Convolutional neural networks in the computer-aided diagnosis of Helicobacter pylori infection and non-causal comparison to physician endoscopists: a systematic review with meta-analysis

BACKGROUND: Helicobacter pylori (H. pylori) infection, if left untreated, can cause gastric cancer, among other serious morbidities. In recent times, a growing body of evidence has evaluated the use of a type of artificial intelligence (AI) known as “deep learning” in the computer-aided diagnosis of H. pylori using convolutional neural networks (CNN). We conducted this meta-analysis to evaluate the pooled rates of performance of CNN-based AI in the diagnosis of H. pylori infection. METHODS: Multiple databases were searched (from inception to June 2020) and studies that reported on the performance of CNN in the diagnosis of H. pylori infection were selected. A random-effects model was used to calculate the pooled rates. In cases where multiple 2×2 contingency tables were provided for different thresholds, we assumed the data tables were independent from each other. RESULTS: Five studies were included in our final analysis. Images used were from a combination of white-light, blue laser imaging, and linked color imaging. The pooled accuracy for detecting H. pylori infection with AI was 87.1% (95% confidence interval [CI] 81.8-91.1), sensitivity was 86.3% (95%CI 80.4-90.6), and specificity was 87.1% (95%CI 80.5-91.7). The corresponding performance metrics for physician endoscopists were 82.9% (95%CI 76.7-87.7), 79.6% (95%CI 68.1-87.7), and 83.8% (95%CI 72-91.3), respectively. Based on non-causal subgroup comparison methods, CNN seemed to perform equivalently to physicians. CONCLUSION: Based on our meta-analysis, CNN-based computer-aided diagnosis of H. pylori infection demonstrated an accuracy, sensitivity, and specificity of 87%.