AI-Based Radiomics for Non-Invasive Microsatellite Instability Detection in Colorectal Cancer

Abstract

Microsatellite instability (MSI) is a critical molecular biomarker in colorectal cancer (CRC), significantly influencing prognosis, therapeutic decision-making, and response to immunotherapy. Conventional MSI detection techniques, such as polymerase chain reaction (PCR) and immunohistochemistry (IHC), are invasive, time-consuming, and require specialized laboratory infrastructure. In recent years, artificial intelligence (AI), particularly the integration of radiomics and deep learning, has emerged as a promising non-invasive alternative for MSI detection. Radiomics enables the extraction of high-dimensional quantitative features from medical imaging modalities, capturing tumor heterogeneity, while deep learning models facilitate automated feature learning and robust prediction. This paper presents a comprehensive review of AI techniques that combine radiomics feature extraction with hyperparameter-tuned pre-trained models for non-invasive MSI detection. The study focuses on recent trends, including transfer learning, multimodal data integration, explainable AI, and advanced optimization strategies. Additionally, it highlights key challenges such as data heterogeneity, lack of standardization, limited external validation, and interpretability issues. The review synthesizes findings from recent studies, demonstrating that hybrid AI models achieve high diagnostic performance, with AUC values frequently exceeding 0.85. Finally, future research directions are discussed, emphasizing the need for standardized frameworks, scalable architectures, and clinically interpretable models to enable real-world deployment.