Accepted for/Published in: Journal of Medical Internet Research
Date Submitted: Apr 17, 2026
Date Accepted: Oct 14, 2026
Imaging-Based Artificial Intelligence for Predicting Pathological Complete Response in Esophageal Cancer After Neoadjuvant Therapy: A Systematic Review and Meta-Analysis
ABSTRACT
Background:
To systematically assess the diagnostic performance of imaging-based artificial intelligence (AI) models for predicting pathological complete response (pCR) in esophageal cancer after neoadjuvant therapy.
Objective:
This meta-analysis aims to assess the predictive performance of imaging-based AI models for pCR in esophageal cancer after neoadjuvant therapy and factors affecting their performance.
Methods:
A systematic search was conducted in PubMed, Embase, Web of Science, and Cochrane Library up to July 22, 2025, for studies using imaging-based AI to predict pCR in esophageal cancer, with pathological assessment as reference. Study quality was evaluated with PROBAST-AI, and evidence certainty was assessed using GRADE. Pooled sensitivity, specificity, diagnostic odds ratio (DOR), and area under the receiver operating characteristic curve (AUC) were calculated using a bivariate random-effects model for internal and external validation cohorts.
Results:
Twenty-four retrospective studies were included, involving 1,207 patients in internal validation and 550 in external validation. For internal validation, pooled sensitivity was 0.80 (95% CI, 0.74–0.85), specificity 0.81 (95% CI, 0.77–0.86), DOR 17.51 (95% CI, 11.67–26.25), and AUC 0.88 (95% CI, 0.85–0.90). For external validation, pooled sensitivity was 0.74 (95% CI, 0.67–0.80), specificity 0.82 (95% CI, 0.76–0.86), DOR 13.00 (95% CI, 8.41–20.11), and AUC 0.84 (95% CI, 0.81–0.87). AUC was significantly higher in internal than external validation, while sensitivity and specificity remained comparable.
Conclusions:
Imaging-based AI models demonstrate promising and consistent performance in predicting pCR after neoadjuvant therapy in esophageal cancer and could complement conventional radiological assessment. However, retrospective design, study heterogeneity, and limited external validation reduce the certainty of current evidence. Future large-scale, prospective, multicenter studies integrating multimodal data are needed to enhance generalizability. Clinical Trial: Registered number: PROSPERO (CRD420251207710)
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.