographic Diagnosis Using the Polynomial IEEE Tnae on Bomedo Discriminant Method of Pattern Recognition. Eng. BME-14:90 (1967). 39. Histogram. JAMA 197: (1966). Smith, Ro A E., and Hyde, C. M.: Computer System for Electrocardio­ ROQky Mountain Bo-Engneng Thind Annual Recognition of Myocardial Sotabata, I., Rickman, H., and Simonson, E.: Amo Health J. 79: Infarction by Means of Frank Lead Q/R Ratios. 508 (1970). Stark L., Dickson, J. Time Diagnosis System. 40. The R-R Interval W., and Ross, R. S.: D. graphic Analysis. PnOQ. Sympoum, May, 1966. 38. by .. 867 37. P-Wave Detection 4:509 (1971). Ann. F., Whipple, G. H., Horibe, H.: of Clinical 06 N. Y. AQad. Remote Real­ Electrocardiograms by a Diqital Computer 06 S. 126:851 (1965). J.: Computer Proqr am for \artak, J., Milliken, J. A., and Karchman o--tomed R. Diagnostic Evaluation of Electrocardlograms. Computo •. 4:225 (1971). 67 41. Wartak, J.: Computers in Electrocardiography. Springfield, IllinQis, (1970). 42. Whiteman, J. R., Gorman, P. A., Calatayud, J. B., Abraham, S., Weibner, A. L., and Caceres, C. A.: Automation of Electro­ cardiographic Diagnostic Criteria. JAMA 200:932 (1967). 43. Zeitlen, R. A., Cady, L. D., Tick, L. J., and Woodbury, M. A.: Combined Analog-Digital Processing of Cardiograms. n. 06 N.Y. 44. Ac.ad. Suo 115:1106 . (1964). Probability. Japane Cc.ulation Amazeen, P. G., Moruzyi, R. L., and Fe 1 dman, C. L.: of R Wave in Noisy Eng. BME 19:63. 46. Thomas Yasui, S., Yokoi, M., Watanabe, Y., Nishyima, K., Azuma, S., Aihara, N., Okamoto, N., Fujino, T., Okayima, M., Miyuno, Y., and Yamado, K.: Computer Diagnosis of Electrocardiograms by Means of the Joint 32: 517 {1968). 45. Charles C. Electrocardiograms. Jonal Phase Detecti on on Bomed. IEEE Tan6. Information Transmission Modulation and Noise. Schwartz, Mischa: Chapter 6, Section 6-4, p. 410, McGraw-Hill (1970). APPENDIX A Mathematical Definitions of Parameters Used Contour Analysis Q+7 1. XSUM E = i=Q X.; where Q is the index of the onset of QRS complex. 1 Q+7 2. YSUM E = i=Q Y. 1 Q+7 3. ZSUM E = i=Q 4. XMAX = 5. XMIN = 6. YMAX = 7. YMIN 8. max Z. 1 {Xi' i=Q, Q+1, ,S-6}; where S is the index of the ... QRS complex. min {Xi' i=Q, Q+1, max {Y i = min ZMAX = 9. ZMIN = 10. A nq 1 - e - ,S} ... =Q, Q+ 1, ... {Vi' i=Q, Q+1, ... max {Zi; i=Q, Q+1,,, min {Zi' i=Q, Q+1, i' YVAL. where YVAL XVAL' 0 ... == ,S-6} ,S} ,S-6} ,S} YMAX if similarly XVALXMAX if 11. VMAX == max {Xi+Zi, IYMAXI>IYtIN IXMAXIIXMINI, i=Q, Q+l,.,o,S-6} , else YVAL=YMIN, - else XVAL=XMIN 69 12. VMIN 130 XQWAV = 14. VQWAV = min = {Xi+Zi, i=Q, Q+1, ,5} ... min {Xi' l=Q, Q+1, ... ,Q+7} min {Yi, i=Q, Q+1, ... ,Q+7} if minimum value of the Q where j is the-fi- wave. rst index no Q If YilOO wave such that present, else it is for any i Y{O, then before j, no Q is present. min 15. RPRIlE = 16. SWAV max 17. XRATIO 18. VRATIO 19. ZRATIO 20. Duration 21. XQMAX = 22. YQDUR = = = - _= {Zs-Zi' {Xs-Xi, i=5-6, 5-5, i=S-6,S-5, ... ... ,5} ,S} XMAX·10 XMIN VMAX·100 VMIN Z10 (S-Q)·5 Xi such t h a t X i Duration of Q 2 +Y wave 2. i lS in Y a max for during the Q wave in Y 70 23. ZRMAX = (RZ-Q)·5; 24. ZRDUR = (RZND-Q)·5; 25. STX = wher-e RZ is index of ZMAX S+7+(ST-1) 1 • ST X. L: i=S+7 , ' interval 26. STY = 27. STZ = where RZND is index of end of R between 2 QRS L: ( 1 + = ( ;=5+5T!1 Y. L: + ' S+ST+2 Z. L 5+11 ( Z. L ' i=S+8 31. TMAXX = max {Xi' + . 33. = min {X i' i S + S T + 5 TMAXY = max {Yi, = )/2 i=S+ST!l , i=S+ST+5, . . . ... , L Yi 5T +3 2 ST +6 2- S+ i=5+5T+5, S+ST+6, TM I N X i=S+ - L Zi 1=S+ ... {71,3·RR/4-DURATION} 32 - TEN 0 } ,TEND} +3 ST +6 2 S+ v'1)/2 - Xi T ;=5+ L: i=S+ST-1 +6 2 L: S=ST+2 i=S+8 = X.)/2- L: ST S+ S+ST+2 X. L: S+l1 CONCAVZ (Q2-Q1) complexes lead Z and RR is i=S+7 ;=5+8 30. DURATION)/2 - S+7+(ST-1) T CONCAVY IRR·SO iS+7Yi S+ll 29. X in S+7+(ST-1) 1 ST CONCAVX= 28. 9 ST=(350 where wave 5T +3 2 ,TEND} where TEND=MIN 71 34. TMINY = 35. TMAYZ = 36 TM I N Z = . min {Vi' i=S+ST+5, ... ,TEND} max {Zi' i=S+ST+5, ... ,TEND} min {Z i' i S + S T + 5 . , = , . . TEN 0 } T+8 37. TAREAX = where T is the index of TMAXX if X., i=T18 ITMAXXIITMINXI, else T is the index of TMINX T+8 38. TAREAY = y. ; where T is the index of TMAXY if i=T18 ITMAXYIITMINYI, else T is the index of TMINY T+8 39. TAREAZ = Zi' where T has similar definition a as in 37 and 38 i=T-8 40. XPlAV = Xi such that Xi is max {IXi I, i=SRST, SRST+l,o.o,Q-8} where SRST=MAX {O, 60-RR/3} 41. YPWAV = Yi such that Yi is max {IYil, i=SRST, SRST+l, 42. ZPWAV = Zi such that Zi is max {I Zi I, i=SRST, SRST+l, 43. POUR = (P-PND)·5 where P is index of end of P index of onset of P wave wave ... ,Q-8} ... ,Q-8} and PNO is APPENDIX B CRITERIA FOR INTERPRETATIONS FOR MORPHOLOGY FINDINGS Criteria for QRS Interpretation 1. Left Bundle Branch Block A. The duration of the QRS must be equal to or greater than 120 milliseconds (DURATION 120 milliseconds); and B. The absence of an R wave in Lead Z; (ZSUM C. The absence of an S wave in Lead X. (SWAVE Note: 2. In the presence of a "Left Bundle interpretation is evaluated. 25 and microvolts) Branch Block" no other magnitude squared of the S wave in Lead Z plus the magnitude squared of the R wave in Lead X plus the magnitude squared of the R wave in Lead Y must be greater than five million microvolts squared. The ZMIN2 4. < microvolts) QRS Criteria for Left Ventricular Hypertrophy A. 3. 200 Ri XMAX2 + + YMAX2 > 5xl06 microvolts squared ght Ventri cul ar Hypertrophy A. ZRATIO> -10 and B. XRATIO> -10 QRS Criteria A. The R B. ZRATIO for Anterior wave < (ZMAX) is 1 Infarction ess than or equa 1 to; or -30 and ZRA TIO .2 -500 or ZRATIO C. < VRATIO/I0 Not criteria for Left Ventricular In Hypertrophy Block the sum o the the presence of Right Bundle Branch .04 seconds lss than -500 mlcrovolts < -500 mlcrovolts) voltages during the first (ZSUM is sufficient criteria. 73 5. 6 . QRS Criteri a for Inferi or Infarcti on A. A Q B. YQDUR -50 microvolts C. XQMAX Rig h t A. wave Ax i s in lead Y A. s Devi ati 8. 0, YVAL and +1800 0) plane between -140 and -1800. < 0) s < 0, YVAL < plane between +1800 and +270°. 0) Bundle Branch Block QRS duration greater than (DURATION 120 or equal milliseconds); to 120 milliseconds and The presence of the R pri me in Lead Z (RPRIME <-25 microvolts) Note: 10. +1030 on An axis in the frontal Right B. and Absence of riteria for inferior infarction. (XVAL A. < between '-1/4, wi th XVAL.:_ 0, YVAL Indetermi nant Axi A. 9. < plane with XVAL An axis in the frontal (ANGLE B. microvolts); De via t ion An axis in the frontal Left Axi -200 -50 microvolts (ANGLE<-1/4, 7. (YQWAV In the presence of a Right Bundle Branch Block only criteria for Axis, Inferior Infarction and Anterior Infarction are evaluated. Intraventricular Conduction Defect, Unspecified A. QRS duration greater than 120 milliseconds; and B. Not a Note: right bundle branch block or left bundle branch block Only criteria for inferior infarction is evaluated in the presence of "Intraventricular Conduction Defect, Unspecified". 74 11. Questionable Q A Q Bo The ratio of the R-wave to the Q-wave in Y is greater than or equal to -4 wave in Y such that Y QWAV < -100 microvolts and not sufficient criteria for inferior infarction; and YMAX -4) QRS Criteria for Lateral Wall Infarction A. The XSUM B. The absolute value of the Q ave in Lead X greater than the magnitude of the R wave in Lead X ( 13. -500 microvolts < I X QWAV I XMAX > . .21 21) QRS Criteria for True Posterior Infarction A. The R wave (ZMAX 14. Consider Inferior Infarction A. (YQWAV 12. in Y, wave > in Lead Z must be greater than 500 microvolts 500 microvolts); and B. The time from onset of QRS to peak of the R greater than or equal to.30 milliseconds (ZRMAX 30 milliseconds); and C. ZRATIO D. The duration of the R wave greater than 45 milliseconds (ZRDUR 45 milliseconds); and E. XRATIO F. Criteria for inferior infarction; (if criteria for anterior infarction is reported). is not present the interpretation of absormal R in Z > < wave in Lead Z is -8 -10 Incomplete Right Bundle Branch Block A. The QRS duration greater than (Duration B. An 100 or equal to 110 milliseconds milliseconds) R-prime present in Lead Z. (RPRIME < -5 microvolts) Note: "Incomplete Right Bundle Branch Blockll is reported only when the QRS duration is less than 120 milliseconds and findings have been found. no other abnormal 75 15. Poor R Wave Progression A. Insufficient criteria for anterior infarction, and B. ZRATIO < C. XSUM -200 microvolts D. Not criteria for LVH < 4VRATIO/I0 or XRATIO < -10 76 CRITERIA FOR ST-T INTERPRETATIONS 1. Upright A. Normal or T Wave in X,yZ. If the ST segment is must be more 50 microvolts ((TMAX positive (ST > 0), then the maximum T wave positive. than-that value by greater than voltage B. C. 2. > 50 microvolts); or IF RR 800 milliseconds, then [TMAX TClINE (ITMAX)] > 50 microvolts, where TCLINE is the line from the start of the QRS complex to a ,point 480 msec beyond the end of the QRS complex and ITMAX is the index of TMAX - The T must not deviate from the ST wave (ITMIN microvolts. STI - segment by 50 microvolts and < ITMAX more - than 50 < 50 microvolts) STI Inverted T Wave in X,Y,Zo A. If the ST voltage B. segment is positive (ST > 0), then the minimum T wave than -50 microvolts (TMIN < -50 microvolts); must be less If the ST segment is must be more than -50 microvolts voltage C. 4. ST) Flat T Wave in X,Y,Z. A. 3. - If the ST segment is negattve (ST < 0) then the maximum T wave voltage must be greater than 50 microvolts (TMAX > 50 microvolts) If RR > or negative (ST < 0), then the minimum T wave negative. than the average ST voltage by more ST] <-50 microvolts) ([TMIN - 800 milliseconds, then [TMIN-TCLINE(ITMIN)J -50 microvolts Elevated ST in X,Y,Z. A. An average ST level in X,Y, 100, 100, 150 microvolts B. The average ST level microvolts C. Z must be elevated greater than or respectively, in X,Y, and or greater than 75, 75, 100 Z must be or respectively, If the T wave in that lead is upright, then the ST segment must be in that lead (CONCAV < -25 microvolts). convex 50 Digitalis Effects A. The average ST voltage in X,Y, (ST <-50 microvolts); and B. The Z must be less in that lead shape of the ST segment concave C. or The T (CONCAV wave > 5 must be either flat microvolts). must not be than -50 microvolts inverted in lead X. or 77 6. ST Depression in X,Y,Zo A. The average ST voltage in X,Y, (ST < -50 microvolts); and B. Not or Z must be less than -50 microvolts digitalis effects. A check is also made for T wave abnormalities in IV4". This is done by combining the values of the T wave parameters from leads X and Z with the following new tolerances being set. maximum or minimum value in leads X and Z TMINX + TMINZ) is used with. the same logic as above for flat, inverted or upright T wave except that the ST segment from IV4" is used and +70 microvolts instead of +50 microvolts. The of the T sum (TMAXX + TMAXZ, wave or 78 CRITERIA FOR P WAVE INTERPRETATIONS 1. Broad P Wave A. Duration of the P (POUR> 2. Abnormal A. 3. Tall A. 160 wave greater than 160 milliseconds milliseconds). P Axis Inverted P wave in lead X with.inversion less than -100 microvolts (XPWAV < -100 microvolts). P Wave of P wave greater than 200 microvolts, 275 microvolts, 200 microvolts in leads X,y and Z respectively. (XPWAV > 200 microvolts or YPWAV> 275 microvolts or ZPWAV > 200 microvolts) Magnitude These P wave diagnoses are only made in the presence of sinus mechanism In the absence of a sinus mechanism, as determined by the rhythm program. the above interpretations are not reported. 79 APPENDIX C CRITERIA FOR RHYTHM DETERMINATION The rhythm program is built around the detection of 17 consecutive QRS complexes. A histogram is generated from the 16 R to R intervals and the major interpretation of the basic rhythm is found from the spread of this histogram. Also, a.search is made before the onset of each.QRS to determine if a P wave is present. P waves are assumed present if the search routine finds a P wave candidate in at least 8 of the 16 complexes at the same location plus or minus 10 milliseconds. If 8 or more are found then a flag is set indicating. the P waves are present and the PR interval is measured as the mode of the PR's from those beats. Average heart rate over the 16 intervals is also measured and the If the heart rate basic rhythms are modified according to the heart rate. is less than 60 beats perinute, then the rhythm is referred to as brady­ beats cardia. If the heart rate is greater than-IOO per minute, then tachycardia is attached to. the basic rhythm. However, in the case of atrial fibrillation, the modifiers are slow ventricular response, and rapid ventricular response. The following 1. basic Normal 2. 3. Less present than 4 non a. P waves present b. Four more Atrial b. reported: zero bins in the R-R histogram Arrhythmia Sinus a. 40 are Sinus Mechanism waves a. b. rhythms or non zero bins in the R-R histogram Fibrillation wave Four or present not more Regular Rhythm non zero bins in the R-R histogram Check P Waves not a. P waves b. Less than 4 present non zero bins in the R-R histogram 80 5. Ventricular Bigeminy a. 60 7. 8. Atrial Every b. P 11. premature ventricular depolarization other beat a premature atrial depolarization waves must be present Three to Two A-V Wenckbach a. Every other beat premature b. Corresponding short long PR interval sequence First Degree AV Block PR interval greater than .21 seconds Supraventricular Rhythm Tachycardia a. 10. a Bigeminy ac a. 9. Every other beat Normal sinus mechanism with heart rate greater than 150 beats per minute Premature Ventricular Depolorizations 28% premature a. R-R interval b. Abnormal c. QRS duration 20 milliseconds wider than median QRS duration configuration or Supraventricular premature depolorizations a. RR interval b. P c. Not waves a 28% premature present premature ventricular depolorization APPENDIX D CRITERIA FOR ECG DIAGNOSTIC STATEMENTS 1. 2. 3. 4. 5. 6. Old Lateral Wall Infarction, Possible a. QRS criteria for lateral wall infarction b. Absence of ST elevation in leads X and Z c. Absence of T wave inversion in leads X and Z Old Anterior Infarction, Possible a. QRS criteria for anterior infarction b. Absence of ST elevation in leads X and Z c. Absence of T wave Inferior Infarction, inversion in leads X and Z Age Undetermined criteria for inferior infarction a. QRS b. Absence of ST elevation in lead Y Anterior Infarction, Age Undetermined Suspect progression a. Poor R b. Absence of ST elevation in lead X and Z c. T wave wave inversion in leads X Anterior Infarction, Age QRS criteria for anterior b. Absence of ST elevation c. T Lateral Wall Z Undetermined a. wave or infarction in lead X and Z inversion in leads X Infarction, Age or Z Undetermined a. wall infarction QRS criteria for lateral b. in leads X and Z Absence of ST elevation c. T wave inversion in leads X or Z 82 7. 8. 9. 10. 11. 12. 13. 14. Acute Anterior Infarction, Suspect a. Poor R b. ST elevation in leads Acute progres s ion wave or Z Inferior Infarction, probable a. QRS criteria for inferior infarction b. ST elevation in lead Y . Acute Anterior Infarction, Probable a. QRS criteria for anterior infarction b. ST elevation in leads X Acute Lateral Wall or Z Infarction, Rrobable a. QRS criteria for lateral wall b. ST elevation in leads X Inferior-Lateral or infarction Z Infarction, Age Undetermined a. Criteria for inferior infarction, age undetermined b. T wave inversion in lead X Inferi or-Latera 1 Acute Infarcti on, Probable a. Criteria for acute inferior infarction, b. T wave probable inversion in lead X Old True Posterior Infarction, Possible true posterior infarction a. QRS criteria for b. Absence of ST elevation c. Absence of T True Posterior wave in leads X and Z inversion in leads X and Z Infarction, Age Undetermined true infarction posterior a. QRS criteria for b. Absence of ST elevation c. T wave in leads X and Z inversion in leads X or Z 83 15. 16. 17. 18. 19. 20. Acute True Posterior Infarction, Probable a. QR5 criteria for b. 5T elevation Left Ventricular true posterior infarction in leads X Z or Hypertrophy, Voltage Criteria Only a. QR5 voltage sugges left ventricular hypertrophy b. No 5T or T wave Left Ventricular abnormalities in leads Z or X Hypertrophy a. QR5 voltage suggests left ventricular hypertrophy b. 5T depression in leads X leads X or Z Z and/or T or wave inversion in Left Anterior Hemi-block -600 a. Left axis deviation less than b. QR5 duration between 100 and 115 milliseconds Consider Bifasicular Block branch block a. Right bundle bo Left axis deviation or indeterminate axis Consider Advanced AV Block a. Rhythm diagnosis b. Heart rate of regular rhythm, P not found, less than 50 beats per minute Intraventricular induction defect unspecified 21. T changes Non specific a. Flat T wave b. T inversion in lead Y c. Absence of wave in leads X or Y, QR5 criteria for or or Z any infarction bradycardia 84 22. 23. Non Specific ao Depressed b. Absence of QRS criteria for any infarction ST Changes ST segment in leads X, Y or Z Ischemia a. T wave inversion b. Absence of any QRS criteria for any infarction in lead X APPENDIX E CRITERIA FOR ECG SERIAL STATEMENTS QRS STATEMENTS 1. QRS duration increased from to msec a. Pesent QRS duration greater than previous QRS duration b. Dlfference between present and previous QRS duration must -- -- be greater than 20 milliseconds 2. QRS duration decreased from to msec a. Previous QRS duration greater than present.QRS duration b. Difference between present and previous.QRS duration must -- -- be greater than 20 milliseconds 3. Criteria for left ventricular hypertrophy satisfied a. Present ECG has left ventricular hypertrophy b. Absence of LVH in previous ECG 4. Criteria for left ventricular hypertrophy a. Previous ECG has LVH b. Absence of LVH in present ECG 5. Criteria for right ventricular hypertrophy satisfied a. Present ECG has RVH b. Absence of RVH in present ECG 6. Criteria for right ventricular hypertrophy Previous ECG has RVH ao b. Absence of RVH in present ECG 7. QRS criteria for anterior infarction satisfied a. b. cO 8. QRS criteria for anterior infarction a. b. cO 9. Present ECG has AI Absence of AI in previous ECG Absence of poor R wave progression in no no longer no b. c. longer satisfied previous ECG longer satisfied Previous ECG has AI Absence of AI in present ECG ECG Absence of poor R wave progression in present satisfied QRS criteria for inferior infarction a. satisfied infarction Present ECG has criteria for inferior ECG Absence of inferior infarction in previous Q in Y in previous ECG Absence of questionable 86 10. QRS a. b. c. 11. Criteria for left axis deviation satisfied a. b. c. 12. for inferior infarction no longer satisfied PrevlOus ECG has inferior infarction Absence of inferior infarction in present ECG Absence of questionable Q in Y in present ECG critria Present ECG has LAD Absence of LAD_in previous ECG Absence of UDA in previous ECG Criteria for left-axis -deviation-no longer satisfied Previous ECG has-LAD a. b. c. Absence of LAD in present. ECG Absence of UDA in present ECG 13. Criteria for right axis deviation satisfied a. Present ECG has RAD b. Absence of RAD in previous-ECG c. Absence of UDA in previous ECG 14. Criteria for right axis deviation Previous ECG has RAD b. Absence of RAD in present ECG c. Absence of UDA in present ECG no longer satisfied a. 15. Criteria for left bundle branch block satisfied a. b. 16. Present ECG has LBBB Absence of LBBB in previous ECG Criteria for left bundle branch block Previous ECG has LBBB no longer satisfied a. b. 17. 18. Absence of LBBB in present ECG Criteria for right bundle branch block satisfied a. Present ECG has RBBB b. Absence of RBBB in previous ECG Criteria for right bundle branch block Previous ECG has RBBB no longer satisfied a. b. 19. Criteria for intraventricular conduction defect satisfied a. b. 20. Absence of RBBB in present ECG Present ECG has IVCD Absence of IVCD in previous ECG Criteria for intraventricular conduction defect Previous ECG has-IVCD ao b. Absence of IVCD in previous ECG no longer satisfied 87 210 QRS criteria for lateral wall infarction satisfied b. Previous ECG has lateral wall infarction Absence of lateral wall infarction in previous ECG QRS critria a. 220 ao bo 23. QRS criteria for poor R wave progression satisfied Present ECG has poor R wave progression Absence of poor R wave progression in previous ECG Absence of anterior infarction in previous ECG a. b. c. 24. QRS criteria for poor R wave progression no longer satisfied Previous ECG has poor R wave progression Absence of poor R wave progression in present ECG Absence of anterior infarction in present ECG a. b. c. 250 QRS criteria for cO true posterior infarction satisfied Present ECG has a true posterior infarction Absence of true posterior infarction in previous ECG Absence of abnormally tall R in Z in previous ECG QRS criteria for true a. ECG has TPI Absence of TPI in present ECG Absence of abnormally tall R in Z in present ECG a. b. 26. be b. cO do 290 longer satisfied abnormally tall R in Z satisfied ECG has abnormally tall R in Z Absence of abnormally tall R in Z in previous ECG Absence of true posterior infarction in previous ECG Present magnitude of R wave in lead Z minus previous magnitude of R wave in lead Z must be greater than or equal to +200 microvolts Present Criteria for abnormally tall R in Z no longer satisfied Previous ECG has abnormally tall R in Z ao Absence of abnormally tall R in Z in present ECG bo Absence of true posterior infarction in present ECG c, Present magnitude of R wave in lead Z minus previous d. microvolts R wave in lead Z must be less than -200 in Y satisfied in Y in previous ECG Absence ECG Absence of inferior infarction in previous Criteria for aD bo cO 30, no Criteria for a. 28. posterior infarction Previous c. 27. for lateral wall infarction no longer satisfied Prevlous ECG has lateral wall infarction Absence of lateral wall infarction in present ECG questionable Q wave questionable Q of questionable Q in Y Present ECG has in Y no longer satisfied Criteria for questionable Q wave Previous ECG has QQ in Y ao ECG Absence of QQ in Y in present b. in present ECG Absence of inferior infarction c . magnitude of 88 ST STATEMENTS 31. Criteria for ST depression in lead X satisfied Present ECG has ST depression in lead X Absence of ST depression in lead X of previous ECG cO Present and previous QRS duration must be less than 120 miliseconds a. b. 32. Criteria for ST depression in lead X no longer satisfied Previous ECG has ST depression in lead X b. Absence of ST depression in lead X in present ECG c. Present and previous QRS duration must be less than 120 milliseconds a. 33. 340 Criteria for ST segment elevation in lead X satisfied a. Present ECG has ST segment elevated in lead X b. Absence of ST elevation in lead X in previous ECG c. Present and previous QRS duration must be less than 120 milliseconds Criteria for ST segment elevation in lead X no longer satisfied Previous ECG has ST segment elevated in lead X b. Absence of ST elevation is lead X in present ECG c. Present and previous QRS duration must be less than 120 milliseconds a. 35. ST elevation increasing in lead X Present ECG has ST segment elevated in lead X and previous ECG has ST segment elevated in lead X or Present ECG has digitalis effect in lead X and b. previous ECG has digitalis effect in lead X or Present ECG has ST depression in lead X and c. previous ECG has ST depression in lead X and Present ST elevation in lead X minus previous ST elevation d. in lead X must be greater than or equal to +100 microvolts a. 36. ST elevation decreasing in lead X Present ECG has ST segment elevated in lead X and previous ECG has 5T segment elevated in lead X or Present ECG has digitalis effect in lead X and b. X or previous ECG has digitalis effect in lead Present ECG has ST depression in lead X and c. X previous ECG has ST depression in lead 5T elevation Present 5T elevation in lead X minus prevl0us d. in lead X must be less than -100 microvolts be less than 120 mllllseconds Present and previous QR5 duration must e. ao nd . Note: ST statement for lead Y and Z similar criteria for those leadso Corresponding are reported for . 89 T STATEMENTS 37. 38. Criteria for flat T wave in lead X satisfied a. Present ECG has flat T wave and previous ECG does not have a flat T wave be Present and previous QRS duration must be less than 120 milliseconds Criteria for upright T wave in lead X satisfied Present ECG has an upright T wave and previous ECG does not have an upright T wave b. Present and previous QRS duration must be less than 120 milliseconds a. 39. Criteria for T wave inverstion in lead X satisfied Present ECG has T wave inversion and previous ECG does not have T wave inversion a. 40. Magnitude of a. b. c. d. 41. b. c. d. Area of inverted T wave increasing in lead X Present ECG has T wave inversion in lead X and a. previous ECG has T wave inversion in lead X or Present ECG has flat T in lead X and b. previous ECG has flat T in lead X and Present area under T wave in lead X minus previous area c. under T wave in lead X must be greater than or equal to +2000 micro d. 43. of T wave inversion decreasing in lead X Present ECG has T wave inversion in lead X and previous ECG has T wave inversion in lead X or Present ECG has flat T in lead X and previous ECG has flat T in lead X and Present magnitude of T wave in lead X minus previous magnitude of T wave in lead X must be less than -200 microvolts Present and previous QRS duration must be less than 120 milliseconds Magnitude a. 42. T wave inversion increasing in lead X Present ECG has T wave inversion in lead X and previ ous ECG has T wave i nvers i on in 1 ead X or Present ECG has fl at Tin 1 ead X and previous ECG has flat T in lead X and Present magnitude of T wave in lead X minus previous magnitude of T wave in lead X must be greater than or equal to +200 microvolts Present and previous QRS duration must be less than 120 milliseconds volts Present and previous QRS duration must be less than 120 milliseconds Area of inverted T wave decreasing in lead X Present ECG has T wave inversion in lead X and a. lead X or previous ECG has T wave inversion in Present ECG has fl at Tin 1 ead X and b. X and previous ECG has flat T in lead area Present area under T wave in lead X minus prevlos c. -2000 under T wave in lead X must be less than must be less than 120 ml111seconds d. Present and previous QRS duration . mlcrovoltso u 90 Note: Corresponding T statements for lead Y and Z similar criteria for those leads. are reported for RHYTHM STATEMENTS 44. Sinus arrhythmia present Present ECG rhythm has sinus arrhythmia b. Absence of sinus arrhythmia in previous ECG rhythm a. 45. Sinus arrhythmia no longer present a. Previous ECG rhythm has sinus arrhythmia b. Absence of sinus arrhythmia in present ECG rhythm 46. Atrial fibrillation present a. Present ECG rhythm has atrial fibrillation b. Absence of sinus arrhythmia in previous ECG rhythm 47. Atrial fibrillation no longer present a. Previous ECG rhythm has atrial fibrillation b. Absence of atrial fibrillation in present ECG rhythm 48. Bigeminal rhythm present Present ECG rhythm has bigeminal rhythm b. Absence of bigeminal rhythm in previous a. 49. Bigeminal rhythm a. b. 500 b. 51. 52. 53. 54. rhythm longer present Previous ECG rhythm has bigeminal rhythm Absence of begeminal rhythm in present ECG rhythm Regular rhythm a. no ECG no longer present Present ECG rhythm has regular rhythm P not found Absence of regular rhythm P not found in previous ECG Regular rhythm P not found no longer present a. Previous ECG rhythm has regular rhythm P not found b. Absence of regular rhythm P not found in present ECG Arrhythmia unspecified Present ECG rhythm has arrhythmia unspecified a. b. Absence of arrhythmia unspecified in previous ECG rhythm rhythm rhythm longer present arrhythmia unspecified of arrhythmia unspecified in present ECG rhythm Arrhythmia unspecified no ao Previous ECG rhythm has b. Absence _ to Heart rate has increased from a. Previous heart rate code=1 and present heart rate code=2 b. Previous heart rate code=2 and present heart rate code=3 -- -- or 3 or 91 55. Heart rate has decreased_from -.to a. b. 56. __ Prevous heart.rate.code3_and.present heart rate code=l PrevlOus heart rate cod=2 and present heart rate code=l or 2, or First degree heart block present. Preent ECG rbythm has PR interval greater than or equal to 25 milliseco b. PR lnterval less than 215 milliseconds in previous ECG rhythm Co Absence of regular rhythm P not found in present or previous ECG a. 57. First degree heart block.no.longer present Previous ECG rhythm has.PR interval greater than a. b. c. 58. or equal to -215 milliseconds PR interval less than 215 milliseconds in. present ECG rhythm Absence of regular rhythm P not found in present or previous ECG Premature ventricular depolarization present Present ECG rhythm.has.I.or more.PVD's b. No PVD's in previous ECG rhythm a. 59. Premature ventricular depolarization no longer present Previous ECG rhythm has 1 or more PVD's. b. No PVD's in present ECG rhythm a. 60. Premature supraventricular depolarization present Present ECG rhythm has I or more APD's b. No APDls in previous rhythm a. 61. Premature supraventricular. depolarization no Previous ECG rhythm has I.or-more APDls b. No APD's in present ECG rhythm longer present a. Heart rate code Heart rate code Heart rate code Note: 1 for heart rates less than 60 beats per minute 2 for heart rate between. 60 and 100 beats per minute 3 for heart rates greater than 100 beats per minute OTHER STATEMENTS 62. Criteria for digitalis effect-satisfied Present ECG has digitalis effect in lead X, lead Y or lead Z Absence of digitalis effect in lead X, Y, or Z on previou EG b. Present and previous QRS duration must be less than 120 mllllseconds c. a. 63. Criteria for digitalis effect no. longer satisfied or lead Z Previous ECG has digitalis effect in lead X, lead Y or Absence of digitalis effect in lead X, Y, or Z b. present,EC than 120 mllllseconds Present and previous QRS duration must be less c. a. 92 GENERAL SERIAL STATEMENTS 64. No changes in morphology and rhythm. since XX/XX/XX (date) No significant changes in morphology or rhythm were found the serial 65. in comparison Changes since XX/XX/XX (date) A significant change was found either or both during the serial comparison in morphology or rhythm NOTES 1. The two ECGls compared by. the serial program are at least one day apart in time. The present ECG referred. to. is the last ECG stored on the patient. The previous ECG is the first ECG at least one day earlier. 2. An ECG rhythm will not be.compared unless it was stored within 15 minutes If two ECG.rhythms are not found after the morphology.data was stored. the morphology data is compared without the rhythm data. APPENDIX F Improvement in the rhythm analysis program finding a new of matched algorithm filtering46 for location of the P model a of the P convolved with the ECG resulted in a However, initial attempts such for several such models; ECG to the next, to model formulated which when a P wave P resulted in wave. a need varied greatly from wave one (unimodal, bimodal, biphasic, inverted, broad, peaked, Thus, of the wave from the baseline) P Table F.l is the model wave. was preceded by Using techniques waves. spike output for each that is, the P etc.). a wave must be model (i.e., selected which abstracted the main features was the width of the P without and the fact that it deviates wave being closely correlated (set of filter to any weights) particular used in this approach. Table F.l Weights for P filter wave -1,0,0,0,0,0,2,0,0,0,0,0,-1 Fig. F.l shows in the Table there the are extremely simple model can be implemented accumulator, of the a in relationship only three for unaffected by d.c. zero zero wave to the model. weights (-1,2-1) giving In implementation. ,omplies two subtracts. a no seen an load of the The fact that the that the output of the filter will bias and thus there is As fact, the matched filter computer assembly language with only binary shift and weights lOS non of the P sum be need to calibrate the ECG 94 before analysis. fi lter. The Fig. F.2 shows the higher frequencies would have been frequency characteristics of little are previously filtered by filter D of Chapter II. Search for the P wave proceeds by applying the filter to from the end of the T wave to the onset of the next QRS. is rate dependent and ifa R-R interval is made and thus no will wave the search interval will be 90 of the importance since the signal an This interval is less than 90 points be found in that interval. no points (450 milliseconds) beyond the first intervals between 90-120 points the start will seconds) In all before the second QRS. or no P will waves a fixed tolerance greater than the first tolerance interval one and all points second a to not mark the (30 points) when a P points (100 milli­ (30 AID units If the filter search is made presently) output is through the where the output of the filter is greater than half the maximum filter output taken be 20 The output of the filter within this in the interval. be found For (90-(RR-120)) points be the end will cases interval must be greater than search The start of QRS for RR intervals greater than 120 points (600 milliseconds). beyond the first QRS. interval same wave P wave marked are twice by as skipping To locate is marked. a P wave. Care is forward 150 milliseconds inverted P waves the absolute outabsolute value of the filter output is used and hence any put greater than Fig. Fo3 shows this half the maximum absolute value will ECG with inverted P waves which has also be marked. been marked using algorithm. With the under an one new development. of this experimental search algorithm Table F.2 program. gives a the new new rhythm analysis program is rhythms included in the logic 95 Table F.2 Additional As seen Rhythm Interpretation in New Arrhythmia Program 1. A-V dissociation 2. Second degree 3. Atri fl utter a 1 in the Table the additions A-V block are P-QRS relationship than in the original I rhythms involving set of rhythms a more \ / \ ---- -----\ I \ I \ / \ I \ I \ I \ '------\ r=:: P FILTER WAVE Fi g. . Relationshlp b e t ween P F.1 wave and P wave complicated (see Appendix D). Matched Filter 96 o Magnitude (db) o 32 16 64 48 80 96 Frequency Fig. F.2 Frequency of Characteristic of To aid the recognition of these rhythms P Wave Filter new parameters have been measured. They include 1, PP interval, 2, PR interval variation (coefficient of variation of PR intervals) and 3, P-P variation. Also, the measurement the coefficient of variation of the of R-R variation is changed to the measure of variation instead of the histogram RR intervals described in as Chapter IV. use 97 Fig. F.3 Results of P-Wave Location Initial has arisen in testing of the new Algorithm Using logic has been encouraging, but difficulty locating examples of the new rhythms. rhythm patterns is available, but has only three The three nized phased were by the examples of new logic. into the clinical Matched Filter A-V dissociation. A cases These digital tape with 67 of the additional were properly With this encouragement the program is setting for continued evaluation. recog- now being rhythrr