AI Model Can Help Classify Atopic Dermatitis Lesion Severity

Artificial intelligence (AI) may be a valuable tool for objectively measuring atopic dermatitis (AD) severity, a new report suggests.¹ The study, which was published as a research letter in the journal Allergy, found that an AI algorithm can successfully score eczema severity, though the severity scores did not always align with patient-reported itch-intensity metrics.

The chronic nature of AD necessitates long-term management and monitoring of the disease. As smartphone technology has evolved, they said phones and social media have made it easier for patients to document their disease and also learn more about it.² In Japan, the authors noted, more than 28,000 users have shared more than 57,000 photos and comments about their symptoms on a popular AD platform called Atopiyo.¹

“However, discrepancies between patient-reported outcomes and objective disease severity measures highlight the need for standardized evaluation methods and the potential role of digital biomarkers,” they wrote.

The apparent divergence between subjective and objective measures of atopic dermatitis severity underscores the need for more precise assessment methods, the authors explained. | Image credit: Evgeniya Primavera - stock.adobe.com

The new report documents the investigators’ attempt to use AI to help fill the gap between patient reports and objective measures of severity. The team created a set of 3 algorithms to analyze user-generated photos. The first algorithm was trained to detect body parts, since user-generated photos usually do not show the patient’s entire body. The second algorithm was built to detect lesions in photographs. The final algorithm was designed to assess the severity of the lesions using the Three Item Severity (TIS) scoring system, which assesses severity on a scale of 0 to 9.

To validate their findings, the investigators used the SCORing Atopic Dermatitis (SCORAD) assessment, which incorporates both objective measures like intensity parameters and subjective measures like pruritus. Users of the AI algorithm were instructed to assess their itch intensity using the Itch-NRS-5 (numeric rating scale) scoring system when they uploaded each photo.

Using a database of images uploaded between August 2018 and January 2024, the investigators identified 9656 images from 900 participants that had itch scores and were clear enough to be analyzed by the model. The AI model had a 98% accuracy rate at detecting body parts and was 100% accurate in detecting eczema areas.

In the case of 220 images, patients who submitted photos also had their lesions assessed by an allergist or dermatologist. The TIS scores generated by the AI algorithm correlated well with the scores determined by physicians (R = 0.73; P < .001).

However, the model’s severity scores showed a weak correlation with patient-reported Itch-NRS-5 scores (R = 0.11; P < .001) in an analysis of 8556 images. The investigators said this suggests that disease severity and pruritus do not always correspond. Indeed, a 2021 analysis suggested that sometimes severe itch is associated with lesions that, according to SCORAD, are mild or moderate in severity.³ The opposite—sometimes severe lesions are associated with only mild or moderate itching—is also true, that study found.

The apparent divergence between subjective and objective measures of disease severity underscored the need for more precise assessment methods, the authors explained, adding that AI could offer a solution.¹

“The AI model developed in this study has the potential to help patients with AD objectively assess their skin condition, facilitating timely and appropriate treatment,” they wrote. However, they added that their models would need to be validated using a more diverse data set, including people with different skin types.

References

1. Okata-Karigane U, Hirota M, Takahashi C, et al. AI-based objective severity assessment of atopic dermatitis using patient photos in a real-world setting: a digital biomarker approach. Allergy. Published online May 19, 2025. doi:10.1111/all.16586

2. Sy W, Bhayana M, Lamb AJ. Atopic dermatitis disease education. Adv Exp Med Biol. 2024;1447:209-215. doi:10.1007/978-3-031-54513-9_17

3. Chovatiya R, Lei D, Ahmed A, Chavda R, Gabriel S, Silverberg JI. Clinical phenotyping of atopic dermatitis using combined itch and lesional severity: A prospective observational study. Ann Allergy Asthma Immunol. 2021;127(1):83-90.e2. doi:10.1016/j.anai.2021.03.019