Development and validation of a deep learning model for predicting gastric cancer recurrence based on CT imaging: a...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Postoperative recurrence significantly impacts gastric cancer (GC) patient prognosis.
• Accurate prediction of GC recurrence is critical for patient management.

Purpose of the Study:

• To develop and validate a novel deep learning model for predicting postoperative recurrence in gastric cancer patients.
• To assess the performance of the deep learning fusion signature (DLFS) compared to existing clinical models.

Main Methods:

• Retrospective analysis of 2813 gastric cancer patients undergoing radical surgery.
• Development of a deep learning signature (DLS) using pretreatment CT images and a deep learning fusion signature (DLFS) integrating DLS with clinical features.
• Validation of the DLFS model's predictive performance, calibration, and clinical usefulness across training, internal, and external cohorts.

Main Results:

• The DLFS demonstrated excellent predictive performance with AUCs ranging from 0.833 to 0.859 across validation sets.
• The DLFS model significantly outperformed traditional clinical models and the DLS alone.
• High-risk patients identified by DLFS exhibited a significantly poorer prognosis.

Conclusions:

• The developed DLFS is a robust and accurate tool for identifying high-risk gastric cancer patients for postoperative recurrence.
• DLFS shows potential as a valuable prognostic factor in gastric cancer management.
• This integrated model aids in personalized risk assessment for GC patients post-surgery.