Development of an image classification pipeline for atherosclerotic plaques assessment using supervised machine...

Area of Science:

• Biomedical Imaging
• Computational Pathology
• Cardiovascular Research

Background:

• Atherosclerosis involves arterial lumen narrowing due to plaque buildup.
• Coherent anti-stokes Raman scattering (CARS) microscopy images lipid-rich atherosclerotic plaques without labeling.
• CARS microscopy visualizes plaque morphology for machine learning characterization.

Purpose of the Study:

• To develop an automated pipeline for classifying atherosclerotic lesion progression.
• To utilize label-free CARS images and machine learning for plaque characterization.
• To accurately differentiate between early and advanced stages of atherosclerosis.

Main Methods:

• Developed an automated pipeline using CARS microscopy images of atherosclerotic plaques.
• Employed image preprocessing, segmentation (Otsu, watershed, K-means, foam cell thresholding), and feature extraction.
• Utilized machine learning classifiers (K-nearest neighbour, support vector machine, decision tree) on refined morphological features.

Main Results:

• The automated pipeline accurately classified three stages of atherosclerosis: early fatty streak (EFS), fatty streak (FS), and advancing atheroma (AA).
• Achieved greater than 85% class accuracy in classifying atherosclerotic lesion stages.
• Demonstrated the pipeline's ability to differentiate between EFS, FS, and AA using morphological features.

Conclusions:

• Combined CARS microscopy and computational methods create a powerful tool for automated atherosclerotic plaque progression classification.
• The developed pipeline can differentiate between key stages of atherosclerosis, including early development.
• Offers potential for earlier detection and classification of atherosclerosis onset.