Your new experience awaits. Try the new design now and help us make it even better

ORIGINAL RESEARCH article

Front. Med.

Sec. Dermatology

Volume 12 - 2025 | doi: 10.3389/fmed.2025.1667794

Differential diagnosis of eczema and psoriasis using routine clinical data and machine learning: development of a web-based tool in a multicenter outpatient cohort

Provisionally accepted
Ning  DingNing Ding1Yinhao  LiYinhao Li2Zheng  ZhaoZheng Zhao2Xiangfu  MengXiangfu Meng2Mingqiang  SunMingqiang Sun3Xueqing  RenXueqing Ren4Ying  WangYing Wang1*
  • 1Department of Dermatology, Shengjing Hospital of China Medical University, Shenyang, China
  • 2Liaoning Technical University - Huludao Campus, Huludao, China
  • 3Shenyang dermatology hospital, Shenyang, China
  • 4The First Affiliated Hospital of Dalian Medical University, Dalian, China

The final, formatted version of the article will be published soon.

Background: Eczema and psoriasis are common chronic dermatoses with overlapping features, making early differential diagnosis difficult. While biopsy is the gold standard, its invasiveness and dependence on clinician expertise restrict routine application, especially in primary care. To overcome these limitations, we developed a machine learning-based diagnostic tool using routine laboratory data, enabling non-invasive, accurate, and practical differentiation between eczema and psoriasis in outpatient settings. Methods: We retrospectively analyzed clinical and routine laboratory data from 57,518 patients with eczema and psoriasis across three medical centers. Patients with confirmed diagnoses and complete laboratory records were included, while those with missing key data were excluded. Eight machine learning models were trained using data from Shengjing Hospital. Model performance was evaluated using accuracy, AUC, sensitivity, specificity, PPV, NPV, F1 score, and confusion matrix. The best-performing model, XGBoost, was externally validated on independent cohorts from two other hospitals. SHapley Additive exPlanation (SHAP) were applied to assess feature importance. Finally, a web-based tool was developed integrating the optimal model with Optical Character Recognition (OCR) for automatic data input. Results: XGBoost demonstrated the best performance, with AUCs of 0.891, 0.830, and 0.812 for the training, internal test, and external test sets, respectively. Key predictive features included dNLR, neutrophil count, SIRI, RDW, and eosinophil count, which were consistent with known clinical patterns. The final model was deployed as an interactive web tool, allowing manual or OCR-based data input to provide real-time prediction probabilities. Conclusions: This machine learning-based diagnostic tool showed strong performance and interpretability in differentiating eczema from psoriasis using routine laboratory data. The user-friendly web interface enables rapid, non-invasive decision support in outpatient clinical settings.

Keywords: Eczema, Psoriasis, machine learning, Clinical decision support, Web-based tool

Received: 22 Jul 2025; Accepted: 03 Oct 2025.

Copyright: © 2025 Ding, Li, Zhao, Meng, Sun, Ren and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

* Correspondence: Ying Wang, wydn20232023@163.com

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.