CT angiography-based radiomics as a tool for carotid plaque characterization: a pilot study

Area of Science:

• Medical Imaging
• Radiology
• Machine Learning

Background:

• Carotid atherosclerosis poses a significant risk for stroke.
• Accurate discrimination of vulnerable plaques is crucial for patient risk stratification.
• Radiomics offers a quantitative approach to extract imaging features for analysis.

Purpose of the Study:

• To develop a CT angiography (CTA)-based radiomics analysis.
• To create a machine learning model for classifying carotid plaques as vulnerable or non-vulnerable.
• To assess the model's performance in discriminating plaque vulnerability.

Main Methods:

• Extracted 230 radiomics features from CTA scans of 30 carotid atherosclerosis patients.
• Utilized Spearman correlation and univariate analysis to identify significant features.
• Developed and cross-validated logistic regression, SVM, and CART models for plaque classification.

Main Results:

• Entropy and volume features were significantly associated with plaque vulnerability (p < 0.001).
• Support Vector Machine (SVM) classifier achieved 86.7% accuracy, 92.9% precision, 81.3% recall, and 86.7% F-score.
• Classification and Regression Tree (CART) model demonstrated 86.7% accuracy and an AUC of 0.987.

Conclusions:

• CTA-based radiomics and machine learning show potential for discriminating carotid plaque composition.
• This approach may offer a reliable method for improving risk stratification in patients with carotid atherosclerosis.
• Further validation is needed to confirm clinical utility.