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

Abstract

Colorectal cancer (CRC) remains one of the leading causes of cancer-related mortality worldwide, necessitating improved diagnostic and prognostic strategies. Microsatellite instability (MSI), a critical biomarker associated with mismatch repair deficiency, plays a significant role in treatment selection and immunotherapy response. Traditional MSI detection methods, including polymerase chain reaction and immunohistochemistry, are invasive, time-consuming, and resource-intensive. In recent years, radiomics and deep learning have emerged as promising non-invasive alternatives for MSI prediction by extracting quantitative imaging features from medical scans. This review explores the integration of radiomics feature extraction with hyperparameter-tuned pre-trained deep learning models for accurate and non-invasive MSI detection in colorectal cancer. The study synthesizes recent advancements from 2020 to 2023, focusing on model architectures, feature engineering, multimodal fusion, and optimization strategies. Furthermore, the review highlights the importance of transfer learning, self-supervised learning, and ensemble techniques in improving predictive performance and generalizability. Challenges such as data heterogeneity, lack of interpretability, and clinical translation barriers are also discussed. The findings suggest that combining radiomics with optimized pre-trained models significantly enhances diagnostic accuracy and offers a scalable solution for precision oncology. Future research should emphasize standardized datasets, explainable AI, and real-world clinical validation to facilitate widespread adoption.