Classification of Otoendoscopic Images with the Developed Textural Based Artificial Intelligence Model

Abstract

Many ear diseases can be diagnosed with the findings obtained by otoendoscopic examination, thus enabling early diagnosis and treatment. The inadequacy of accurate diagnosis rates in primary health care institutions and the difficulties in reaching otolaryngologists in rural areas can cause delays in the diagnosis of otological pathologies and sometimes complications. These obligatory needs make it even more important to use computer-aided systems to accurately identify ear diseases. In this paper, a texturalbased hybrid model has been developed for the classification of otoscope eardrum images. In the developed model, feature extraction was performed using LBP and HOG methods. From the features obtained from the feature maps obtained by the LBP and HOG methods, 500 features were selected each by using the Relief method. After the selected features were combined, the combined feature map was classified at different classifiers accepted in the literature. In the study, an accuracy value of 93.8% was obtained. This result demonstrates the potential of the suggested methodology for categorizing otoscope eardrum images.