eSkinHealth: A Multimodal Dataset for Neglected Tropical Skin Diseases
2025-08-27
Summary
The article presents eSkinHealth, a new dermatological dataset specifically designed to address the data scarcity in diagnosing Neglected Tropical Diseases (NTDs) prevalent in West Africa. Comprising 5,623 images from 1,639 cases across Côte d’Ivoire and Ghana, the dataset focuses on 47 skin diseases and integrates AI-expert collaboration for multimodal data annotation, including semantic masks, captions, and clinical concepts to enhance AI-driven diagnostic tools.
Why This Matters
This dataset is vital because it addresses critical gaps in existing dermatological data, particularly for underrepresented populations affected by skin NTDs. By providing a resource that includes rich, diverse data and expert-verified annotations, eSkinHealth can significantly advance the development of more equitable, accurate, and interpretable AI tools for global dermatology, offering potential improvements in diagnostic accessibility and efficiency.
How You Can Use This Info
Healthcare professionals and AI developers can leverage eSkinHealth to train and fine-tune AI models that are more culturally and demographically relevant for diagnosing skin conditions in West African populations. Additionally, the dataset can be used for benchmarking AI models, developing specialized medical image captioning tools, and exploring parameter-efficient fine-tuning methods to adapt large models to specific tasks efficiently.