Figure 2
From: Weakly supervised end-to-end artificial intelligence in gastrointestinal endoscopy
Classification performance for end-to-end artificial intelligence in the gastroscopy and colonoscopy data set. (A) Classifier performance in the gastroscopy data set as measured as the mean area under the receiver operating curve (AUROC) of three cross-validation experiments for each disease category with 95% confidence intervals. (B) Receiver operating curve (ROC) for C15, (C) for K91, (D) for B37 and (E) for R13. (F) AUROCs for all disease categories in the colonoscopy data set and ROCs for (G) disease category I84, (H) for K91, (I) for K51 and (J) for C20.
