Classification of Dermatological Data with Self Organizing Maps and Support Vector Machine

Creative Commons License

Fidan U., Uzunhisarcıklı E., Çalıkuşu İ.

Afyon Kocatepe Üniversitesi Fen ve Mühendislik Bilimleri Dergisi, vol.19, no.3, pp.894-901, 2019 (Peer-Reviewed Journal)


The frequency incidence of dermatological diseases is increasing in parallel with the fact that human skin is exposed to different chemicals. Examined many skin diseases, many of them are similar in shape and appearance, although the reasons for their appearance are different. In dermatology, the differential diagnosis of Erythemato-squamous diseases is frequently encountered by doctors. Doctors try to differentiate and diagnose diseases by evaluating clinical findings and histopathological parameters together. Many researchers have developed different algorithms on the classification and clustering of diseases and data that have been diagnosed from the UCI database. In the present study, unlike previous studies, clinical and histopathological findings of 6 different Erythamo Squamos skin diseases were clustered by applying to SOM network separately. As a result of this clustering process, it is determined that Psoriasis - Cronic Dermatitis and Seborreic Dermatitis - Pitriasis Rosea diseases were found in the same cluster and the diagnoses are confused. In order to prevent this confusion, clinical and histopathological findings of the diseases were clustered by SOM method. Clustering parameters of clinical and histopathological findings were classified with SVM. As a result of the study, it was achieved that the classification of Psoriasis - Cronic Dermatitis diseases was classified as 0.89 with an accuracy of 0.93 and that of Seborreic Dermatitis - Pitriasis Rosea with an accuracy of 0.79 and 0.80.