Radiomics-Based Classification of Pathological Patterns in Common Carotid Artery Wall.

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Vascular Biology

Background:

• Abnormal echogenic patterns in the common carotid artery (CCA), like the triple signal pattern, are linked to fibromuscular dysplasia, hypertension, and cardiovascular risk factors.
• These patterns may indicate vascular aging, but visual detection is time-consuming and subjective.
• Automated detection methods are needed to assess these vascular markers efficiently.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for identifying carotid wall patterns using ultrasound image features.
• To assess the potential of ML in detecting abnormal vascular patterns in a general population cohort.

Main Methods:

• Analysis of ultrasound data from 784 participants.
• Extraction of 178 radiomic features from the CCA far wall.
• Visual classification of vascular patterns (healthy/abnormal) by a physician.
• Feature selection based on reproducibility, correlation, and relevance.
• Training and testing of Logistic Regression (LR) and Support Vector Machine models.

Main Results:

• The dataset comprised 56% healthy and 44% abnormal vascular patterns.
• Logistic Regression achieved an AUC of 0.78 on the training set and 0.72 on the test set.
• The model demonstrated good performance in discriminating between healthy and abnormal carotid wall patterns.

Conclusions:

• Machine learning classifiers can effectively differentiate between healthy and abnormal vascular wall patterns.
• This automated tool shows promise for future investigations into the clinical significance of carotid wall patterns.
• Further research in larger cohorts is warranted to explore the clinical relevance of these findings.