A comparative study of artificial intelligence-based approaches for colorectal cancer diagnosis

Abstract

Colorectal cancer (CRC) remains one of the leading causes of cancer-related morbidity and mortality worldwide, underscoring the need for early and accurate diagnosis to improve patient outcomes. Artificial Intelligence (AI), encompassing Machine Learning (ML) and Deep Learning (DL), has transformed CRC detection, classification, and segmentation. This study employs the PRISMA framework to conduct a systematic and comparative review of AI-driven approaches for CRC diagnosis, covering literature published between 2019 and 2025. Unlike prior reviews the primarily summarize algorithms, this work critically synthesizes trends across Convolutional neural Network (CNN), transformer based architectures, hybrid and ensemble models, and Explainable AI(XAI) frameworks. Comparative analyses using benchmark datasets such as KVASIR, NCT-CRC-HE-100, CVC-ClinicDB, and CVC-colonDB highlight key evaluation metrics including accuracy, precision, recall, F1-score, and AUC. The review identifies persistent challenges small and imbalanced datasets, limited interpretability, and constrained clinical adoption and proposes solution such as federated learning, attention-driven architectures, and multi-center validation. Overall, this review bridges algorithmic innovation with clinical applicability, offering a unified taxonomy and future roadmap for advancing trustworthy and generalize AI systems in CRC diagnostics.