Object-Specific Four-Path Network for Stroke Risk Stratification of Carotid Arteries in Ultrasound Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Disease

Background:

• Atherosclerotic carotid plaques are a significant risk factor for stroke.
• Accurate classification of carotid plaques is crucial for stroke risk stratification, diagnosis, and treatment planning.

Purpose of the Study:

• To develop and validate a deep learning model for classifying carotid plaques to assess stroke risk.
• To improve the accuracy and clinical applicability of automated carotid plaque analysis.

Main Methods:

• Proposed an object-specific four-path network (OSFP-Net) integrating transverse and longitudinal ultrasound images of carotid plaques.
• OSFP-Net utilizes feature extraction and object-specific pooling subnetworks to handle varying plaque sizes and shapes.
• The model was trained and tested on a clinical dataset of 333 subjects with 1332 carotid plaques.

Main Results:

• OSFP-Net demonstrated superior performance compared to several state-of-the-art deep learning methods.
• Experimental results showed high clinical agreement between the model's predictions and the ground truth.
• The model effectively captured informative context for improved classification accuracy.

Conclusions:

• OSFP-Net is a promising tool for accurate carotid plaque classification and stroke risk stratification.
• The developed model has significant potential for clinical application in monitoring patients at risk for stroke.