An effective colorectal polyp classification for histopathological images based on supervised contrastive learning.

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Oncology

Background:

• Early detection of colon adenomatous polyps is critical for reducing colon cancer risk.
• Accurate differentiation between adenomatous polyp subtypes (tubular, tubulovillous) and hyperplastic polyps is essential for appropriate patient management.
• Existing diagnostic methods may benefit from enhanced accuracy and efficiency in histopathological analysis.

Purpose of the Study:

• To develop and validate a novel computer-aided diagnosis (CADx) system for classifying colon polyp subtypes.
• To enhance the discriminatory power of image classification models for colon histopathology.
• To improve the accuracy of distinguishing between adenomatous and hyperplastic colon polyps.

Main Methods:

• Utilized Supervised Contrastive learning for precise classification of colon histopathology images.
• Integrated the Big Transfer (BiT) model, known for its adaptability in medical imaging tasks.
• Developed a novel approach to discern between in-class and out-of-class images for improved discrimination.
• Validated the system on a custom dataset and the public UniToPatho dataset.

Main Results:

• The developed CADx system achieved classification accuracies of 87.1% on the custom dataset and 70.3% on the UniToPatho dataset.
• Demonstrated superior performance compared to traditional deep convolutional neural networks.
• Successfully differentiated between adenomatous and hyperplastic polyp subtypes with high precision.

Conclusions:

• The proposed computer-aided diagnosis system shows significant potential for accurate colon polyp classification.
• This AI-driven approach can aid pathologists in early and precise detection of precancerous lesions.
• The integration of advanced deep learning techniques offers a transformative tool for colon cancer prevention strategies.