Role of MRI radiomics in deep learning-based prediction of intestinal diseases

Magnetic resonance imaging (MRI) is widely used for the diagnosis, evaluation, and follow-up of intestinal diseases. With advances in artificial intelligence, MRI radiomics and deep learning have emerged as promising tools for prognostic assessment and treatment guidance. This review synthesizes current evidence on MRI radiomics and deep learning for prognostic assessment of intestinal diseases, with a focus on inflammatory bowel disease and colorectal cancer.

We conducted a narrative review of studies published between January 2005 and March 2025, retrieved from PubMed/MEDLINE, Web of Science, and Embase. Eligible studies applied deep learning or radiomics approaches to MRI data to predict treatment response, recurrence, metastasis, or survival outcomes. Methodological quality and clinical relevance were critically appraised with reference to established artificial intelligence-specific evaluation frameworks.

The reviewed studies indicate that deep learning models, including convolutional neural networks, vision transformers, and multimodal fusion approaches, can effectively exploit multiparametric MRI data to improve prognostic prediction across multiple clinical endpoints. These applications encompass image preprocessing, treatment planning, prediction of therapeutic response, disease relapse, and survival outcomes. MRI-based deep learning models generally outperform conventional imaging and traditional radiomics methods, particularly when integrated with clinical variables. However, most studies remain retrospective, with limited external validation and challenges related to interpretability and generalizability.

MRI-based radiomics and deep learning hold substantial potential for enhancing precision medicine in intestinal diseases. Future progress will depend on standardized imaging protocols, multicenter prospective validation, and the development of explainable and clinically trustworthy artificial intelligence models.