Predicting Rectal Cancer Response to Total Neoadjuvant Treatment Using Artificial Intelligence Modeling Based on Magnetic Resonance Imaging and Clinical Data

Abstract Purpose To develop a model for predicting response of Total Neoadjuvant Treatment (TNT) for patients with locally advanced rectal cancer (LARC) based on baseline MRI and clinical data using artificial intelligence method.Methods Patients with LARC who received TNT were enrolled retrospectively. We defined two groups of response to TNT as pCR vs non-pCR (Group 1), and high sensitivity vs moderate sensitivity vs low sensitivity (Group 2). We extracted and selected clinical and radiomic features on baseline T2WI. Then we built logistic regression (LR) models and deep learning (DL) models. Receiver operating characteristic (ROC) curves analysis was performed to assess predictive performance of models.Results Eighty-nine patients were assigned to the training cohort, and 29 patients were assigned to the testing cohort randomly. Four predictive models were built. The area under the ROC curve (AUC) of LR models, which were predictive of high sensitivity and pCR, were 0.853 and 0.866. While AUC of DL models were 0.829 and 0.838. The accuracy of the models with group 1 are higher than group 2. Conclusion There was no significant difference between LR model and DL model. The prediction model constructed by the grouping method of pCR vs non-pCR has a higher accuracy.