Real-time use of artificial intelligence in the evaluation of cancer in Barrett’s oesophagus

摘要: Based on previous work by our group with manual annotation of visible Barrett oesophagus (BE) cancer images, a real-time deep learning artificial intelligence (AI) system was developed. While an expert endoscopist conducts the endoscopic assessment BE, AI captures random images from camera livestream and provides global prediction (classification), as well dense (segmentation) differentiating accurately between normal BE early oesophageal adenocarcinoma (EAC). The showed accuracy 89.9% 14 cases neoplastic BE. This paper follows up prior publication application in evaluation BE.1 2 In initial publications, we developed computer-aided diagnosis (CAD) model demonstrated promising performance scores classification segmentation domains during assessment.1 However, these results were achieved optimal which may not mirror real-life situation sufficiently. To enable seamless integration AI-based image into clinical workflow, further to increase speed analysis for resolution prediction, shows color-coded spatial distribution probabilities.1 Still based convolutional neural nets (CNNs) residual net (ResNet) architecture DeepLab V.3+, state-of-the-art encoder–decoder network adapted.3 transfer system, capture card (Avermedia, Taiwan) plugged monitor. Online supplementary video 1 setting endoscopy room University Hospital Augsburg (figure 1). can be started at any time using either button keyboard or foot switch. clip examples …