(A-18) Reduced Order Modeling of an Aortic Hemodynamic Fingerprint in the Penn Medicine Biobank

Introduction: Pathologic arterial hemodynamics have been linked to the risk, onset, and progression of cardiovascular diseases. Aortic disease affects at least 5.7% of people.^1,2,3 While catheterization, computational fluid dynamics, and non-invasive imaging can provide a detailed assessment of arterial hemodynamics, they are not clinically indicated for latent arterial disease. We provide a novel opportunistic screening approach to the assessment of arterial hemodynamics. We generate a patient-specific fingerprint describing the multifactorial response to a wide range of physiologic peripheral arterial resistance, pressure, and flow parameters using a reduced order model (ROM). Since the full range of physiologic parameters can be rapidly assessed using ROMs, invasive pressure measurements are not required and without computationally expensive multidimensional computational fluid dynamics (CFD).^4,5 To demonstrate the feasibility of this approach, we apply the technique to an initial cohort of patients with chest CT from the Penn Medicine Biobank (PMBB).


Materials and Methods: 198 patients with thoracic CT scans terminating before the iliac arteries were selected from the PMBB, with an average radius of 13.9 ± 2.3 mm, and an average aorta length of 32.8 ± 5.0 cm. CT scans were performed with and without contrast, and a spatial resolution of 1.024 pixels/mm. Segmentations of the thoracic aorta were obtained from CT scans from the PMBB using a machine-learning approach.⁶ Vascular Modeling Toolkit (VMTK) was used to acquire centerlines and geometry. Subsequently, an aortic ROM was used to model hemodynamic (pressure and flow) waveforms for each geometry, which can be seen in figure 1. Model parameters were gathered from literature.^7,8 The inflow boundary condition was a pulse wave with a maximum flow of 200 cm³/s and was varied up and down by ten percent. This process was iterated for each CT scan, resulting in a range of pressure waveforms for each different resistance, compliance, and inflow.


Results, Conclusions, and Discussions: Average pressure values for each time point and resistance value were generated in figure 2. A plot detailing the pulse pressure variations for different peak volumetric flow rates as well as resistances was also generated in figure 3. For the lowest resistance (920 dynes/cm⁵) the average systolic and diastolic pressures were 106 ± 5.2 and 78 ± 2.5 mmHg respectively. There was variation when compared with the highest resistance (2770 dynes/cm⁵) where the average systolic and diastolic pressures were 175 ± 5.1 and 148 ± 2.6 mmHg respectively (p < 0.05 for both systole and diastole). A full list of each resistance’s average systolic and diastolic pressures can be seen in table 1. The segmentation with the highest pressures for each resistance had an average radius of 13.2 ± 2 mm with a length of 28.1 cm, while the segmentation with the lowest pressures for each resistance had an average radius of 9.88 ± 1.3 mm, with a length of 22.8 cm.


Discussion and Conclusion: The impact of this work is the development of a non-invasive approach to determine arterial hemodynamics from CT using ROMs and it may help identify subclinical arterial disease using opportunistic screening. Previous studies have developed patient-specific ROM models; however, these were not applied on a large scale like is possible using the PMBB.⁹ Previous studies have been done to examine geometric measurements such as coronary calcium and ventricular mass, and the effect they have on aortic disease progression. Studies such as these provide information on how geometric quantities lead to aortic disease.³ A limitation of this technique is the lack of flow data due to using CT scans. This was addressed by simulating a range of boundary conditions for each patient. In the future, this process may be applied to study aortic hemodynamic properties in a large-scale population dataset.




Acknowledgements (Optional): : Acknowledgments: The authors would like to thank the National Institutes of Health National Heart Lung and Blood Institute for their support of this project R01 HL137984, T32 HL007954, and T32 EB009384


References (Optional): : 1) Stather, P. W., Sidloff, D. A., Rhema, I. A., Choke, E., Bown, M. J., & Sayers, R. D. (2014). A review of current reporting of abdominal aortic aneurysm mortality and prevalence in the literature. European Journal of Vascular and Endovascular Surgery, 47(3), 240–242. https://doi.org/10.1016/j.ejvs.2013.11.007

2) Mokashi, S.A., Svensson, L.G. Guidelines for the management of thoracic aortic disease in 2017. Gen Thorac Cardiovasc Surg 67, 59–65 (2019). https://doi.org/10.1007/s11748-017-0831-8Lima João A.C., Desai, M. Y., Steen, H., Warren, W. P., Gautam, S., & Lai, S. (2004). Statin-induced cholesterol lowering and plaque regression after 6 months of magnetic resonance imaging–monitored therapy. Circulation, 110(16), 2336–2341. https://doi.org/10.1161/01.cir.0000145170.22652.51

3) Bild, D. E. (2002). Multi-ethnic study of atherosclerosis: Objectives and Design. American Journal of Epidemiology, 156(9), 871–881. https://doi.org/10.1093/aje/kwf113

4) Liang, L., Mao, W., & Sun, W. (2020). A feasibility study of deep learning for predicting hemodynamics of human thoracic aorta. Journal of Biomechanics, 99, 109544. https://doi.org/10.1016/j.jbiomech.2019.109544

5) Tsao, N. L., McLean, M., Judy, R., Rader, D. J., Witschey, W., & Damrauer, S. M. (n.d.). Automated Measurement of Aortic Diameter based on Clinically Obtained CT Imaging Subclinical/CHD and Biobank WG.

6) Haluska, B. A., Jeffriess, L., Brown, J., Carlier, S., & Marwick, T. H. (2010). A comparison of methods for assessing total arterial compliance. Journal of Human Hypertension, 24(4), 254–262. https://doi.org/10.1038/jhh.2009.92

7) Behnam, V., Rong, J., Larson, M. G., Gotal, J. D., Benjamin, E. J., Hamburg, N. M., Vasan, R. S., & Mitchell, G. F. (2019). Windkessel measures derived from pressure waveforms only: The Framingham Heart Study. Journal of the American Heart Association, 8(14). https://doi.org/10.1161/jaha.119.012300

8) L. Itu, P. Sharma, V. Mihalef, A. Kamen, C. Suciu and D. Lomaniciu, "A patient-specific reduced-order model for coronary circulation," 2012 9th IEEE International Symposium on Biomedical Imaging (ISBI), Barcelona, Spain, 2012, pp. 832-835, doi: 10.1109/ISBI.2012.6235677.