Related Experiment Video
Updated: Aug 6, 2026
06:08
Quantification of Hypopigmentation Activity In Vitro
Published on: March 6, 2019
Clinically Interpretable Deep Learning for Differentiating Vitiligo and Postinflammatory Hypopigmentation: Diagnostic
Amal Adel Alzu'bi1, Shadi Al Khateeb2, Bassam M Elzaghmouri3
1Department of Computer Information Systems, Faculty of Computer & Information Technology, Jordan University of Science and Technology, Irbid, 22110, Jordan, 962 790051705.
JMIR Medical Informatics
|July 24, 2026
Summary
This study developed an interpretable deep learning model to differentiate vitiligo from postinflammatory hypopigmentation (PIH). The framework achieved high accuracy, offering a transparent tool for diagnosing these similar-looking skin conditions.
Area of Science:
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background:
- Distinguishing vitiligo from postinflammatory hypopigmentation (PIH) is clinically challenging due to similar depigmented lesions.
- Deep learning shows promise in dermatology but lacks interpretability, hindering clinical use.
Purpose of the Study:
- To develop an interpretable deep learning framework for accurate differentiation between vitiligo and PIH.
- To utilize a lightweight convolutional neural network and ensemble explainability methods.
Main Methods:
- A dataset of 332 clinical images (176 vitiligo, 156 PIH) was used.
- A MobileNetV2 model was fine-tuned and combined with Grad-CAM, integrated gradients, and SmoothGrad for interpretability.
- 5-fold cross-validation was employed to assess performance using accuracy, precision, recall, F1-score, and AUC.
Main Results:
- The model achieved high diagnostic performance with an overall accuracy of 94.88% and an AUC of 0.9885.
- The ensemble explainability framework provided clinically meaningful explanations in 98.48% of cases.
- High performance metrics including precision, recall, and F1-score were observed across validation folds.
Conclusions:
- The proposed framework effectively combines high diagnostic accuracy with robust interpretability for differentiating vitiligo from PIH.
- The integration of multiple explanation methods enhances clinical transparency.
- This approach may aid dermatologists in the differential diagnosis of pigmentary disorders.