Automated Classification of Coronary Plaque on Intravascular Ultrasound by Deep Classifier Cascades

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Image Analysis

Background:

• Intravascular ultrasound (IVUS) is crucial for visualizing coronary arteries and atherosclerotic plaques.
• Accurate plaque classification is vital for assessing rupture risk, but manual methods are inefficient.
• Machine learning offers potential for automated plaque characterization.

Purpose of the Study:

• To develop and evaluate a novel serial classifier pipeline for automated IVUS plaque classification.
• To distinguish between five IVUS plaque categories: normal, calcified, attenuated, fibrous, and echolucent.
• To enhance the efficiency and accuracy of coronary plaque analysis in clinical practice.

Main Methods:

• A pipeline of serial classifiers, including densely connected models and traditional machine learning classifiers, was developed.
• Over 100,000 IVUS frames from 471 patients, representing five lesion types, were used for training and validation.
• The framework employs a cascade approach for staged classification of IVUS images.

Main Results:

• The proposed automated classifier achieved an overall accuracy of 0.877 in distinguishing IVUS plaque types.
• The system demonstrated robust capacity for identifying the nature and category of coronary plaques.
• The accuracy indicates significant potential for real-time plaque identification.

Conclusions:

• The developed machine learning pipeline offers an accurate and efficient method for classifying coronary plaques in IVUS images.
• This automated approach can significantly aid clinicians in characterizing plaque components and assessing risk.
• The framework has the potential to streamline clinical decision-making in routine cardiovascular practice.