Deep Learning and Optimization Approaches in Combining the Advantages of Radiomics Feature Extraction and Non-Invasive Detection of Microsatellite Instability in Colorectal Cancer Using Hyperparameter Tuned Pre-trained Model: A Review

Abstract

Microsatellite instability (MSI) is a critical biomarker in colorectal cancer (CRC), influencing prognosis, therapeutic decisions, and immunotherapy response. Conventional MSI detection techniques such as polymerase chain reaction (PCR) and immunohistochemistry (IHC) are invasive, time-consuming, and resource-intensive. In recent years, the integration of radiomics and deep learning has emerged as a promising non-invasive alternative for MSI prediction. Radiomics enables the extraction of high-dimensional quantitative features from medical imaging, capturing tumor heterogeneity, while deep learning models automate feature learning and improve predictive performance. This review explores recent advances in combining radiomics with deep learning architectures, particularly hyperparameter-tuned pre-trained models, to enhance MSI detection accuracy. The study synthesizes literature from 2020 to 2023, focusing on optimization techniques, multimodal data integration, and model generalization. Evidence suggests that radiomics-based machine learning models achieve high diagnostic performance, with area under the curve (AUC) values often exceeding 0.80, although challenges in reproducibility and external validation remain . Deep learning approaches, especially those leveraging histopathology and imaging data, demonstrate improved sensitivity and specificity in MSI classification . This review highlights current trends, limitations, and future directions for developing robust, clinically deployable non-invasive MSI detection systems.