Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Pathology

Background:

• Optical coherence tomography (OCT) is vital for assessing atherosclerotic lesions.
• Detecting these lesions in high-resolution OCT is challenging and requires specialized expertise.
• Deep learning offers automated identification of atherosclerotic lesions for risk stratification.

Purpose of the Study:

• To develop and validate a deep-learning algorithm (DeepAD) for automated atherosclerotic lesion detection in OCT.
• To leverage co-registered histopathology for high-quality annotation and training.
• To assess DeepAD's performance against expert manual analysis in a clinical setting.

Main Methods:

• A U-net based deep convolutional neural network (CNN) ensemble was trained using two datasets: histopathology-annotated OCT frames and clinically annotated OCT frames.
• The algorithm, DeepAD, was trained with co-registered histopathology for enhanced lesion prediction.
• Performance was evaluated using intersection over union (IOU) for segmentation accuracy.

Main Results:

• DeepAD achieved a median IOU of 0.68 ± 0.18 for atherosclerotic lesion segmentation.
• Training without histopathology annotations resulted in a significant performance drop (>0.25 IOU).
• In a clinical cohort, DeepAD demonstrated high sensitivity and specificity, comparable to expert analysis.

Conclusions:

• Histopathology-guided deep learning significantly enhances automated atherosclerotic lesion detection in OCT.
• DeepAD facilitates accurate lesion identification in clinical practice.
• This automated decision-support tool can aid in patient risk prediction and interventional treatment planning.