Application of artificial intelligence in colonoscopy imaging for polyp analysis-A systematic review

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Colorectal cancer is a leading cause of death, making early polyp detection crucial.
• Colonoscopy is the gold standard but its accuracy varies by operator.
• Deep learning (DL) offers potential to improve accuracy, consistency, and objectivity in polyp detection.

Purpose of the Study:

• To provide a comprehensive analysis of DL applications in colorectal polyp analysis.
• To review state-of-the-art DL methodologies and their performance.
• To identify strengths, limitations, clinical relevance, and challenges of AI in this field.

Main Methods:

• Systematic review of DL architectures (CNNs, transformers, hybrid models).
• Examination of performance on public benchmark datasets.
• Analysis of clinical relevance and prevailing challenges.

Main Results:

• AI-based systems show potential to reduce inter-observer variability and increase diagnostic efficiency.
• DL enhances accuracy, consistency, and objectivity in polyp detection, segmentation, and classification.
• Current techniques face challenges like data imbalance and generalizability.

Conclusions:

• DL is transforming colorectal lesion assessment, improving polyp detection and diagnosis.
• AI-assisted tools offer significant clinical relevance for decision-making.
• Future research should address challenges to optimize AI deployment in real-world settings.