JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Time-sequential prediction of postoperative complications after gastric cancer surgery using machine learning: a multicenter cohort study

  • 0Department of Gastroenterological Surgery, Cancer Institute Hospital, Japanese Foundation for Cancer Research, 3-8-31 Ariake, Koto-Ku, Tokyo, 135-8550, Japan.
Gastric Cancer : Official Journal of the International Gastric Cancer Association and the Japanese Gastric Cancer Association +

|

September 8, 2025

|

September 8, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Machine learning models accurately predict post-gastric cancer surgery complications using comprehensive perioperative data. Hour-based models offer time-sequential predictions, aiding clinical decisions and improving patient outcomes.

Area Of Science

  • Surgical Oncology
  • Machine Learning in Medicine
  • Predictive Analytics

Background

  • Existing logistic regression models lack time-sequential prediction capabilities for post-gastric cancer surgery complications.
  • Comprehensive perioperative data integration with machine learning (ML) has not been fully explored for time-sequential predictions.

Purpose Of The Study

  • To develop and evaluate time-sequential machine learning models for predicting complications after gastric cancer surgery.
  • To assess the accuracy of hour-based and postoperative day-based ML models using comprehensive perioperative data.

Main Methods

  • Developed four ML models: Postoperative Day (POD) 1, POD 3, 24-hour, and 8-hour prediction models.
  • Utilized comprehensive perioperative data from 4139 patients undergoing gastric cancer surgery (2013-2019).
  • Assessed model performance using Area Under the Receiver Operating Characteristic Curve (AUC) with repeated validation.

Main Results

  • The 8-hour ML model achieved the highest AUC (0.737) for overall complications.
  • POD 3 models showed AUCs > 0.8 for pancreatic fistula (0.869) and intra-abdominal abscess (0.821).
  • The 8-hour model demonstrated superior AUCs for specific complications (e.g., pancreatic fistula 0.889, intra-abdominal abscess 0.842).
  • Key predictors in the 8-hour model included C-reactive protein, pulse rate, and intraoperative blood loss.

Conclusions

  • Hour-based ML models effectively predict post-gastric cancer surgery complications with high accuracy.
  • These models offer time-sequential prediction capabilities, enhancing clinical decision-making.
  • The integration of comprehensive perioperative data improves predictive performance for better patient outcomes.

Keywords: