Ability of artificial intelligence to detect T1 esophageal squamous cell carcinoma from endoscopic videos and the effects of real-time assistance

Abstract Diagnosis using artificial intelligence (AI) with deep learning could be useful in endoscopic examinations. We investigated the ability of AI to detect superficial esophageal squamous cell carcinoma (ESCC) from esophagogastroduodenoscopy (EGD) videos. retrospectively collected 8428 EGD images cancer develop a convolutional neural network through learning. evaluated detection accuracy diagnosing system compared that 18 endoscopists. used 144 videos for two validation sets. First, we 64 observation ESCCs both white light imaging (WLI) and narrow-band (NBI). then 80 40 patients (20 ESCC 20 non-ESCC). In first set, correctly diagnosed 100% ESCCs. second it detected 85% (17/20) Of these, 75% (15/20) 55% (11/22) were by WLI NBI, respectively, positive predictive value was 36.7%. The endoscopists 45% (25–70%) With real-time assistance, sensitivities significantly improved without assistance (p < 0.05). can high sensitivity endoscopist support.