Perivascular Adipose Tissue Radiomics Predicts Abdominal Aortic Aneurysm Rupture: A Multicenter Study.

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Abdominal aortic aneurysms (AAA) exhibit exacerbated inflammatory responses in perivascular adipose tissue (PVAT) preceding rupture.
• These inflammatory changes can cause functional and structural alterations, leading to detectable imaging disparities.
• Radiomics, combined with machine learning (ML), offers a powerful approach for extracting quantitative image features for clinical decision support.

Purpose of the Study:

• To investigate the potential of radiomic features derived from PVAT in predicting AAA rupture.
• To develop and validate ML models for AAA rupture prediction using PVAT radiomics.

Main Methods:

• Retrospective analysis of aortic Computed Tomography Angiography (CTA) images from two centers.
• Radiomic feature extraction from PVAT, followed by statistical analysis to identify significant differences between stable and ruptured AAA.
• Dimensionality reduction and construction/validation of ten ML models using internal and external datasets.

Main Results:

• 18 out of 107 extracted radiomic features showed statistically significant differences between ruptured and non-ruptured AAA groups.
• Five representative features were selected after dimensionality reduction.
• ML models achieved an average accuracy of 0.76 (AUC 0.81) internally and 0.73 (AUC 0.77) externally.

Conclusions:

• PVAT radiomic features exhibit significant differences between ruptured and non-ruptured AAA patients.
• The findings support the feasibility of using PVAT radiomics for predicting AAA rupture.
• ML models based on PVAT radiomic features demonstrate reasonable accuracy for clinical application.