Back to 2026 Abstracts
Computational Analysis Of Thoracic Aortic Aneurysm With Radiomics Can Predict Endoleak Following Thoracic Endovascular Aortic Repair
Mohsen Bannazadeh, MD,
Seyed Morsal Aghili, MD, Tej Sura, MD, Stevan S. Pupovac, MD, Derek Brinster, MD
Northwell Health, Manhasset, NY, USA.
OBJECTIVES: Radiomics has emerged as a powerful imaging biomarker approach, enabling quantitative analysis of radiologic features to support risk stratification and predictive modeling. In this study, we applied radiomic analysis of the aneurysm sac on pre-thoracic endovascular aortic repair (TEVAR) computed tomography angiography to evaluate its potential for predicting endoleak development.
METHODS:A retrospective review was undertaken of all consecutive patients who underwent TEVAR from January 2018 to January 2023. Patients with Isolated thoracic aortic aneurysms were included. Pre-TEVAR CT angiograms were segmented from the ascending aorta to the common femoral artery utilizing 3D Slicer software. Radiomics features were extracted using PyRadiomics, including both original features (n = 120) and filter-derived features (wavelet and Laplacian of Gaussian, n = 1,316). For each dataset, features were preprocessed, screened with univariate Mann-Whitney U tests (p < 0.20), and reduced using LASSO logistic regression. Model performance was evaluated using leave-one-out cross-validation.
RESULTS:We included 40 patients who underwent TEVAR during this period for an isolated thoracic aortic aneurysm with available appropriate CTA. Endoleak was identified in 13 patients (7 type I, 5 type II, and 1 type III). Sixteen radiomics features were selected using univariate analysis (p<0.2), of which LASSO selected 10 predictors of endoleak, including dependence non-uniformity, dependence variance, gray-level non-uniformity, and several GLCM-based texture descriptors. The resulting model achieved an AUC of 0.73 in cross-validation.
CONCLUSIONS:Radiomics analysis of pre-TEVAR CTA can identify sac texture heterogeneity as the most consistent imaging biomarker associated with subsequent endoleak. These preliminary findings could potentially be utilized in TEVAR planning and risk stratification. Larger studies are needed to confirm and refine these results.
Back to 2026 Abstracts
