Breaking Update: Here’s a clear explanation of the latest developments related to Breaking News:AI Matches Experts in Diagnosing Early Gastric Cancer– What Just Happened and why it matters right now.
AI exhibits high diagnostic performance in detecting early gastric cancer (EGC) using white light endoscopy (WLE) images, a 2026 systematic review and meta-analysis has found.
The deep learning (DL) algorithms matched the diagnostic accuracy of expert endoscopists.
Early Gastric Cancer and Endoscopy
EGC is defined as adenocarcinoma that infiltrates the stomach, associated with a favourable prognosis and 5-year survival rate of approximately 95%. Early detection is critical for improving patient outcomes.
Upper gastrointestinal endoscopy has been established as the gold standard for the diagnosis of EGC. WLE is the preferred technique due to widespread availability and ease of use.
Screening upper gastrointestinal endoscopy has reduced mortality by approximately 50%. Although, due to the intricacy of EGC lesions, accuracy of detection relies heavily on endoscopic expertise.
Diagnostic Accuracy of Deep Learning Models
Analysis of 15 studies published between 2018 and 2025 provided more than 37,000 WLE images, making up the internal validation set.
For external validation, 4 studies published in the same period offered more than 3,500 images.
In the internal validation data set, DL algorithms based on WLE images correctly identified 91% of patients with EGC. It accurately ruled out 93% of patients without EGC.
For the external validation set, DL models correctly diagnosed 82% of patients with EGC and accurately ruled out 83% of patients without.
There was no significant difference between DL models and expert endoscopists in diagnostic sensitivity and specificity, a random-effects model found.
Real-Time Endoscopy
Included studies all used retrospective datasets, resulting in an inherent risk of selection bias and spectrum bias. The high accuracy of DL models likely represents a “best-case” scenario.
There was also variability in definitions of EGC across included studies, which focused on detecting gastric lesions from static, high-quality WLE images. These scans cannot reproduce the complexity of real-time endoscopy, that is often affected by motion blur, variations in illumination, transient interference from blood, and so on.
Future Practice in Early Gastric Cancer Diagnosis
Results indicate that DL algorithms exhibit excellent diagnostic performance in EGC detection using WLE images.
However, researchers emphasised that the diagnostic accuracy of AI in this context might overestimate its effectiveness in real-world, real-time clinical settings.
DL algorithms nonetheless have the potential to serve as a clinical decision-support tool in routine practice.
Reference
Liu J et al. Deep learning for detecting early gastric cancer with white-light endoscopy: a systematic review and meta-analysis. Front Artif Intell. 2026;DOI:10.3389/frai.2026.1734591.