Reduced-Order Learning of Cerebral Aneurysm Hemodynamics: Comparing POD, PCA, and Autoencoder Models Coupled with TCN

Barzegar Gerdroodbary, Mostafa; Sajadi, Seyedeh Zahra

doi:10.22059/jcamech.2026.424526.1983

Reduced-Order Learning of Cerebral Aneurysm Hemodynamics: Comparing POD, PCA, and Autoencoder Models Coupled with TCN

Document Type : Research Paper

Authors

Mostafa Barzegar Gerdroodbary ¹ ^{, 2}

Seyedeh Zahra Sajadi ³

¹ Department of Electromechanical Engineering, C-MAST-Center for Mechanical and Aerospace Science and Technology, Universidade da Beira Interior, Covilha, Portugal

² Department of Mechanical Engineering, Iranian Research Organization for Science and Technology (IROST), Tehran, Iran

³ Department of Computer Science, University of Mazandaran, Iran

10.22059/jcamech.2026.424526.1983

Abstract

Accurate characterization of cerebral aneurysm hemodynamics is important for understanding flow-related mechanisms associated with aneurysm progression and for quantifying the hemodynamic modifications induced by endovascular treatment. However, high-fidelity transient computational fluid dynamics (CFD) simulations are computationally demanding, particularly when time-resolved three-dimensional hemodynamic fields are required. This study presents a comparative evaluation of three reduced-order modeling (ROM) strategies—proper orthogonal decomposition (POD), principal component analysis (PCA), and Autoencoder—coupled with a temporal convolutional network (TCN) for reconstruction and temporal prediction of hemodynamics in a patient-specific internal carotid artery saccular aneurysm before and after endovascular coiling. Pulsatile, non-Newtonian blood flow was simulated using the Casson model, while the coiled aneurysm region was represented through a porous-medium formulation. Wall shear stress (WSS/AWSS), oscillatory shear index (OSI), and pressure fields were extracted from CFD simulations and compressed into low-dimensional representations using the three ROM approaches. A common TCN architecture was subsequently employed to learn the temporal evolution of the reduced coefficients and predict future hemodynamic states. The results demonstrate that coiling substantially suppresses intra-aneurysmal flow activity and modifies the spatial distributions of WSS, OSI, and pressure. The modal analysis further shows that approximately 99% of the dominant information can be retained using a relatively small number of modes or latent variables, although the required dimensionality varies among hemodynamic quantities and ROM techniques. All three hybrid frameworks successfully reconstructed the major spatial characteristics of the CFD fields, while prediction accuracy decreased with increasing forecast horizon. Among the investigated approaches, PCA–TCN showed the most stable predictive behavior for the present dataset and effectively reproduced the principal distributions of WSS, OSI, and pressure in both original and coiled configurations. These findings demonstrate the potential of ROM–TCN frameworks as computationally efficient surrogate models for time-resolved assessment of cerebral aneurysm hemodynamics and treatment-induced flow alterations.

Keywords

Cerebral saccular aneurysm

Endovascular coiling

Computational fluid dynamics (CFD)

Reduced-order modeling (ROM)

Machine learning

Subjects

BioMechanics

[1] A. Rostamian, K. Fallah, Y. Rostamiyan, Reduction of rupture risk in ICA aneurysms by endovascular techniques of coiling and stent: numerical study, Scientific Reports, Vol. 13, No. 1, pp. 7216-1118, 2023.

[2] H. Jiang, Z. Lu, M. B. Gerdroodbary, A. Sabernaeemi, S. Salavatidezfouli, The influence of sac centreline on saccular aneurysm rupture: computational study, Scientific Reports, Vol. 13, No. 1, pp. 11288, 2023.

[3] I. Shiryanpoor, A. Kheiri, M. B. Gerdroodbary, P. Valipour, R. Moradi, Using computational fluid dynamic for evaluation of rupture risk of micro cerebral aneurysms in the growth process: hemodynamic analysis, International Journal of Modern Physics C, Vol. 36, pp. 2450184, 2025.

[4] R. Bozorgpour, P. Kim, CFD-Based Quantification of Hemodynamic Variables in Cerebral Aneurysms: How Hemodynamics Shape Aneurysm Fate, arXiv preprint, 2025.

[5] S. V. Mousavi, M. B. Gerdroodbary, A. Sabernaeemi, S. Salavatidezfouli, P. Valipour, Impacts of the aneurysm deformation induced by stent on hemodynamic of blood flow in saccular internal carotid artery aneurysms, AIP Advances, Vol. 14, pp. 095035, 2024.

[6] A. Mantha, C. Karmonik, G. Benndorf, C. Strother, R. Metcalfe, Hemodynamics in a cerebral artery before and after the formation of an aneurysm, American Journal of Neuroradiology, Vol. 27, No. 5, pp. 1113-1118, 2006.

[7] L.-D. Jou, H. M. Shaltoni, H. Morsi, M. E. Mawad, Hemodynamic relationship between intracranial aneurysm and carotid stenosis: review of clinical cases and numerical analyses, Neurological Research, Vol. 32, No. 10, pp. 1083-1089, 2010.

[8] A. Sabernaeemi, M. B. Gerdroodbary, S. Salavatidezfouli, P. Valipour, Influence of stent-induced vessel deformation on hemodynamic feature of bloodstream inside ICA aneurysms, Biomechanics and Modeling in Mechanobiology, Vol. 22, No. 4, pp. 1193-1207, 2023.

[9] P. M. Munarriz, P. A. Gómez, I. Paredes, A. M. Castaño-Leon, S. Cepeda, A. Lagares, Basic principles of hemodynamics and cerebral aneurysms, World Neurosurgery, Vol. 88, pp. 311-319, 2016.

[10] L. Song, P. Du, E. Bao, Influence of bleb formation on intracranial aneurysm hemodynamics: a patient-specific CFD study, Chinese Journal of Physics, Vol. 99, pp. 412-425, 2025.

[11] W. Rajhi, Z. Ahmed, A. Basem, A. a. Alizadeh, S. A. Hussein, H. Rajab, B. Louhichi, W. Aich, Hemodynamic response to stent-induced aneurysm deformation in patient-specific internal carotid artery cases: a computational study, Scientific Reports, Vol. 16, pp. 921, 2026.

[12] H. Rajab, J. Djuansjah, S. A. Aminian, H. I. Hassan, H. Mushtaq, W. M. Taher, M. Alwan, R. Khan, Computational characterization of geometry-driven hemodynamics in intracranial aneurysms, The European Physical Journal Plus, Vol. 141, No. 3, pp. 312, 2026.

[13] M. B. Gerdroodbary, L.-C. Zhao, Investigation of pulsatile blood flow dynamics in anterior cerebral artery aneurysms under different physiological conditions: hemodynamic analyses, Journal of Applied and Computational Mechanics, 2025.

[14] J. R. Cebral, A. Radaelli, A. Frangi, C. M. Putman, Hemodynamics before and after bleb formation in cerebral aneurysms, Proceedings of SPIE, Vol. 6511, pp. 65112C, 2007.

[15] S. F. S. Ashkezari, F. Mut, M. Slawski, B. Cheng, A. K. Yu, T. G. White, H. H. Woo, M. J. Koch, S. Amin-Hanjani, F. T. Charbel, B. R. Jahromi, M. Niemelä, T. Koivisto, J. Frosen, Y. Tobe, S. Maiti, A. M. Robertson, Prediction of bleb formation in intracranial aneurysms using machine learning models based on aneurysm hemodynamics, geometry, location, and patient population, Journal of NeuroInterventional Surgery, Vol. 14, No. 10, pp. 1002-1007, 2022.

[16] T. Satoh, Hemodynamic analysis of a thrombosed bleb in an unruptured cerebral aneurysm, Surgical Neurology International, Vol. 15, pp. 357, 2024.

[17] W. Aich, J. Djuansjah, A. B. M. Ali, A. a. Alizadeh, M. A. Hussein, N. S. S. Singh, K. Hajlaoui, W. Rajhi, Pulsatile hemodynamic prediction in cerebral fusiform aneurysms using proper orthogonal decomposition model and long short-term memory networks, International Communications in Heat and Mass Transfer, Vol. 169, pp. 109833, 2025.

[18] I. Chatziprodromou, V. Butty, V. B. Makhijani, D. Poulikakos, Y. Ventikos, Pulsatile blood flow in anatomically accurate vessels with multiple aneurysms: a medical intervention planning application of computational haemodynamics, Flow, Turbulence and Combustion, Vol. 71, No. 1-4, pp. 333-346, 2003.

[19] B. M. W. Cornelissen, J. J. Schneiders, W. V. Potters, R. van den Berg, B. K. Velthuis, G. J. E. Rinkel, C. H. Slump, E. van Bavel, C. B. L. M. Majoie, H. A. Marquering, Hemodynamic differences in intracranial aneurysms before and after rupture, American Journal of Neuroradiology, Vol. 36, No. 10, pp. 1927-1933, 2015.

[20] M. B. Gerdroodbary, S. Salavatidezfouli, A predictive surrogate model based on linear and nonlinear solution manifold reduction in cardiovascular FSI: a comparative study, Computers in Biology and Medicine, Vol. 189, pp. 109959, 2025.

[21] N. G. Hammouda, J. Djuansjah, A. B. M. Ali, A. a. Alizadeh, M. A. Hussein, N. S. S. Singh, K. Hajlaoui, W. Aich, Patient-specific deep learning reduced-order model of blood hemodynamic indices in thoracic aortic aneurysms via dynamic mode decomposition, Chinese Journal of Physics, Vol. 99, 2026.

[22] A. Kheiri, S. Hariri, M. B. Gerdroodbary, M. Sorkhi, P. Valipour, A CFD–machine learning framework for efficient prediction of cerebral aneurysm hemodynamic under vessel dilation, Chinese Journal of Physics, Vol. 102, pp. 748-767, 2026.

[23] N. G. Hammouda, K. Kriaa, A. a. Alizadeh, H. Rajab, A. M. Sadeq, N. S. S. Singh, K. Hajlaoui, W. Aich, Physically informed machine learning for predicting non-Newtonian blood flow in giant MCA aneurysms after endovascular treatment, International Communications in Heat and Mass Transfer, Vol. 172, pp. 110296, 2026.

[24] M. Liang, J. Liu, H. Wang, S. An, C. Chen, H. Chu, M. Zhu, X. Su, P. Liang, Y. Zong, M. Wan, Ultrasound super-resolved hemodynamic estimation in microvessel using physics-informed neural networks and data assimilation, Computer Methods and Programs in Biomedicine, Vol. 274, pp. 109136, 2026.

[25] AneuriskWeb project website, Accessed; http://ecm2.mathcs.emory.edu/aneuriskweb.

[26] W. Rajhi, Z. Ahmed, A. B. M. Ali, A. a. Alizadeh, Z. A. Hussein, N. S. S. Singh, K. Hajlaoui, W. Aich, Predicting hemodynamic complexity in thoracic aortic aneurysms with a POD–LSTM surrogate model, Chinese Journal of Physics, Vol. 98, pp. 73-85, 2025.

[27] Z. Shi, G. Z. Chen, L. Mao, X. L. Li, C. S. Zhou, S. Xia, Y. X. Zhang, Machine learning–based prediction of small intracranial aneurysm rupture status using CTA-derived hemodynamics: a multicenter study, American Journal of Neuroradiology, Vol. 42, No. 4, pp. 648-654, 2021.

[28] S. Tanioka, F. Ishida, A. Yamamoto, S. Shimizu, H. Sakaida, M. Toyoda, N. Kashiwagi, H. Suzuki, Machine learning classification of cerebral aneurysm rupture status with morphologic variables and hemodynamic parameters, Radiology: Artificial Intelligence, Vol. 2, No. 1, pp. e190077, 2020.

[29] G. Li, X. Song, H. Wang, S. Liu, J. Ji, Y. Guo, A. Qiao, Y. Liu, X. Wang, Prediction of cerebral aneurysm hemodynamics with porous-medium models of flow-diverting stents via deep learning, Frontiers in Physiology, Vol. 12, pp. 733444, 2021.

[30] P. J. Roache, Perspective: a method for uniform reporting of grid refinement studies, Journal of Fluids Engineering, Vol. 116, No. 3, pp. 405-413, 1994.

[31] M. Barzegar Gerdroodbary, S. Salavatidezfouli, A predictive surrogate model of blood haemodynamics for patient-specific carotid artery stenosis, J R Soc Interface, Vol. 22, No. 224, pp. 20240774, Mar, 2025. eng

[32] I. ANSYS, 2020, ANSYS Fluent User’s Guide, Release 2020 R2, ANSYS, Inc., Canonsburg, PA

[33] H. Abdi, L. J. Williams, Principal component analysis, Wiley Interdisciplinary Reviews: Computational Statistics, Vol. 2, No. 4, pp. 433-459, 2010.

[34] P. Li, Y. Pei, J. Li, A comprehensive survey on design and application of autoencoder in deep learning, Applied Soft Computing, Vol. 138, pp. 110176, 2023.

[35] A. Chatterjee, An introduction to the proper orthogonal decomposition, Current Science, Vol. 78, No. 7, pp. 808-817, 2000.