DeepLabV3-DenseNet and Radiomics for Non-Invasive Microsatellite Instability Detection in Colorectal Cancer: A Survey

Abstract

Microsatellite instability (MSI) is a critical biomarker in colorectal cancer (CRC), significantly influencing prognosis and immunotherapy response. Conventional MSI detection techniques such as polymerase chain reaction (PCR) and immunohistochemistry (IHC) are invasive, costly, and time-consuming. Recent advancements in artificial intelligence (AI), deep learning, and radiomics have enabled non-invasive MSI prediction using medical imaging and histopathological data. Deep learning models, particularly convolutional neural networks (CNNs), have demonstrated high diagnostic accuracy in MSI prediction from histopathological slides, achieving area under the curve (AUC) values above 0.95 in several studies. Additionally, radiomics-based approaches extract high-dimensional quantitative features from CT and PET images, enabling non-invasive characterization of tumour heterogeneity and MSI status. Hybrid frameworks combining segmentation models such as DeepLabV3 and classification models such as Dense Net have further enhanced predictive performance by improving feature extraction and localization. This survey reviews recent advancements in MSI detection using deep learning, radiomics, and hybrid architectures. It provides a comprehensive comparison of methodologies, identifies key challenges such as data heterogeneity and computational complexity, and outlines future research directions for developing scalable and clinically deployable MSI prediction systems.