Operational Strategies for Developing a Deep Learning and MRI Radiomics Model to Predict Treatment Response and Survival Outcomes in Rectal Cancer

Rectal cancer is one of the most important diseases of the distal colon, challenging for healthcare professionals and requiring accurate and timely treatment decisions at the early stages of diagnosis. The current treatment approach, which often includes preoperative radiotherapy and chemotherapy, is associated with different treatment responses in patients due to the diversity and biological heterogeneity of tumors, and exposes some of them to severe treatment complications or disease recurrence without receiving adequate results. This policy brief was developed based on the findings of a radiogenomic modeling study that examined the potential of artificial intelligence models in predicting treatment response and survival outcome in patients with the participation of 200 patients. The results of the study showed that the integration of histological and structural features from pretreatment MRI images with pathology data and genomic markers such as KRAS and BRAF mutation status provides a high accuracy of about 85% in predicting disease-free survival and overall survival of patients. Deep learning-based models also provided more accurate prediction of treatment response. However, the implementation of these achievements in the clinical field required the development of operational policy options, which, according to the research findings and expert opinions, were proposed as follows: "Improving radiogenomic deep learning models in clinical decision-making systems of specialized hospitals, developing a national protocol for validation and multicenter data collection of rectal cancer, and developing new clinical guidelines for screening and predicting treatment response using combined biomarkers (MRI and genomics)."