ACCURATE SEGMENTATION OF PSORIASIS DISEASES IMAGES USING K-MEANS ALGORITHM BASED ON CIELAB (L*A*B) COLOR SPACE

摘要

Context: Psoriasis turned out to be one of the debilitating and enduring inflammatory skin diseases. Often misinterpreted as a casual skin disease, it is estimated that approximately 125 million people worldwide suffers due to this infection. The case is made worse when there is no known cure in the status quo. The communal category of psoriasis has been considered as abruptly demarcated scaly and erythematous plaque at patient's skin. This disease could ensue anywhere on the human body. Problem: Diagnosis of psoriasis requires an experienced specialist in the field of dermatology because of the presence of other skin diseases similar to a large extent which lead to majority cases of an error in diagnostic. As doctors are still mere human and depends on factors such as eye and physical touch that is not error free. In addition, the drugs for psoriasis disease contain quantities of Chemical materials dangerous to other body organs that may put the functionality of critical organs such as the liver and spleen in jeopardy. Meanwhile, over-treatment leads to loss of life of the patient so it must be re-diagnosis multiple times until the confirmation of a high proportion of the dangerous disease. Time is not the greatest threat for this disease rather the accuracy of diagnosis is much crucial and the accuracy of diagnostic plays a pivotal role in combating this atrocious disease. Regular re-diagnosis is considered a must in order to ensure the survivability of patients from the threat it poses. However, re-diagnosis often consumed a great amount of financial expenditure just to ensure that it is indeed a disease of psoriasis and that the appropriate treatment is given may only lead to another issue which is a financial deficiency. Approach: In this paper, the researcher is interested in separating the image and concentrate on the lesion region and extricating disease district. The process itself is an enormous challenge in light of the fact that there is no discovery of this minute segmentation algorithms division executes and all in all dataset. The proposed strategy is based on K-Means clustering as initial segmentation and gets a divided region, including areas of diseased and the proposed K-Means based on CIE Lab L*a*b color spaces instead of using Red, Green and Blue (RGB) color space. Post segmentation based on color feature will be filtered out as non-interesting objects. Finding: The findings from this study have shown that: Firstly the method is depending on the L*a*b color spaces instead of using RGB color spaces, secondly, the method is based on color feature to select disease region of psoriasis or the correct object. The results of this research confirmed that this method works effectively where we have been implementing this method on a database containing 80 medical images of RGB psoriasis diseases image and shows the accuracy of this method was at 95% when we did a comparison between our method and other ways to find that the proposed strategy gives more effective results in the segmentation. The researcher compared accurate segmentation of K-Means cluster formation with color spaces L*a*b on medical imaging and K-Means cluster formation with color spaces RGB on the same images.