Authors
Y. Harth
Topics
Hair Growth and DisordersHerpesvirus Infections and TreatmentsDermatologic Treatments and Research198 Reliable and interpretable segmentation for remote assessment of atopic dermatitis severity using digital images K Pan , R Attar , G Hurault , H Williams and R Tanaka 1 Department of Bioengineering, Imperial College London, London, United Kingdom and 2 Centre of Evidence Based Dermatology, University of Nottingham, Nottingham, United Kingdom Assessing the severity of atopic dermatitis (AD) traditionally relies on a face-to-face assess ment by healthcare professionals and may suffer from inter- and intra-rater variability. With the increasing demand for better patient self-care and the growing popularity of telemedicine post-pandemic, the ability to assess AD severity remotely from digital images, such as those taken with a smartphone camera, is becoming increasingly important. Previously, we developed EczemaNet, a fully automated computer vision pipeline for detecting and assessing AD severity. It demonstrated good performance for assessing AD severity in real world images while being robust to suboptimal imaging conditions. However, we recognize that it had limitations in practical use due to its lack of interpretability in AD area segmen tation and to the need for more reliable AD segmentation data provided by specialists. At the same time, we also found a poor agreement for AD segmentation in digital images, with the average interclass correlation coefficient among four dermatologists being 0.45 on 80 digital images. To address these challenges, we improved EczemaNet pipeline to perform AD seg mentation in a more reliable and interpretable fashion. The new pipeline uses pixel-level segmentation and data augmentation to improve the quality and robustness of AD lesion detection. We achieved pixel-level AD segmentation using U-Net architecture and evaluated the reliability of the pipeline using various data augmentation methods such as Pix2Pix. Our investigation found that the use of whole-skin images for model training is a viable alternative as a data collection strategy, which would allow the data acquisition to be more cost-effective without affecting the system’s final performance.199 Computer vision AI-based androgenetic alopecia analysis using a novel mobileweb appY Harth 1,2 1 MDalgorithms, San Francisco, California, United States and 2 Dermatology, Ramat Yam, Medical Center, Herzlya, Israel Introduction: Mobile digital health is developing rapidly and is expected to dominate how we inform, treat, and monitor skin and hair conditions. We tested the world’s first direct-to consumer fully automated AI-based service app that uses image analysis to analyze the severity of androgenetic alopecia and monitor hair regrowth. Objectives: Compare the ac curacy of a fully automated mobile, AI-based, mobile-hair loss analysis to hair loss assessment by a human dermatologist. Materials and Methods: We tested MDhair.co, an AI-based analysis system (MDalgorithms Inc., San Francisco, USA), on 30 females with androgenetic alopecia. The users answered a few questions about their hair loss and used their smartphones to take a selfie of their heads. Savin’s grading for women’s hair loss was used by a machine learning model of the AI-based system and the human dermatologist. The severity score generated by the MDhair analysis system was compared to a blinded assessment by a board certified dermatologist. The data was collected and processed anonymously. Results: Data was collected from 30 women with a gradual development of hair loss compatible with fe male pattern androgenetic alopecia. The group included four women 18-29 years old, 13, 30-44, 12, 45-65, and one older than 65. Sixteen had straight hair, 10 had wavy hair, and 4 had curly hair. Based on human dermatologists assessing the hair loss severity, 17 had mild hair loss (Savin’s hair loss grades I-1 and I-2), and 13 had moderate hair loss (Savin’s hair loss grades II-1 and II-2). AIebased assessment matched the human dermatologist diagnosis in 28 of 30 cases (accuracy - 94%). 20% of the study participants underestimated the severity of their hair loss, and one overestimated the severity of hair loss. Conclusions: MDhair’s has been proven to be a valuable tool for assessing the severity of androgenetic alopecia. Our study has found that self-assessment resulted in a significant underestimation of hair loss. The automatic assessment by the MDhair system shows high accuracy in evaluating hair loss compared to a human dermatologist.200 Deep learning model to predict sentinel lymph node status in melanomapatientsT Okamoto1 , M Kawai2 , M Le2 , S Shimada
About
PublishedApr 17, 2023
TypeArticle
Citations0