Abstract:Background Conventional MRI staging can be challenging in the preoperative assessment of rectal cancer. Deep learning methods based on MRI have shown promise in cancer diagnosis and prognostication. However, the value of deep learning in rectal cancer T‐staging is unclear. Purpose To develop a deep learning model based on preoperative multiparametric MRI for evaluation of rectal cancer and to investigate its potential to improve T‐staging accuracy. Study Type Retrospective. Population After cross‐validation, 260 patients (123 with T‐stage T1‐2 and 134 with T‐stage T3‐4) with histopathologically confirmed rectal cancer were randomly divided to the training (N = 208) and test sets (N = 52). Field Strength/Sequence 3.0 T/Dynamic contrast enhanced (DCE), T2‐weighted imaging (T2W), and diffusion‐weighted imaging (DWI). Assessment The deep learning (DL) model of multiparametric (DCE, T2W, and DWI) convolutional neural network were constructed for evaluating preoperative diagnosis. The pathological findings served as the reference standard for T‐stage. For comparison, the single parameter DL‐model, a logistic regression model composed of clinical features and subjective assessment of radiologists were used. Statistical Tests The receiver operating characteristic curve (ROC) was used to evaluate the models, the Fleiss' kappa for the intercorrelation coefficients, and DeLong test for compare the diagnostic performance of ROCs. P‐values less than 0.05 were considered statistically significant. Results The Area Under Curve (AUC) of the multiparametric DL‐model was 0.854, which was significantly higher than the radiologist's assessment (AUC = 0.678), clinical model (AUC = 0.747), and the single parameter DL‐models including T2W‐model (AUC = 0.735), DWI‐model (AUC = 0.759), and DCE‐model (AUC = 0.789). Data Conclusion In the evaluation of rectal cancer patients, the proposed multiparametric DL‐model outperformed the radiologist's assessment, the clinical model as well as the single parameter models. The multiparametric DL‐model has the potential to assist clinicians by providing more reliable and precise preoperative T staging diagnosis. Evidence Level 3 Technical Efficacy Stage 2

Computed Tomography for Preoperative Colorectal Staging Prediction Based on Deep Multi-task Learning: Colorectal Staging Prediction Based on Deep Multi-task Learning

MTSC2: Multi-task Survival Prediction on Colorectal CT Images with Clinical Information

A Transformer-Based Model for Preoperative Early Recurrence Prediction of Hepatocellular Carcinoma with Muti-modality MRI

Using Multi-Task Learning to Improve Diagnostic Performance of Convolutional Neural Networks

Multi-Size Deep Learning Based Preoperative Computed Tomography Signature for Prognosis Prediction of Colorectal Cancer

A computed tomography-based multitask deep learning model for predicting tumour stroma ratio and treatment outcomes in patients with colorectal cancer: a multicentre cohort study

A Contrast-Enhanced CT-Based Deep Learning System for Preoperative Prediction of Colorectal Cancer Staging and RAS Mutation

Diagnosis and Survival Prediction of Colorectal Cancer Using Deep Learning from the Histopathological HE Images

Robust Deep Multi-task Learning Framework for Cancer Survival Analysis.

Cancer Classification With Multi-Task Deep Learning

CST: A Multitask Learning Framework for Colorectal Cancer Region Mining Based on Transformer

Multi-task Deep Convolutional Neural Network for Cancer Diagnosis

Cancer Classification with Multitask Deep Learning

Deep Learning‐Based Multiparametric MRI Model for Preoperative T‐Stage in Rectal Cancer

A CT-based multitask deep learning model for predicting tumor stroma ratio and treatment outcomes in patients with colorectal cancer: a multicenter cohort study

A Fully Automated Multimodal MRI-Based Multi-Task Learning for Glioma Segmentation and IDH Genotyping.

Multi-Task Learning With Hierarchical Guidance for Locating and Stratifying Submucosal Tumors

Predicting rectal cancer prognosis from histopathological images and clinical information using multi-modal deep learning

Multi-task Learning for Segmentation and Classification of Tumors in 3D Automated Breast Ultrasound Images

MPBD-LSTM: A Predictive Model for Colorectal Liver Metastases Using Time Series Multi-phase Contrast-Enhanced CT Scans