Related Experiment Video
Updated: Apr 25, 2026
03:05
Author Spotlight: Advancing Early Detection and Treatment of Gastrointestinal Tumors
Published on: February 16, 2024
1.8K
Deep Learning Model Based on Tumor and Visceral Adipose Tissue CT Features for Predicting Peritoneal Metastasis Risk
Yueyue Li1, Ximiao Wang2, Qiuying Chen1
1Department of Radiology, The First Affiliated Hospital of Jinan University, No. 613 Huangpu Avenue West, Tianhe District, Guangzhou 510632, People's Republic of China.
Radiology. Imaging Cancer
|April 24, 2026
Summary
A new deep learning model integrating CT scan features and clinical data accurately predicts peritoneal metastasis in gastric cancer patients. This multimodal approach aids in noninvasive risk stratification for better postoperative management.
Area of Science:
- Radiology and Medical Imaging
- Oncology
- Artificial Intelligence in Medicine
Background:
- Serosa-invasive gastric cancer poses a high risk of peritoneal metastasis.
- Accurate prediction of metastasis is crucial for effective treatment planning and patient management.
- Current prediction methods may lack sufficient accuracy and rely on invasive procedures.
Purpose of the Study:
- To develop and validate a deep learning model for predicting postoperative peritoneal metastasis in serosa-invasive gastric cancer.
- To integrate computed tomography (CT) scan features of tumor and visceral adipose tissue (VAT) with clinical indicators.
- To assess the model's performance against existing clinical and deep learning-only approaches.
Main Methods:
- A multicenter, retrospective study involving 416 patients with pathologically confirmed serosa-invasive gastric cancer.
- Development of a multimodal deep learning radiomics model (MDLR) using ResNet18 for feature extraction and a sparse Bayesian extreme learning machine.
- Integration of deep features from tumor and VAT CT scans with clinical variables.
- Model validation using receiver operating characteristic curves, integrated discrimination improvement, calibration, decision curve analysis, and recurrence-free survival.
Main Results:
- The developed multimodal deep learning radiomics model (MDLR) achieved high predictive performance with area under the curve (AUC) values of 0.86 in both internal and external test sets.
- The MDLR significantly outperformed clinical and deep learning-only models (P < .001).
- High-risk patients identified by the MDLR exhibited significantly shorter recurrence-free survival (P < .001).
Conclusions:
- The MDLR effectively enables noninvasive prediction of peritoneal metastasis risk in serosa-invasive gastric cancer.
- This AI-driven approach can aid in postoperative risk stratification.
- The model holds potential for improving patient management and treatment strategies.
