A groundbreaking AI model developed through a collaboration between Keio University School of Medicine, Kyoto Prefectural University of Medicine, Teikyo University, and Atopiyo LLC now enables objective assessment of eczema severity using patient-uploaded smartphone photos. The findings, recently published in Allergy, the official journal of the European Academy of Allergy and Clinical Immunology, mark a major step forward in digital dermatology.
Atopic dermatitis (AD), a chronic skin condition known for recurring flare-ups, often demands continuous monitoring and treatment adjustments. Although modern smartphone apps and social media platforms have made it easier for patients to log their symptoms and progress, subjective symptom reports—like itching or sleep disruption—don’t always reflect the visible severity of the condition. This disconnect has emphasized the need for more objective and standardized assessment tools. Digital biomarkers are increasingly seen as a solution to bridge this gap.
To meet this need, the researchers utilized data from Atopiyo, Japan’s largest platform dedicated to AD. Since 2018, more than 28,000 users have uploaded over 57,000 photos and shared personal symptom-related comments. Drawing on this data, the team designed an AI model that combines three primary algorithms: one for identifying body parts, another for detecting eczema lesions, and a third for scoring severity using the Three Item Severity (TIS) scale—which assesses redness, swelling, and excoriation.
The model was trained on a dataset of 880 images paired with self-reported itch scores. In a subsequent validation phase with 220 new images, the AI-generated TIS scores (AI-TIS) showed a strong correlation with dermatologist-assessed TIS scores (R = 0.73, P < 0.001). It also demonstrated a meaningful connection with SCORAD, a standard objective scoring tool (R = 0.53, P = 0.04).
“Many patients with eczema struggle to evaluate their disease severity on their own,” said Dr. Takeya Adachi, the study’s lead author. “Our AI model enables objective, real-time assessment using just a smartphone, which empowers patients and could improve overall disease management.”
One notable finding was that the AI’s severity scores did not strongly correlate with patient-reported itch levels, underscoring the gap between subjective symptoms and visible inflammation. This further reinforces the need for digital biomarkers that can deliver more precise insights in dermatological care.
The team plans to broaden the model’s capabilities by including a more diverse range of skin types, expanding age demographics, and incorporating additional clinical scoring systems such as SCORAD and EASI. Their long-term vision is to develop comprehensive AI-based teledermatology tools that can support patients and healthcare professionals alike in everyday clinical practice.
Related Topics
