4D flow characteristics of left atrial blood flow after atrial fibrillation

Atrial fibrillation is the most prevalent cardiac arrhythmia and causes increased risk for stroke, yet accurate pathophysiological diagnosis remains a challenge. This study aimed to further investigate the physiologic biomarkers in the left atrium (LA) related to atrial fibrillation (AF) using 4D flow data in an animal model with experimentally induced atrial fibrillation. Velocity-encoded phase contrast magnetic resonance angiography was used to gather 4D flow data in sinus rhythm and after 6 months of AF. Several metrics including mean velocity, peak velocity, stasis and vorticity were analyzed with custom created MatLab code and compared across regions of the LA including fibrotic regions determined from LGE MRI of the LA. Decrease in mean velocity, peak velocity, and increase in stasis were comparable to previous studies. Vorticity increased on average across the cardiac cycle, except in the LAA. Vorticity decreased on average at peak flow time in certain regions, but not the entirety of the LA. Vorticity and velocity were similar in fibrotic and non-fibrotic regions of the LA wall. The results support previous research indicating stasis and peak velocity as consistently relevant biomarkers, while suggesting AF also causes more rotational, stagnant blood flow patterns between contractions of the LA. Vorticity therefore has the potential to indicate prothrombotic conditions in the LA but needs further investigation with larger sample sizes.

ABSTRACT Atrial fibrillation is the most prevalent cardiac arrhythmia and causes increased risk for stroke, yet accurate pathophysiological diagnosis remains a challenge. This study aimed to further investigate the physiologic biomarkers in the left atrium (LA) related to atrial fibrillation (AF) using 4D flow data in an animal model with experimentally induced atrial fibrillation. Velocity-encoded phase contrast magnetic resonance angiography was used to gather 4D flow data in sinus rhythm and after 6 months of AF. Several metrics including mean velocity, peak velocity, stasis and vorticity were analyzed with custom created MatLab code and compared across regions of the LA including fibrotic regions determined from LGE MRI of the LA. Decrease in mean velocity, peak velocity, and increase in stasis were comparable to previous studies. Vorticity increased on average across the cardiac cycle, except in the LAA. Vorticity decreased on average at peak flow time in certain regions, but not the entirety of the LA. Vorticity and velocity were similar in fibrotic and non-fibrotic regions of the LA wall. The results support previous research indicating stasis and peak velocity as consistently relevant biomarkers, while suggesting AF also causes more rotational, stagnant blood flow patterns between contractions of the LA. Vorticity therefore has the potential to indicate prothrombotic conditions in the LA but needs further investigation with larger sample sizes. 2 3 TABLE OF CONTENTS ABSTRACT ii INTRODUCTION 1 BACKGROUND 2 METHODS 6 RESULTS 9 DISCUSSION 18 REFERENCES 21 4 1 INTRODUCTION Atrial fibrillation (AF), is the most common and prevalent cardiac arrhythmia and a public health concern. The condition is projected to affect 12.1 million people by 2030 [22], and can be associated with stroke, heart failure and excess mortality with a stroke-related death occurring every 4 minutes in the US [1]. AF in particular causes ischemic strokes which account for 87% of all strokes [1], and the condition increases risk of stroke by five times [2]. This is due to thromboembolism (clot) formation in the left atria, which then circulates to the brain and causes a stroke. While this relationship is well known, the mechanism of thrombogenesis (clot formation) in the atria, and evaluation of its risk, is complex and still being developed. Anticoagulant treatment can reduce thrombus formation, but comes with risk of bleeding in the patient, leaving this risk to be balanced with the risk of stroke in determining treatment [3]. Alternative preventative strategies such as mechanical occlusion or surgical excision of the left atrial appendage (LAA) reduce bleeding risks, but the adequacy of these strategies for stroke prevention remains controversial due to lack of understanding of the mechanism and location of clot formation [4]. Clinical risk assessment of stroke primarily utilizes CHA2DS2-VASc score, which identifies risk factors such as other heart conditions, age, sex, and previous strokes [5]. While comprehensive for correlational predictive factors, the assessment does not include physiological evaluation of thrombogenic risk in AF that can lead to stroke. The left atrium is the most common site of thrombus formation [6], and more accurate risk assessment for patients could potentially be derived from analysis of blood flow in this region. 2 Investigation of thrombogenic risk is focused on a portion of the LA, the left atrial appendage (LAA), which is well established as the primary site of thrombus formation in the heart [7]. The mechanism of clot formation is described by a modern version of Virchow’s triad which lists the primary factors contributing to clot formation as endocardial and endothelial dysfunction and damage, increased blood stasis, and hypercoagulability [6]. Insults to the heart such as inflammation or oxidative stress can cause fibrotic tissue formation and cardiomyopathy [3]. These physiological changes are what developing research has termed atrial myopathy, which in turn leads to fibrotic tissue development and electrophysiological remodeling of the left atrium, shown in Figure 1. These effects contribute to increased atrial fibrillation via inconsistent electrical signaling. Atrial fibrillation in turn leads to more development, fibrotic tissue inflammation, and remodeling of atrial tissue [3]. All of these changes environment contribute prone to to an thrombus formation and cyclically worsen each other. Furthermore, paroxysmal AF leads to atrial myopathy that is increasing risk for 3 stroke from prothrombotic conditions even when the arrhythmia is not present, while also increasing the risk for AF and subsequently the risk for stroke [8]. The nature of atrial myopathy and atrial fibrillation suggest that clinical investigation should target the identification of these conditions that are prothrombotic. In particular, a noninvasive method of evaluating stroke risk based on quantifiable physiological conditions in the LA is the aim of research in this field, and blood flow analysis has provided the most promising results so far [3]. Blood flow analysis of the left atrium in relation to thrombus formation traditionally uses Doppler transesophageal echocardiography (TEE). This method has successfully highlighted blood flow values relevant to thrombogenesis, namely LAA flow velocity which was found to be predictive of thrombus formation [9], and LAA peak emptying flow velocity which was associated with spontaneous echo contrast (SEC), LAA thrombus, and thromboembolic stroke [10]. The agreement between various studies on left atrial appendage peak emptying velocity and stasis (SEC) association with increased risk of thromboembolic stroke indicates their potential value for stroke risk evaluation. Though these data points could provide valuable insight into thrombus formation risk, the method of study (TEE) lacks a holistic analysis of blood flow in the LA/LAA as it is limited to 2 dimensional data [11]. Additionally, TEE is an invasive procedure, and these two limiting factors have led to the use of 4D flow MRI, which is able to measure blood flow in 3 dimensions without invasive procedures. 4D flow, which utilizes velocity-encoded phase contrast MRI, is noninvasive and provides more comprehensive blood flow measurement. Consensus on the validity and efficacy of 4D flow methods is displayed throughout various works. Three dimensional 4 flow analysis of the left atrium has been used with consistent success in providing detailed blood flow metrics, [12, 13, 14] and has further demonstrated sensitivity to hemodynamic patterns that were matched to AF [15]. More recently, 4D flow MRI was used to again show a relationship between AF conditions and parameters such as stasis and velocity in the LA, though significant individual variability between subjects limited the applicability of these results [15]. LA velocity, stasis, and vorticity have further demonstrated technical reproducibility while also showing variability over longitudinal cases, indicating potential as valuable biomarkers for AF [16]. With the establishment of 4D flow as an effective measurement tool and the reproducibility of metrics relevant to AF, developing a more detailed understanding of blood flow changes under AF remains the next step in utilizing these values for clinical risk assessment. This study aimed to further apply established left atrial 4D flow methods in a direct comparison setting, allowing for detailed differentiation of the effects of AF on LA/LAA blood flow while limiting confounding variables. We applied 4D flow data with certain metrics including stasis, mean and peak velocity, and vorticity to segmented regions of the LA. This allowed for comparison of the flow changes across different regions of the atrium including the LAA, analysis of abnormal flow effects on tissue, and direct comparison of blood flow conditions in normal heart condition versus post-AF. Calculation of vorticity in particular was based on previous work in which 4D flow was used to assess vorticity and other blood flow data in healthy subjects [17]. The 4D flow parameters studied can also be further contextualized with comparisons to fibrosis. Atrial fibrosis has been associated with increased risk of stroke in patients with AF [18]. Investigation into the mechanism of this association shows a 5 relationship between fibrosis, increased clot forming regions, and aberrant blood flow (increased oscillatory shear index and endothelial cell activation potential) in patients with AF [19]. While there appears to be some association, the relationship between fibrosis and blood flow metrics has not been explored. By comparing fibrosis data from late-gadolinium-enhanced (LGE) MRI to 4D flow data and blood flow parameters in the LA, impacts of fibrotic tissue regions on the mechanisms of clot formation can be further understood. We aimed to compare changes in blood flow parameters in the AF condition to prevalence of fibrosis and ascertain any possible statistical relationship. Drawing from these developments in the study of the left atrium in atrial fibrillation, this study sought to directly compare 4D flow data before and after AF using an animal model with experimentally induced AF. We hypothesized that the post AF condition would be associated with increased measures of stagnant blood flow such as higher vorticity, lower peak velocity, lower mean velocity, and increased stasis, and that fibrotic regions in the post-AF condition would impact these measurements more than non-fibrotic regions. Characterizing these impacts on left atrial blood flow in detail and exploring more metrics in a matched-pair method aimed to help better identify prothrombotic conditions in the left atrium for further use in a clinical setting. METHODS Subjects 6 Initial data was collected from six different animal models in the CVRTI lab. We used a rapid atrial paced canine model. Animals had neurostimulators serving as pacemakers with a lead implanted in the right atrium. AF was induced by pacing at 50 Hz for 1 second and 1 second pacing off. These models were studied at two different points; first in sinus rhythm (SR) before any treatment to establish a baseline condition. Then, they were studied after six months of rapid pacing-induced atrial fibrillation. After six months of AF, animals were cardioverted back into sinus rhythm for the MRI scan [20]. Measurement In each condition, phase-contrast magnetic resonance angiography (pcMRA) with three-directional velocity encoding was collected to provide three dimensional blood flow speed data over time (4D flow MRI) relative to the pcMRA of the subjects heart. The 4D flow data was compiled to provide an average dataset for one cardiac cycle for each case in each condition consisting of twenty time points resulting in a 20-30ms difference between each data point. Additionally, late gadolinium-enhanced magnetic resonance imaging (LGE MRI) was collected for each sample in the post-AF condition to provide identification of fibrotic regions of the heart [20]. Processing Once collected, the pcMRA data was primarily processed using Seg3D2, a volume segmentation and processing tool developed by the NIH Center for Integrative Biomedical Computing at the Scientific Computing Institute at University of Utah [21]. Each pcMRA scan was segmented into the following regions: Left atrium (LA), left ventricle (LV), aorta, left atrial appendage (LAA), left atrial wall (LAW), left atrial blood pool (LABP), left atrial main chamber (LAMC), mitral valve (MV), left superior 7 pulmonary vein (LSPV), left inferior pulmonary vein (LIPV), right superior pulmonary vein (RSPV), and right inferior pulmonary vein (RIPV). These segmentations were made in 640x400x128 (x,y,z) resolution and then resampled to 160x120x60 to match the 4D flow MRI resolution. The LGE MRI data were also processed in Seg3D2 by establishing an intensity threshold for each scan and segmenting areas of the left atrial wall that were above this threshold. These were deemed the fibrotic regions of the wall and the segment was resampled to the 160x120x60 4D flow MRI. The fibrotic region segment was then subtracted from the original left atrial wall (LAW) segment to provide a non-fibrotic wall segment. These areas were compared in later processing. Secondary processing of the data was accomplished in MatLab using custom coding to carry out various calculations. 4D flow data was converted to MatLab format with X, Y, and Z vector components for each voxel at each time point, as well as velocity magnitude for each voxel at each time point. The pcMRA segmentations were coded to 8 MatLab as binary filters for their corresponding coordinates in the scan. A custom MatLab function was created to apply each regional filter to the 4D flow data, and calculate several variables from the 4D flow data for that region. Peak velocity was calculated as 95% of the maximum velocity magnitude recorded over the time period for each voxel. Mean velocity was calculated as the average velocity magnitude over the time period for each voxel. Stasis was calculated as the percentage of frames (out of 20) that the velocity magnitude for the voxel was below 0.1cm/s, a threshold previously established in AF 4D flow research [15]. Vorticity was calculated by the following formula (vi = velocity in the i direction): 𝑉 = ( მ𝑣𝑧 მ𝑦 − მ𝑣𝑦 მ𝑧 , მ𝑣𝑥 მ𝑧 − მ𝑣𝑧 მ𝑥 , მ𝑣𝑦 მ𝑥 − მ𝑣𝑥 მ𝑦 ) The vorticity equation calculates the curl of the flow velocity, resulting in a matrix of vectors that indicate the rotational motion of each voxel. The magnitude of the vorticity was also taken according to the following equation: 𝑉𝑚𝑎𝑔 = √[( მ𝑣𝑧 მ𝑦 − მ𝑣𝑦 2 ) მ𝑧 + ( მ𝑣𝑥 მ𝑧 − მ𝑣𝑧 2 ) მ𝑥 + ( მ𝑣𝑦 მ𝑥 − მ𝑣𝑥 2 )] მ𝑦 A higher vorticity magnitude indicates more rotational motion at that voxel, while lower vorticity magnitude indicates more linear motion at that voxel. By evaluating vorticity at each voxel, the rotational characteristics of the blood flow are quantified and comparable across samples. Analysis After applying each metric calculation to each sample and each region within the samples, data visualization and analysis was carried out with custom-coded MatLab 9 scripts. Mean values of vorticity, velocity, and stasis were calculated for each voxel in each sample, and these values were compared across the matched cases of pre and post-AF. In particular, we compared these values over the course of the cardiac cycle (all 20 time points) and at the peak flow time, which was determined by the peak velocity and corresponds to the point after systolic contraction in the left ventricle resulting in high flow out of left atrium through the mitral valve. Three-dimensional vector maps were also generated for the velocity vectors of each sample to visualize the general flow patterns. Color maps were used to visually compare vorticity, velocity, and stasis values in two-dimensional planes at different time values. Statistical analysis was done via the T-test to compare difference in means across the matched cases, while variance and standard deviation were also compared. RESULTS While vorticity is a relatively newer metric in this field, velocity, stasis, and peak velocity have been used thoroughly to evaluate blood flow in AF in the left atrium. Therefore, we were interested in comparing our results for these metrics with previous work to show validity of the matched pairs design of our study and also to confirm the trends that are seen in these values from previous research. Out of the six cases used for study, five provided usable data while one did not record a complete pcMRA scan and was not considered in the study. In the comparison of all results, the pulmonary veins segments (RSPV, RIPV, LSPV, LIPV) were not considered. These were eliminated from individual comparison due to the nature of the scan resolution which was such that these regions only represented a small number of 10 pixels. This resulted in a very low sample size of 4D flow data for these regions which led us to eliminate them from the comparisons as they did not provide large enough datasets for significant information. However, these areas are included in the segment for the entire left atrium (LA). The aorta segment was also not considered alongside the other segments since its peak flow time is different from that of the left atrium. Velocity Velocity in the post AF condition showed a general decrease for all samples in all regions of the left atrium over the cardiac cycle when compared to the pre-AF condition. The specific region with the largest decrease over all samples was the left atrial wall (LAW). These comparisons can be seen in Figure 2. 11 While the trend of decreasing velocity was consistent for the determined peak flow times, this difference was not seen in the comparison of mean velocity across cases for the entire 12 cardiac cycle. Instead, the LAA saw a significant decrease while the LABP and LAMC saw a significant increase in velocity. The entirety of the left atrium and left atrial wall did not see a significant change in mean velocity over the cardiac cycle. Average percent change in mean velocity is shown in Figure 3 for peak flow time and cardiac cycle. Comparing samples individually, variation in the change in peak flow mean velocity was seen. Only one sample (60) showed a decrease in peak flow mean velocity in all regions of the LA (including LAA, LABP, LAMC, LAW). Another sample (59) showed a decrease in all regions except the LAA, while another (58) showed an increase in all regions except the LAA. Similar inconsistency was seen in the velocity means over the entire cardiac cycle. Three samples (59, 60, 64) showed an increase in all regions, while two (58, 65) showed a decrease in all regions. Stasis Stasis comparisons showed an increase in stasis over the course of the cardiac cycle. Stasis was considered as blood flow velocity below 0.1cm/s and each voxel was evaluated at this threshold. Average stasis over the cardiac cycle describes the average number of time points (out of 20) that the voxel was in stasis. This value was averaged for all voxels in each scan and compared. Comparisons can be seen by region in Figure 4. The largest increase in stasis was seen in the blood pool region of the left atrium (LABP), while the smallest regional change was seen in the main chamber of the left atrium (LAMC). Individual samples all showed an increase in average time in stasis in all regions except for the LABP of 64 and LAW of 65. 13 Peak Velocity Comparing peak velocities showed a consistent decrease in the peak velocity across all regions when averaged over the cardiac cycle in all samples. Individual samples all showed a decrease in average peak velocity of the voxels except for one sample (64), which showed an increase in peak velocity averages in all regions. The 14 largest decrease in peak velocity was seen in the LAW. Vorticity Vorticity was also evaluated at peak flow time and over the entire cardiac cycle. As shown in the left of Figure 6, there was an average increase in vorticity in the LABP and LAMC, with a small increase in LAW average vorticity. However, there was a somewhat small decrease in average vorticity throughout the cardiac cycle in the LAA. As shown in the second plot in Figure 6, average change in vorticity at the peak flow time between samples was negative in all regions indicating consistent decrease in vorticity at peak flow time. 15 The vorticity change within individual samples was inconsistent, however two samples (59, 64) showed an increase in vorticity over the cardiac cycle in all regions. Sample 60 showed an increase in all regions except the LA, while 58 and 65 showed a smaller decrease in all regions except the LAW in 58. Sample 58 and 60 are shown in Figure 7. Throughout each sample there was no consistent trend in vorticity at peak flow time for any of the regions. In the vorticity means over the cardiac cycle, the LABP and LAW, as 16 well as the LAMC (equivalent to both LABP and LAW combined) were the most consistent with four out of five samples showing an increase in vorticity in these regions. In particular, as shown in the percentage comparison, the LABP showed the most notable increase in vorticity across all samples. Out of all the metrics, vorticity in particular produced extremely wide ranges of values and high standard deviations. For instance, the average standard deviation across all samples in the LA was ​264.7118 rad/s pre-AF and 213.6799 rad/s post-AF, for means 160.1465 rad/s and 162.0338 rad/s respectively. Fibrosis Areas of fibrotic tissue in the LAW determined by LGE MRI for each sample (LAF) were compared to the rest of the LAW region that was not fibrotic (LANF). These two regions were also compared to the entirety of the left atrial chamber (LAMC) for reference. The same metrics of stasis, mean velocity, peak velocity and vorticity were applied to these regions. In each category, no strong difference was observed between fibrotic and non-fibrotic regions. Both the non-fibrotic and fibrotic areas showed higher mean vorticity over the cardiac cycle, as well as higher mean velocity over the cycle, compared to the entire chamber. Stasis was lower in the fibrotic and non-fibrotic regions compared to the entire left atrial chamber, and peak velocity was higher. In comparison of individual samples, vorticity was at least slightly higher across the entire cardiac cycle in the fibrotic region compared to the non-fibrotic in all except one sample. Vorticity was also higher in the fibrotic region compared to the main chamber in three of the five samples. 17 Stasis showed inconsistent differences throughout the individual samples with no clear trend, as did velocity at peak flow time. DISCUSSION 18 The results of this study provided a number of discussion points regarding the effects of AF on blood flow patterns in the left atrium. Primarily, our results for stasis and peak velocity appeared consistent with previous research evaluating these metrics in AF compared to a healthy heart. In particular, Markl et al found AF to be associated with lower average peak velocities and higher average stasis in both the LA and LAA when compared to normal hearts [15]. Our data indicate a 17.2% average increase in stasis in the LA and a 38.5% average increase in stasis in the LAA, while also showing a 16% decrease in LA peak velocity and 12% decrease in LAA peak velocity on average. As suggested by previous research, the changes in stasis and peak velocity suggest that AF impacts the blood flow in the left atrial appendage as well as the entirety of the left atrium with a reduction in the maximum speed of blood flow and overall more stagnant blood flow [15]. These conditions are congruent with effects of underlying atrial myopathy discussed by previous work [3]. In addition to confirming previous work, our research demonstrates the consistency of these metrics for potential as biomarkers. Previous work utilizing clinical data relies on patients with AF compared to other healthy patients as controls, and variability between these patients implicated heterogeneity in the results. In this study, the matched pairs design eliminates this heterogeneity and we observed consistent changes for peak velocity and stasis in the subjects when post-AF was compared to pre-AF. This supports the use of peak velocity and stasis in the left atrium as biomarkers for clinical evaluation. With further longitudinal study in a clinical setting these metrics could become widely used to assess atrial myopathy and stroke risk from AF. 19 Evaluation of velocity and vorticity showed more variability in these data points than stasis and peak velocity. Mean velocity across samples did consistently decrease at the peak flow time (during LA contraction), which suggests that there is weaker contraction of the left atrium in the post-AF condition. This is also congruent with pathophysiology of atrial myopathy [3]. However, the inconsistency of the mean velocity for the duration of the cardiac cycle suggests that the effects of AF and atrial myopathy do not extend to slowing the overall blood flow. This is sensible considering that peak velocity occurs during LA contraction and AF appears to decrease peak velocity, therefore decreasing mean velocity at this time. However, previous research found significant differences in mean LA and LAA velocity between healthy patients and AF patients [15]. Therefore, it is possible that mean velocity is a less distinct metric compared to peak velocity and stasis and our sample size was too small to provide an accurate representation of velocity change over the cardiac cycle. Vorticity only showed a significant and consistent increase in the blood pool (LABP) region of the LA in the post-AF condition when calculated over the cardiac cycle. Conversely, vorticity was consistently lower, indicating more linear flow, at the peak flow time for the post-AF condition. This contradicts what we predicted, and perhaps suggests that vorticity is more of a factor in between LA contractions. Higher vorticity in the blood pool during the cardiac cycle implies that the blood has less linear motion in between contractions which could contribute to increased risk of thrombus formation. The fact that this was not seen at the peak flow time could suggest that the LA contraction in the post-AF condition still provides enough force to drive linear blood flow toward the mitral valve, even if it is at a lower speed. 20 Analysis of the LAA vorticity was particularly contradictory as it was expected that LAA vorticity would increase in the post-AF condition. The LAA is a well-known site for thrombus formation due to the high local stasis [3]. If vorticity were to be directly associated with clot formation, there should be a significant increase in vorticity in this region. However, we saw only one sample that had an increase in vorticity over the cardiac cycle in this region while multiple samples saw little change and one saw a significant decrease in vorticity. Furthermore, as seen in the results section, the vorticity data had extremely large standard deviations and a number of very high outliers in the distribution of values for each scan. These inconsistencies could be due to the resolution of the pcMRA and 4D flow data, which were such that the LAA region in each scan was only represented by 5-10 voxels per 2D slice. Since vorticity calculation relies on the neighboring velocity values, this region had few vorticity values that could be calculated. This could account for the heterogeneity we observed in this region, and warrants further research and re-evaluation to see if more consistent results can be achieved with improved methods such as higher resolution or adjusted calculation. Finally, analysis of fibrotic regions provided little insight into the effects of these regions on blood flow metrics in the LA. There was little to no observable difference in vorticity in the fibrotic regions of the LA compared to the non-fibrotic regions, and inconsistent differences in stasis and peak velocity. Fibrotic tissue is a well-established mark of atrial myopathy and associated with AF, and contributes to a prothrombotic environment. However, our results provide little insight as to whether this type of tissue impacts blood flow conditions in the LA [3]. Resolution and sample size again provide 21 possible explanations for the lack of significant results, as fibrotic regions were represented by a small number of voxels in each scan. The application of 4D flow for evaluation of the effects of AF on blood flow patterns continues to build strength as a promising clinical tool. In this study, we were able to support previous results showing AF increases stasis in the LA and LAA, while decreasing peak velocity in these regions. The finding of decreased mean velocity was also consistent with previous results when evaluated at LA contraction. These similarities provide more evidence to support the application of these metrics for AF evaluation in a clinical setting once further evaluated in larger sample sizes of patients over extended time periods. The variability of vorticity between pre and post-AF demands further research with larger sample sizes and higher resolution to further identify any applications. While vorticity did show potential with significant change in the LABP, our results were not homogenous enough to attribute this change to AF directly and therefore more data is needed. Additionally, investigation of these same metrics in these subjects after returning to sinus rhythm for a sustained period could provide insight into the long term effects of AF and the risks associated with sustained vs. paroxysmal AF. Peak velocity, stasis, mean velocity and vorticity could be useful in differentiating severity of AF and atrial myopathy in this sort of comparison. REFERENCES 22 1. Virani, S. S., Alonso, A., Benjamin, E. J., & Bittencourt, M. S. (2020, January 29). Heart Disease and Stroke Statistics—2020 Update: A Report From the American Heart Association. Circulation. https://www.ahajournals.org/doi/10.1161/CIR.0000000000000757 2. McGrath ER, Kapral MK, Fang J, Eikelboom JW, Conghaile Aó, Canavan M, O'Donnell MJ. Association of atrial fibrillation with mortality and disability after ischemic stroke. Neurology, 2013 Aug 27;81(9):825-32. https://doi.org/10.1212/WNL.0b013e3182a2cc15 3. Shen, M. J., Arora, R., & Jalife, J. (2019). Atrial Myopathy. JACC: Basic to Translational Science, 4(5), 640–654. https://doi.org/10.1016/j.jacbts.2019.05.005 4. Katsanos, A. H., Kamel, H., Healey, J. S., & Hart, R. G. (2020). Stroke Prevention in Atrial Fibrillation. Circulation, 142(24), 2371–2388. https://doi.org/10.1161/circulationaha.120.049768 5. Lane, D. A., & Lip, G. Y. (2012). Use of the CHA2DS2-VASc and HAS-BLED Scores to Aid Decision Making for Thromboprophylaxis in Nonvalvular Atrial Fibrillation. Circulation, 126(7). https://doi.org/10.1161/circulationaha.111.060061 6. Watson, T. (2009, January 10). Mechanisms of thrombogenesis in atrial fibrillation: Virchow’s triad revisited. PubMed. https://pubmed.ncbi.nlm.nih.gov/19135613/ 7. Fuster, V., Rydén, L. E., Cannom et al, (2011). 2011 ACCF/AHA/HRS Focused Updates Incorporated Into the ACC/AHA/ESC 2006 Guidelines for the 23 Management of Patients With Atrial Fibrillation. Circulation, 123(10). https://doi.org/10.1161/cir.0b013e318214876d 8. Calenda B.W., Fuster V., Halperin J.L. and Granger C.B. : "Stroke risk assessment in atrial fibrillation: risk factors and markers of atrial myopathy". Nat Rev Cardiol 2016; 13: 549. 9. Handke, M. (2005, December 18). Left atrial appendage flow velocity as a quantitative surrogate parameter for thromboembolic risk: determinants and relationship to spontaneous echocontrast and thrombus formation--a transesophageal echocardiographic study in 500 patients with cerebral ischemia. PubMed. https://pubmed.ncbi.nlm.nih.gov/16376768/ 10. Goldman, M. E. (1999, December 12). Pathophysiologic correlates of thromboembolism in nonvalvular atrial fibrillation: I. Reduced flow velocity in the left atrial appendage (The Stroke Prevention in Atrial Fibrillation [SPAF-III] study). PubMed. https://pubmed.ncbi.nlm.nih.gov/10588784/ 11. Markl, M. (2016, September). Left Atrial and Left Atrial Appendage 4D Blood Flow Dynamics in Atrial Fibrillation. PubMed. https://pubmed.ncbi.nlm.nih.gov/27613699/ 12. Fyrenius, A. (2001, October). Three dimensional flow in the human left atrium. PubMed. https://pubmed.ncbi.nlm.nih.gov/11559688/ 13. Markl, M. (2011, January 14). Comprehensive 4D velocity mapping of the heart and great vessels by cardiovascular magnetic resonance. Journal of Cardiovascular Magnetic Resonance. 24 https://link.springer.com/article/10.1186/1532-429X-13-7?error=cookies_not_sup ported&code=5816ac5e-bbc8-4f9e-b7f2-888b31891c56 14. Ebbers, T. (2002, June). Noninvasive measurement of time-varying three-dimensional relative pressure fields within the human heart. PubMed. https://pubmed.ncbi.nlm.nih.gov/12071263/ 15. Markl, M., Lee, D. C., Ng, J., Carr, M., Carr, J., & Goldberger, J. J. (2016, March). Left Atrial 4D Flow MRI: Stasis and Velocity Mapping in Patients with Atrial Fibrillation. PubMed. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4742429/ 16. Spartera, M., Pessoa-Amorim, G., Stracquadanio, A. et al. Left atrial 4D flow cardiovascular magnetic resonance: a reproducibility study in sinus rhythm and atrial fibrillation. J Cardiovasc Magn Reson 23, 29 (2021). https://doi.org/10.1186/s12968-021-00729-0 17. Kamphuis, V.P., Westenberg, J.J.M., van der Palen, R.L.F. et al. Scan–rescan reproducibility of diastolic left ventricular kinetic energy, viscous energy loss and vorticity assessment using 4D flow MRI: analysis in healthy subjects. Int J Cardiovasc Imaging 34, 905–920 (2018). https://doi.org/10.1007/s10554-017-1291-z 18. Daccarett, M., Badger, T. J., Akoum, N., Burgon, N. S., Mahnkopf, C., Vergara, G., Kholmovski, E., McGann, C. J., Parker, D., Brachmann, J., MacLeod, R. S., & Marrouche, N. F. (2011). Association of Left Atrial Fibrosis Detected by Delayed-Enhancement Magnetic Resonance Imaging and the Risk of Stroke in 25 Patients With Atrial Fibrillation. Journal of the American College of Cardiology, 57(7), 831–838. https://doi.org/10.1016/j.jacc.2010.09.049 19. Paliwal, N., Ali, R. L., Salvador, M., O’Hara, R., Yu, R., Daimee, U. A., Akhtar, T., Pandey, P., Spragg, D. D., Calkins, H., & Trayanova, N. A. (2021). Presence of Left Atrial Fibrosis May Contribute to Aberrant Hemodynamics and Increased Risk of Stroke in Atrial Fibrillation Patients. Frontiers in Physiology, 12. https://doi.org/10.3389/fphys.2021.657452 20. Yamashita, K., Silvernagel, J., Kwan, E., Kamali, R., Ghafoori, E., MacLeod, R., Dosdall, D. J., & Ranjan, R. (2019). Changes in atrial electrophysiological and structural substrate and their relationship to histology in a long‐term chronic canine atrial fibrillation model. Pacing and Clinical Electrophysiology, 42(7), 930–936. https://doi.org/10.1111/pace.13730 21. CIBC (2016). Seg3D: Volumetric Image Segmentation and Visualization. Scientific Computing and Imaging Institute (SCI), Download from: http://www.seg3d.org 22. Colilla, S. (2013, October 15). Estimates of current and future incidence and prevalence of atrial fibrillation in the U.S. adult population. PubMed. https://pubmed.ncbi.nlm.nih.gov/23831166/ Name of Candidate: Colin Hinchman Date of Submission: May 2, 2022