Computer-aided diagnosis of esophageal cancer and neoplasms in endoscopic images: a systematic review and meta-analysis of diagnostic test accuracy

Background and AimsDiagnosis of esophageal cancer or precursor lesions by endoscopic imaging depends on endoscopist expertise is inevitably subject to interobserver variability. Studies computer-aided diagnosis (CAD) using deep learning machine are the increase. However, studies with small sample sizes limited inadequate statistical strength. Here, we used a meta-analysis evaluate diagnostic test accuracy (DTA) CAD algorithms cancers neoplasms images.MethodsCore databases were searched for based presenting data performance, systematic review DTA performed.ResultsOverall, 21 19 included in meta-analysis, respectively. The pooled area under curve, sensitivity, specificity, odds ratio image-based analysis 0.97 (95% confidence interval [CI], 0.95-0.99), 0.94 CI, 0.89-0.96), 0.88 0.76-0.94), 108 43-273), Meta-regression showed no heterogeneity, publication bias was detected. invasion depth 0.96 0.86-0.99), 0.90 0.88-0.92), 0.83-0.91), 138 12-1569), respectively.ConclusionsCAD high automatic neoplasms. limitation lack performance external validation clinical applications should be overcome. Diagnosis images. Core performed. Overall,