Innovative AI Smartphone Tool for Assessing Eczema Severity in Real-World Settings

Researchers from Keio University School of Medicine, Kyoto Prefectural University of Medicine, and Teikyo University have developed a cutting-edge artificial intelligence (AI) system designed to evaluate the severity of eczema through photographs taken with smartphones. This advancement addresses the need for more objective and standardized assessment methods for atopic dermatitis (AD), a persistent chronic skin condition characterized by flare-ups that often require continuous monitoring and therapeutic adjustments.

The AI model utilizes data from Atopiyo, Japan's largest platform dedicated to AD, where over 28,000 users have uploaded more than 57,000 images and comments since 2018. The system incorporates three core algorithms: detection of specific body parts, identification of eczema lesions, and severity assessment based on the Three Item Severity (TIS) scale, which considers redness, swelling, and excoriation.

In testing, the AI was trained on a dataset of 880 images with self-reported itch scores, demonstrating high accuracy. Validation with 220 images revealed a strong correlation between AI-generated severity scores and dermatologist assessments (correlation coefficient R = 0.73), as well as a significant association with the objective SCORAD scoring system (R = 0.53). This signifies that the AI is effective in objectively quantifying eczema severity using simple smartphone images.

Dr. Takeya Adachi, the study’s lead author, emphasized the potential impact: "Many eczema patients struggle to self-assess their condition. Our AI tool offers an objective, real-time evaluation that can be performed conveniently via smartphone, empowering patients with better disease management." The findings also uncovered a weak correlation between AI-based severity scores and patients’ subjective itch reports, highlighting the discrepancy between perceived symptoms and observable inflammation, thereby advocating for the importance of digital biomarkers in dermatological care.

The team plans to refine the model to include diverse skin types, age ranges, and integrate additional clinical features from established scoring systems like SCORAD and EASI. This progress suggests a promising future for AI-powered teledermatology, enhancing the collaboration between patients and healthcare providers outside traditional clinical settings.

The full study is available in the journal Allergy (2025). More information can be found at source.