2026-10-14T10:34:50Z https://data.mendeley.com/oai

oai:data.mendeley.com/5xtc2z3vs2.1 2026-07-08T17:25:09Z

Wan, Li Non-invasive prediction of Dupilumab response in facial Atopic Dermatitis lesions based on early post-treatment Reflectance Confocal Microscopy features Mendeley Data Background: Non-invasive prediction of the efficacy of dupilumab on facial lesions in patients with atopic dermatitis (AD) remains poorly defined. Objectives: This study aimed to explore reflectance confocal microscopy (RCM), a non-invasive method, for predicting dupilumab efficacy in treating facial AD lesions and to realize effective prediction via deep learning. Methods: Data from a prospective study included 49 AD patients received dupilumab treatment conducted between May 2023 and June 2025 were collected and analyzed. Patients were categorized into 'responder' and 'nonresponder' groups based on whether achieved EASI75 at 16 week. RCM was utilized to evaluate facial lesions prior to and following treatment to identify intergroup differences. Deep learning was employed to construct a prediction model. Results: Dermal papillary dilation (AUC=0.83, P<0.001) and tortuous papillary capillary dilation ((AUC=0.81, P<0.001) at 4 weeks post-treatment correlated significantly with therapeutic non-response. The ResNet101-based deep learning model, trained on RCM images from the 4-week interval, demonstrated robust predictive performance across 5-fold cross-validation, achieving a mean accuracy of 0.7280 ± 0.0215 and a final test AUC of 0.796. The model specifically identified non-responders with a precision of 0.8120 ± 0.0153 and an F1-score of 0.7400 ± 0.0239. Saliency heatmaps further visualized the model's decision-making basis, aligning with key pathological RCM features. Limitations: The present study is characterized by a relatively modest sample size, which is consistent with the exploratory nature of this investigation. Conclusions: Early post-treatment RCM features can effectively predict the efficacy of dupilumab in facial AD lesions combined with deep learning. Supplementary Data Figures for Publication Image Classification Dataset doi:10.17632/5xtc2z3vs2.1 oai:data.mendeley.com/5xtc2z3vs2.1 Creative Commons Attribution 4.0 International http://creativecommons.org/licenses/by/4.0 https://data.mendeley.com/datasets/5xtc2z3vs2 2026-07-08T17:25:09Z