A deep learning–based digital biopsy for predicting early recurrence in gastric cancer

2026-04-15 · Source: Machine learning : nature.com subject feeds · Field: Health & Wellbeing — Medical Specialties & Subspecialties, Artificial Intelligence & Machine Learning, Health & Medical Research · Depth: Expert, long

Summary

A new multimodal prediction model, Recurrence Stratification and Assessment (RSA), has been developed to predict early postoperative recurrence in patients with locally advanced gastric cancer (LAGC). Published on April 15, 2026, the RSA model integrates deep learning-derived histopathological features from routine hematoxylin and eosin (H&E) slides with conventional clinical variables. It was developed using a retrospective multicenter cohort of 1,763 patients and validated across two internal cohorts, two external cohorts, and a prospective clinical trial population (NCT01516944). The model demonstrated robust performance with area under the curve (AUC) values ranging from 0.843 to 0.887. Interpretation using Shapley Additive Explanations identified key histological features contributing to recurrence risk, while transcriptomic sequencing and immune profiling revealed immune-enriched microenvironments in low-risk patients.

Key takeaway

For oncologists and pathologists assessing gastric cancer prognosis, the RSA model offers a robust, biologically informed tool for predicting early recurrence. Your clinical decisions regarding postoperative surveillance and potential immune checkpoint inhibitor utility can be guided by its high-performance predictions. Consider integrating digital pathology AI solutions like RSA to enhance risk stratification and personalize patient management strategies.

Key insights

The RSA model combines deep learning with clinical data to predict gastric cancer recurrence.

Principles

• Multimodal data improves recurrence prediction.
• Histopathological features are key risk indicators.

Method

The RSA model integrates deep learning features from H&E slides with clinical variables, validated across multiple cohorts, and interpreted using Shapley Additive Explanations for feature importance.

In practice

• Use H&E slides for deep learning feature extraction.
• Integrate clinical data for enhanced prediction.

Topics

• Deep Learning
• Gastric Cancer Recurrence
• Digital Pathology
• Multimodal Prediction Model
• Histopathological Features

Code references

• hebeidpa/GC-RC-RSA

Best for: AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• Efficient AI-Driven Multi-Section Whole Slide Image Analysis for Biochemical Recurrence Prediction in Prostate Cancer — A novel AI framework predicts prostate cancer biochemical recurrence using multi-section whole slide images, outperforming clinical benchmarks.
• Multimodal deep learning model for AI-based functional prognostic risk stratification in patients undergoing radical nephrectomy — A multimodal deep learning model predicts post-surgical kidney function decline in complex renal cell carcinoma patients.
• Cascade Classification of Dermoscopic Images of Skin Neoplasms with Controllable Sensitivity and External Clinical Validation — Cascade classification with a tunable triage threshold offers superior sensitivity control and better mimics clinical diagnosis logic.
• Scaling Recurrence-aware Foundation Models for Clinical Records via Next-Visit Prediction — RAVEN is a generative pretraining strategy for EHRs that predicts next-visit events while accounting for recurrence.
• RECOVER: Designing a Large Language Model-based Remote Patient Monitoring System for Postoperative Gastrointestinal Cancer Care — LLM-powered RPM systems can enhance postoperative GI cancer care through conversational agents and staff dashboards.
• AI identifies early risk patterns for skin cancer — AI models can predict melanoma risk with high accuracy using routine healthcare data, enabling targeted screening.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.