Trismegisto – An Aortic Dissection Support Software with Automated Segmentation, SVM based Classification and OpenFOAM Flow Simulation

Fernando José Domínguez Morales, Enrique Emiliano Manjarrez-Peña, Brandon Orduña Velazquez, JUAN PABLO ROSADO-AÍZA, Nezih Nieto Gutierrez, Valdo Jahuey-Moncada, Luis Alberto Muñoz Ubando

Keywords: Machine Learning, Medical Imaging, Non-contrast enhanced tomography, Aortic Dissection, Semantic Segmentation, Radiomics

TL;DR: Integrated ML-based framework for aortic dissection classification and segmentation with Non-contrast enhanced computer tomography.

Abstract: Aortic dissection (AD) is a critical cardiovascular emergency that conventionally relies on contrast-enhanced computed tomography for diagnosis, which poses limitations for contraindicated patients. To address this gap, Trismegisto is presented: a diagnostic support software that analyzes non-contrast enhanced CT (NCE-CT) scans by integrating automated segmentation, machine learning classification, and OpenFOAM-based flow simulation. The proposed methodology extracts geometric and morphological features from segmented, multicentric volumetric data, utilizing ANOVA and Kruskal-Wallis feature selection to train predictive algorithms. Evaluation of the models demonstrated that Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) architectures achieved 98.3% validation accuracy, with the SVM model uniquely yielding zero false negatives. Ultimately, this work highlights the capability of machine learning to accurately identify pathological outliers from non-contrast imaging, providing a strong foundation for accessible diagnostic tools and hemodynamic visualization in global clinical settings.

Submission Category: Full Paper

Overaged Verification: Yes

Latin American Hispanic Heritage: Yes

Icml Proceedings Status: No

Submission Number: 6