Analysis of the clinical tumor stage and survival prognosis of rectal cancer patients on the basis of deep learning and imaging characteristics: An observational study

: This study aimed to investigate the value of deep learning (DL) imaging omics based on axial high-resolution T2-weighted imaging (T2WI) for preoperative prediction of stage T2 and T3 rectal cancer. : We analyzed complete data from 361 patients with pathologically confirmed stage T2 (n=100) and T3 (n=261) rectal cancer treated at our hospital between January 2019 and December 2023. Patients were randomly divided into training (n=262) and test (n=99) sets using 7:3 stratified sampling. : Univariate and multivariate logistic regression analyses revealed that carbohydrate antigen levels (carbohydrate antigen 199, CA19-9) [95% confidence interval (CI): 1.150-1.820, P=0.002] and lesion longest diameter (LD) (95% CI: 1.159-22.584, P=0.031) were independent risk factors for predicting T2 and T3 rectal cancer. Among all constructed models, combined feature models demonstrated superior efficacy compared to single feature models, with the XGBoost classifier using combined features in the training set showing the highest performance with an AUC of 0.998 (95% CI: 0.995-1.000), which was selected as the output model for this study. : A deep learning imaging omics machine learning model based on axial high-resolution T2W images can effectively predict T2 and T3 stages of rectal cancer, with the XGBoost classifier utilizing combined features demonstrating the best performance.