Automatic Urticaria Activity Score: Deep Learning-Based Automatic Hive Counting for Urticaria Severity Assessment

Area of Science:

• Dermatology
• Artificial Intelligence
• Medical Imaging

Background:

• Chronic urticaria affects 1% globally, with chronic spontaneous urticaria being most common.
• The Urticaria Activity Score (UAS) assesses disease severity but is manual, time-consuming, and prone to variability.
• Objective assessment is crucial for guiding treatment and evaluating therapies.

Purpose of the Study:

• To develop an automated Urticaria Activity Score (UAS) using deep learning.
• To validate the performance of the automated UAS against expert physicians.
• To explore the potential of AI in clinical decision support and clinical trial endpoints.

Main Methods:

• A deep learning model, Legit.Health-UAS-HiveNet, was developed for lesion detection and UAS calculation.
• The model was trained and evaluated on chronic urticaria patient data.
• Performance was compared to manual UAS assessments by expert physicians.

Main Results:

• The Automatic UAS model demonstrated performance comparable to expert physicians in assessing chronic urticaria severity.
• The AI model offers a consistent and objective method for UAS assessment.
• The system has potential for integration into Computer-Aided Diagnosis (CADx) systems.

Conclusions:

• Artificial intelligence, specifically deep learning, can effectively automate UAS assessment for chronic urticaria.
• AI-powered tools can support clinicians in daily practice and serve as novel endpoints in clinical trials.
• This technology advances evidence-based medicine by empowering clinicians and standardizing trial outcomes.