Objective: To investigate the value of a mechanical learning model based on multiparametric MRI imaging histology combined with clinical and conventional radiological features for preoperative prediction of perineural invasion (PNI) in rectal cancer. Methods: We retrospectively collected a total of 123 patients diagnosed with rectal cancer by postoperative pathology from January 2016 to December 2019 in our hospital. Based on the postoperative pathology results, the patients were categorized into PNI(-) and PNI(+).Clinical data and imaging data of patients were collected. All patients were randomly divided into training cohort (n = 86) and validation (n = 37) cohort according to a ratio of 7:3. The volumes of interest were manually delineated in the T2-weighted images (T2WI) and T1-weighted images (T1WI) images, from which a total of 1476 radiomics features were extracted. Thereafter, we used Spearman correlation analysis and Mann-Whitney U test and Least Absolute Shrinkage and Selection Operator (LASSO) for feature selection sequentially, and logistic regression algorithm (LR, Logistic Regression) for the PNI prediction model construction. Three single-mode models and two mixed-mode models were included. The predictive performance and clinical utility of the models were evaluated by receiver operating characteristic curve (ROC), calibration curve and decision curve analysis (DCA). Results: The integrated Clinical-mMRI model showed the best predictive efficacy, which yielded an AUC of 0.72 (95% CI 0.612-0.827) in the training cohort and 0.901 (95% CI 0.802-0.999) in the validation cohort. Calibration curve showed good agreement between predicted results of the model and actual events, and DCA indicated good clinical usefulness. Conclusions: The integrated Clinical-mMRI model is better than other predictive models, and it has value in predicting the PNI status of patients with rectal cancer before surgery.