Observer-independent assessment of psoriasis-affected area using machine learning

Area of Science:

• Dermatology
• Medical Imaging
• Machine Learning

Background:

• Psoriasis severity assessment is subjective and lacks objective tools.
• High-cost psoriasis therapies require accurate, objective severity measures for reimbursement.

Purpose of the Study:

• Develop an objective psoriasis assessment method using machine learning image segmentation.
• Establish an accurate and reproducible tool for evaluating psoriasis severity.

Main Methods:

• Retrospective diagnostic accuracy study using 259 standardized patient photographs.
• Machine learning algorithm trained and validated on 203 images, tested on 56.
• Algorithm's lesion area assessment compared to manual markings and dermatologist estimates.

Main Results:

• Algorithm achieved >90% accuracy in 77% of tested images.
• Average difference between algorithm and manual marking was 5.9%.
• Average difference between algorithm and physician estimates was 8.1%.

Conclusions:

• Machine learning technology shows significant potential for objective psoriasis assessment.
• This method offers an objective alternative to the subjective Psoriasis Area and Severity Index (PASI).
• Further development is recommended to establish this as a reliable assessment tool.