Computer generation of a patient specific clinical summary for radiology

Studies reported in the literature demonstrate that diagnostic interpretations of test results such as chest x-rays and ECGs can be improved by supplying important clinical information about the patient to the clinician making the interpretations. Physicians at LDS Hospital, Salt Lake City, Utah, do not routinely give this information to radiologists when ordering chest x-rays. The project described in this thesis creates a computer program that is able to generate a clinical summary relevant to the domain of chest x-rays. The computer system at LDS Hospital, the HELP System,"" includes an expert system that uses 211 clinical findings stored in the patient database to diagnose 30 pulmonary diseases. This expert system provides the medical knowledge necessary to make an intelligent selection of data for a clinical summary. A sample of 106 patients was used. The database contained history, physical examination, vital signs, and laboratory data for these 106 patients. The computer system created to produce clinical summaries has five distinct and important components. These are 1) the expert-diagnostic system already mentioned, 2) an entry-sequence file, 3) a processor program, 4) a modular-scoring-algorithm frame, and 5) a report-generator program. To create the clinical summary, the processor or driver program sequentially reads formatted records for each of the 211 findings from the entry-sequence file, determines if the finding exists in the patient's clinical database, isolates this finding in a place in memory called a ""blackboard,"" runs the 30 diagnostic frames of the expert system using the blackboard as the patient's database, calls a modular-scoring-algorithm frame that generates a score or measure of importance for the finding, sorts all findings and scores into a ""sorted"" key-sequence file after removing redundant information, and finally calls a report-generator program that produces the clinical summary report using the sorted-key-sequence file as its input. The time required to produce the summaries using four different versions of the processor program ranged from 1 hr. 33.8 mins. to 3.8 mins. The fastest version used preprocessing to tabulate finding scores for given ranges of values.""