A-line-based thin-cap fibroatheroma detection with multi-view IVOCT images using multi-task learning and contrastive...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Disease

Background:

• Thin-cap fibroatheroma (TCFA) poses a significant risk for acute coronary syndrome.
• Accurate and automated detection of TCFA is crucial for timely clinical intervention.

Purpose of the Study:

• To develop and validate a novel method for the automatic detection of TCFA using multi-view intravascular optical coherence tomography (IVOCT) images.
• To enhance the accuracy of TCFA detection by simultaneously performing image-level and A-line-based classification.

Main Methods:

• A multi-input-output network was designed for simultaneous image-level and A-line-based classification of IVOCT data.
• A contrastive consistency term was incorporated to ensure alignment between the two classification tasks.
• A novel architecture and regional connectivity constraint were proposed to capture spatial and global information for precise A-line classification.

Main Results:

• The proposed method achieved state-of-the-art performance on the 2017 China Computer Vision Conference IVOCT dataset.
• Achieved a total score of 88.7±0.88%, overlap rate of 88.64±0.26%, precision of 84.34±0.86%, and recall of 93.67±2.29%.

Conclusions:

• The developed method demonstrates high efficacy in the automatic detection of TCFA from IVOCT images.
• The approach effectively reduces false positives and improves the characterization of TCFA extent, aiding in the prevention of acute coronary syndrome.