Objective erythema assessment of Psoriasis lesions for Psoriasis Area and Severity Index (PASI) evaluation

Psoriasis severity assessment is usually performed based on the computation of the Psoriasis Area and Severity Index (PASI). Physicians subjectively classify the erythema parameter into several grades of severity. To support the decision and the evaluation of the psoriasis lesions' evolution in time, this study proposes an approach for the objective assessment of erythema degree. There were processed seventeen images depicting psoriasis lesions from mild to severe and the erythema parameter was classified into three categories using machine learning algorithms. The classification was based on color and texture features extracted from the digital images. Three classifiers were trained and tested with these features: Naïve Bayes, Neural Networks and Support Vector Machine. A comparative analysis on the classification accuracy and computation time was made to the classification algorithms. The best classification accuracy (92%) was obtained when using a two-layer feed-forward Neural Network. The proposed method can be used to objectively assess the psoriasis lesion's erythema. The major interest of this approach is to be cheap, fast, robust and easy to use in a dermatological context, with very few constraints on the acquisition protocol.