Clustering and classification of dermatologic data with Self Organization Map (SOM) method

Nowadays, skin diseases have increased due to various external effects such as chemical, radiological and so on. In dermatology, the distinguished diagnosis of Erythemato-squamos diseases is a situation that doctors often confront. When many skin diseases are examined, it is seen that many of them are quite similar in shape and appearance although their reasons of emergence are different. Doctors try to distinguish diseases from each other and diagnose by evaluating the clinical findings with pathological parameters. It is observed that many researchers have conducted studies on the Erythemato-Squamous diseases to develop decision support systems using different classification algorithms for detection and diagnosis. Unlike the cited studies in literature, the aim of the present study is to extract Self Organization Maps (SOM) of clinical and pathological findings and investigate cluster of condition various diseases from reduced data. SOM is a size reduction process which aim to simplify the problem. Basically, SOM provides less size reduction output using multidimensional input. In this study, the clinical and pathological classification was realized separately and together. As the result, classification of six types of Erythamato-Squamos skin disease was performed with SOM artificial intelligence application. In addition, clinical and pathological effects of SOM application was seen clearly by showing as a graphically display instead of a matrix. As a result, in the diagnosis of Erythemao-Squamos diseases, it was determined that a dermatologist diagnose mostly depending on the clinical findings although pathological findings contain quantative data.