Automatic Scale Severity Assessment Method in Psoriasis Skin Images Using Local Descriptors.

Area of Science:

• Dermatology and Medical Imaging
• Computer Vision and Machine Learning

Background:

• Psoriasis is a chronic skin condition requiring clinical assessment of erythema, scales, induration, and area.
• Accurate assessment of psoriasis severity is crucial for effective treatment and patient management.

Purpose of the Study:

• To introduce a scale severity scoring framework for two-dimensional psoriasis skin images.
• To evaluate the performance of different feature descriptors and machine learning classifiers for psoriasis scale scoring.

Main Methods:

• Leveraged the bag-of-visual words (BoVWs) model for lesion feature extraction using superpixels.
• Employed three-class machine learning classifiers, including support vector machine (SVM) and random forest.
• Examined eight local color and texture descriptors, including color histogram, local binary patterns, and color and edge directivity descriptor (CEDD).

Main Results:

• Color descriptors demonstrated the highest performance for scale severity scoring, followed by combined color and texture descriptors.
• The K-means algorithm showed superior results in vocabulary building compared to Gaussian mixed models.
• The proposed method achieved a scale severity scoring accuracy of 80.81% using specific parameters.

Conclusions:

• The developed framework effectively scores psoriasis scale severity using image analysis.
• Color and combined color-texture descriptors are most effective for this task.
• The findings support the use of automated image analysis for psoriasis assessment.