Home Publications Research Data Research Grants Equipment Inventions & Patents Awards Additional Research Activities Faculties & Institutions Research Areas

Scalable Automated Framework for Left Atrial Appendage Segmentation, Clustering, and CFD Analysis

Raghunath A, Castaneda Gavina E, Jacob A, Weiss M, Voigt I, Passerini T, Mihalef V, Maier A, Lluch È (2026)

---

Publication Language: English

Publication Type: Conference contribution

Publication year: 2026

Journal

Original Authors: Adarsh Raghunath, Eduardo Castañeda, Athira Jacob, Maximillian Weiss, Ingmar Voigt, Tiziano Passerini, Viorel Mihalef, Andreas Maier, Éric Lluch

Publisher: Springer

Series: Lecture Notes in Computer Science

City/Town: Cham

Book Volume: 16459

Pages Range: 128-138

Conference Proceedings Title: Statistical Atlases and Computational Models of the Heart. Regular and CMRxRecon Challenge Papers. STACOM 2025

Event location: Daejeon

ISBN: 9783032177339

DOI: 10.1007/978-3-032-17734-6_13

Abstract

The left atrial appendage (LAA) is a major site for thrombosis, responsible for 90% of strokes in patients with atrial fibrillation (AF). Variations in LAA morphology can significantly impact blood stasis and thrombosis risk, making an accurate assessment of its shape crucial for stroke prevention. In this work, we present a fully automated computational pipeline that integrates deep learning (DL)-based segmentation, computational mesh generation, statistical shape analysis, clustering, and computational fluid dynamics (CFD) simulations to analyze patient-specific LAA morphology and flow dynamics. Our approach achieves accurate 3D segmentation of the left atrium (LA), pulmonary veins, and LAA from CTA scans, with an average Dice score of 0.93. The LAA is then extracted at the ostium following clinical guidelines and aligned using iterative closest-point registration to establish a standardized template for shape analysis. To categorize different LAA morphologies, principal component analysis (PCA) is used, followed by clustering techniques. Our pipeline is fully automated, enabling the transition from CT cardiac images to CFD simulations with a 93% success rate across 812 patients - the largest cohort of LA simulated patients in the literature. For each patient, two CFD simulations are conducted to evaluate the effect of boundary conditions under both sinus rhythm and AF. The resulting hemodynamic insights can aid in identifying high-risk LAA morphologies, potentially guiding personalized stroke prevention strategies and device interventions.

Authors with CRIS profile

Involved external institutions

How to cite

APA:

Raghunath, A., Castaneda Gavina, E., Jacob, A., Weiss, M., Voigt, I., Passerini, T.,... Lluch, È. (2026). Scalable Automated Framework for Left Atrial Appendage Segmentation, Clustering, and CFD Analysis. In Oscar Camara, Esther Puyol Antón, Maxime Sermesant, Charlène Mauge, Marta Varela, Yingliang Ma, Rasmus Paulsen, Chengyan Wang, Qian Tao, Alistair Young (Eds.), Statistical Atlases and Computational Models of the Heart. Regular and CMRxRecon Challenge Papers. STACOM 2025 (pp. 128-138). Daejeon, KR: Cham: Springer.

MLA:

Raghunath, Adarsh, et al. "Scalable Automated Framework for Left Atrial Appendage Segmentation, Clustering, and CFD Analysis." Proceedings of the 16th International Workshop, STACOM 2025, Held in Conjunction with MICCAI 2025, Daejeon Ed. Oscar Camara, Esther Puyol Antón, Maxime Sermesant, Charlène Mauge, Marta Varela, Yingliang Ma, Rasmus Paulsen, Chengyan Wang, Qian Tao, Alistair Young, Cham: Springer, 2026. 128-138.

BibTeX: Download