Deep learning driven colorectal polyp analysis: a review of detection, classification and segmentation methods

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Colorectal polyps are precursors to colorectal cancer, posing detection challenges during colonoscopy due to variations in appearance.
• Accurate polyp identification is crucial for effective colorectal cancer screening and prevention.

Purpose of the Study:

• To systematically review artificial intelligence (AI)-based methods for colorectal polyp detection, classification, and segmentation.
• To analyze current state-of-the-art AI models, datasets, and preprocessing techniques in this field.

Main Methods:

• Comprehensive literature review of AI applications in colorectal polyp analysis.
• Analysis of publicly available datasets, data augmentation strategies, and common challenges like low contrast and class imbalance.
• Evaluation of AI model architectures, performance trends, and standard metrics for benchmarking.

Main Results:

• AI, particularly deep learning, shows significant promise in automating polyp detection, classification, and segmentation.
• Identified strengths and weaknesses of various AI models and highlighted the importance of robust datasets and preprocessing.
• Assessed standard performance metrics for consistent evaluation of AI algorithms.

Conclusions:

• AI-based approaches offer a structured reference for analyzing colorectal polyps, enhancing diagnostic accuracy.
• Discussed existing research gaps and future directions to facilitate the clinical translation of AI tools.
• Emphasized the need to bridge the gap between experimental AI performance and real-world clinical deployment.