129 Fully automated assessment of Atopic Dermatitis severity from real-world digital images

Assessing the severity of Atopic Dermatitis (AD) traditionally relies on face-to-face assessments by healthcare professionals. Such approaches are resource-intensive for participants and staff, challenging during pandemics, prone to inter- intra-observer variation. We aim investigate what extent computer vision algorithms can help standardise automate detection assessment AD using real-world digital images, without human intervention. developed EczemaNet, a deep learning pipeline detect assess from camera images. first trained model that lesions in images data provided four dermatologists who delineated (“segmented”) regions 1345 287 children. then second seven disease signs identified. EczemaNet demonstrated good performance assessing while being robust poor imaging conditions. noted inter-rater reliability segmentation dermatologists, i.e. rarely reached consensus location potential Nevertheless, we highlighted challenge accurately detecting lesions. It may limit attempting