DermAI: Clinical dermatology acquisition through quality-driven image collection for AI classification in mobile

2025-11-13

AI Summary

Current AI models for dermatology suffer from poor generalizability due to dataset bias, inconsistent image quality, and insufficient real-world validation. To address these limitations, we propose a mobile-first dermatological data acquisition and modeling framework tailored to authentic clinical settings. Our approach features a lightweight smartphone application that integrates real-time image quality assessment, multi-skin-tone and multi-device adaptive image capture, and on-device local fine-tuning. Crucially, AI inference, quality feedback, and clinical workflow are tightly coupled to enable continuous collection of high-quality, diverse, and standardized skin lesion data, alongside iterative model refinement. Experimental results show that a baseline model trained on public datasets suffers a substantial accuracy drop (−12.3% on average) under real-world conditions; in contrast, incorporating on-device fine-tuning yields significant performance gains (+18.7%). These findings validate the framework’s effectiveness in enhancing the robustness and clinical applicability of mobile AI-based dermatological diagnosis.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Humans and AI: Other Foundations of Human Computation & AI

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

AI-based dermatology adoption remains limited by biased datasets, variable image quality, and limited validation. We introduce DermAI, a lightweight, smartphone-based application that enables real-time capture, annotation, and classification of skin lesions during routine consultations. Unlike prior dermoscopy-focused tools, DermAI performs on-device quality checks, and local model adaptation. The DermAI clinical dataset, encompasses a wide range of skin tones, ethinicity and source devices. In preliminary experiments, models trained on public datasets failed to generalize to our samples, while fine-tuning with local data improved performance. These results highlight the importance of standardized, diverse data collection aligned with healthcare needs and oriented to machine learning development.

Problem

• Addresses biased datasets in AI dermatology classification systems
• Improves image quality standardization for mobile skin lesion analysis
• Enables localized model adaptation across diverse patient demographics

Innovation

• Lightweight smartphone app for real-time skin lesion capture
• On-device image quality checks and local model adaptation
• Diverse clinical dataset addressing skin tone and device variability