Related Experiment Video
Updated: Jul 27, 2025
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
An MRI Deep Learning Model Predicts Outcome in Rectal Cancer
Xiaofeng Jiang1, Hengyu Zhao1, Oliver Lester Saldanha1
1From the Departments of Colorectal Surgery and General Surgey (X.J., H.Z., X.W., J.K.) and Radiology (X.M.), the Sixth Affiliated Hospital, Sun Yat-sen University, Guangzhou, China; Guangdong Institute of Gastroenterology, Guangdong Provincial Key Laboratory of Colorectal and Pelvic Floor Diseases, Guangzhou, China (X.J., H.Z., X.W., X.M., J.K.); Department of Medicine III (X.J., O.L.S., J.N.K.), Department of Diagnostic and Interventional Radiology (X.J., S.N., C.K., D.T.), and Department of Surgery and Transplantation (I.A., S.A.L.), University Hospital RWTH Aachen, Aachen, Germany; and Else Kröner Fresenius Center for Digital Health, Medical Faculty Carl Gustav Carus, Technical University Dresden, Dresden, Germany (X.J., O.L.S., J.N.K.).
Deep learning models using MRI scans can predict rectal cancer survival. This artificial intelligence tool aids in preoperative risk stratification for better patient outcomes.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Deep learning (DL) models show potential for improving rectal cancer prognostication.
- Systematic assessment of DL models for rectal cancer survival prediction is lacking.
Purpose of the Study:
- To develop and validate a magnetic resonance imaging (MRI) based DL model for predicting survival in rectal cancer patients.
- To utilize segmented tumor volumes from pretreatment T2-weighted MRI scans for prognostication.
Main Methods:
- DL models were trained and validated on retrospective MRI scans from 507 patients (2003-2021).
- Exclusion criteria included concurrent malignancies, prior treatment, incomplete neoadjuvant therapy, or no radical surgery.
- The Harrell C-index assessed model performance; a multimodal model incorporating DL risk score and carcinoembryonic antigen was also evaluated.
Main Results:
- The best DL algorithm achieved a C-index of 0.82 for overall survival in the validation set (n=218).
- The model demonstrated significant hazard ratios in high-risk groups: 3.0 (internal test set, n=112) and 2.3 (external test set, n=58).
- A multimodal model improved performance with C-indices of 0.86 (validation) and 0.67 (external test set).
Conclusions:
- A DL model utilizing preoperative MRI can effectively predict survival in rectal cancer patients.
- This AI-driven approach can serve as a valuable preoperative risk stratification tool.
More Related Videos
07:13Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018
