Currently accepted at: Journal of Medical Internet Research
Date Submitted: Apr 29, 2026
Date Accepted: Sep 11, 2026
This paper has been accepted and is currently in production.
It will appear shortly on 10.2196/99764
The final accepted version (not copyedited yet) is in this tab.
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Diagnostic Performance of Deep Learning Algorithms for Gastric Cancer: A Systematic Review and Meta-analysis
ABSTRACT
Background:
Despite advances in multimodal treatment strategies, the prognosis of patients with advanced gastric cancer (GC) remains poor, with markedly reduced five-year survival rates compared with cases detected at an early stage. Deep learning (DL)-based artificial intelligence (AI) is emerging as a promising auxiliary tool.
Objective:
This study conducted a systematic review and meta-analysis to evaluate the GC diagnostic performance of DL models using endoscopic images.
Methods:
Literature was retrieved from five databases, MEDLINE, EMBASE, Cochrane Library, IEEE, and Web of Science, according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy (PRISMA-DTA) guidelines. Data was extracted relative to GC diagnostic performance from DL models using endoscopic images. Diagnostic performance was synthesized using bivariate and hierarchical summary receiver operating characteristic (HSROC) models.
Results:
In total, 48 studies comprising 110,152 patients and 554,714 endoscopic images were analyzed. The meta-analysis showed the pooled sensitivity of the DL model was 0.954 (95% CI, 0.935–0.968) and the pooled specificity was 0.955 (0.918–0.976). The diagnostic odds ratio was 400.736, and the area under the HSROC curve was 0.977, indicating very high diagnostic performance.
Conclusions:
DL-based AI models showed excellent accuracy and discrimination in GC diagnosis using endoscopic images, with high clinical applicability. AI contributes to reducing lesion omission rates and inter-observer variability by playing an auxiliary role rather than by replacing the judgment of an endoscopic specialist. However, given the retrospective nature of the reviewed studies, further validation using large-scale prospective studies is warranted before clinical application. Clinical Trial: The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) prior to the initiation of the review (registration ID: CRD420251243998; protocol version 1.0)
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.