Enhancing Atopic Dermatitis Assessment: The Role of Automated Diagnostics

The use of ultrasound in dermatology has gained attention in recent decades, particularly for conditions like atopic dermatitis (AD). AD is a chronic inflammatory skin disease that causes dry, itchy, and inflamed eczematous lesions, often appearing in areas where the skin flexes, such as hands, feet, or the inner folds of the knees and elbows. AD, like many skin conditions, is commonly perceived as a superficial condition and traditionally relies on visual examination; however, what happens beneath the surface can be equally significant.

The subepidermal low echogenic band (SLEB) is a characteristic feature of AD that indicates inflammation within the skin. High-frequency ultrasounds (HFUS) can measure SLEB formation and growth, providing a more comprehensive tool for monitoring disease progression and treatment response. However, HFUS for SLEB measurement is underutilized due to the time-consuming manual analysis, poor objectivity and repeatability, and specialized expertise required for image analysis and interpretation. Recent research has aimed to address these gaps, allowing for the use of HFUS in AD management.

Testing Automated HFUS in AD Patients

A new study developed and evaluated a computer-aided diagnostic (CAD) system for assessing AD using HFUS images of SLEB. The proposed method leverages advanced machine learning techniques in a two-step automated framework:

1. Segmenting HFUS images to detect inflamed areas
2. Classifying severity using the Investigator Global Assessment (IGA) five-point scale

The study included 80 patients diagnosed with AD, with severity assessed by visual examination and through interviews before and after treatment using the IGA five-point scale. The patients’ HFUS data included 20 MHz images before and after treatment, and images were annotated twice by two expert clinicians under blinded conditions. The results showed that the CAD performed at a level comparable to human experts:

• SLEB detection: The automated system achieved a 98 percent accuracy in detecting SLEB presence, slightly outperforming the experts' 96 percent accuracy.
• AD severity classification: The overall accuracy of AD assessment was 69 percent for the CAD system, compared to 64-70 percent for experts’ assessment.

Implications for AD Management

An automated approach to examining and diagnosing AD HFUS images would help support clinical decision-making and provide more objective assessments. As automated systems continue to evolve for medical purposes, they hold promise for enhancing dermatological care by improving efficiency and broadening access to advanced diagnostic methods.

References:

Czajkowska, J., Juszczyk, J., Bugdol, M. N., Glenc-Ambroży, M., Polak, A., Piejko, L., & Pietka, E. (2023). High-frequency ultrasound in anti-aging skin therapy monitoring. Sci Rep, 13(1), 17799. https://doi.org/10.1038/s41598-023-45126-y

Czajkowska, J., Polańska, A., Slian, A., & Dańczak-Pazdrowska, A. (2025). The usefulness of automated high frequency ultrasound image analysis in atopic dermatitis staging. Sci Rep, 15(1), 163. https://doi.org/10.1038/s41598-024-84051-6

Dini, V., Iannone, M., Michelucci, A., Manzo Margiotta, F., Granieri, G., Salvia, G.,…Romanelli, M. (2023). Ultra-High Frequency UltraSound (UHFUS) Assessment of Barrier Function in Moderate-to-Severe Atopic Dermatitis during Dupilumab Treatment. Diagnostics (Basel), 13(17). https://doi.org/10.3390/diagnostics13172721

Hurault, G., Pan, K., Mokhtari, R., Olabi, B., Earp, E., Steele, L.,…Tanaka, R. J. (2022). Detecting Eczema Areas in Digital Images: An Impossible Task? JID Innov, 2(5), 100133. https://doi.org/10.1016/j.xjidi.2022.100133