Objective assessment of psoriasis treatment through skin images

© 2014 Dr. Juan LuPsoriasis, characterised by white rough scaling and red inflamed erythema, is a common chronic inflammatory skin disease without a known cure. Since there is no consensus on the current subjective assessment methods due to the unavoidable inter- and intra- observer variances, an objective assessment of psoriasis treatment efficacy is urgently needed. The aim of this thesis is to build a severity scoring system through 2D digital skin images to evaluate the treatment efficacy of psoriasis reliably, automatically and reproducibly. Previous research focuses on the segmentation of plaque psoriasis, the assessment of its severity and the assessment of severity changes in a short time. What is much needed is a general method for assessing the treatment efficacy of psoriasis, not only applicable to plaque psoriasis, but also applicable to other types of psoriasis for different points in time. This thesis attempts to fill the gap by considering psoriasis segmentation as erythema segmentation and scaling segmentation separately. Moreover, additional psoriasis severity features are explored and models are built for severity change analysis during a long-term treatment. This thesis contributes to: • An erythema segmentation method is developed to segment out erythema from the psoriasis skin images. This research applies a skin model building on composition of skin pigments to do the colour analysis. The erythema segmentation accurately locates the position of erythema in digital skin images, and is not restricted to just plaque psoriasis. • A scaling segmentation method is proposed to segment out scaling from psoriasis skin images. Scaling often comes along with erythema, but sometimes scaling appears alone. As far as we know, this is the first study to accurately detect the position of scaling in the skin images for both situations, while previous works require the segmentation of psoriatic lesions as a first step. • Features of erythema and scaling severity are proposed for the assessment of psoriasis severity. The erythema severity features are associated with the haemoglobin and melanin pigments; the scaling severity features are associated with the relative scaling area and the roughness degree of the scaling. The severity features correlate well with dermatologists' observation, and are shown to be better than existing methods. • Evolution of a psoriatic lesion is evaluated in a long-term treatment. The evaluation uses a number to indicate the degree of changes of a psoriatic lesion. This research allows direct measurement of treatment efficacy in a long-term treatment rather than existing works that are only applicable to a short-term treatment. This thesis presents the first work to reliably evaluate the treatment efficacy of a general psoriatic lesion. It shows the potential of using a computer-aided image processing system to objectively and quantitatively evaluate the psoriasis severity. It is hoped that based on this work, the future research will advance the psoriasis diagnosis as well as the treatment research