Akademik Veri Yönetim Sistemi
1st International Future Engineering Conference, Şırnak, Türkiye, 25 - 26 Aralık 2023, ss.273
Deep learning represents an alternative approach to conventional diagnostic methods in the diagnosis of ear
diseases and offers the potential to achieve more effective and faster results. Deep learning algorithms are
known for their ability to analyze large datasets and recognize complex patterns. These algorithms, used in
the diagnosis of ear diseases, can recognize and classify symptoms by learning from the data rather than
relying on predefined features. This enables a more precise identification of the characteristic features of
different ear diseases and increases the potential for improving treatment processes. Especially in ear imaging
procedures, such as otoacoustic emissions or intracavitary cameras, deep learning serves as a valuable aid
for physicians to accurately diagnose and create appropriate treatment plans. Otoscopic examination is
particularly important for the diagnosis of diseases such as myringosclerosis, earwax, and chronic otitis
media. In this study, EfficientNet models were employed to classify these diseases, and the results showed a
remarkable success rate of 99% on the validation data during training. The test data also exhibited an
impressive accuracy rate of 95%. In the research, the dataset consists of training images from 160 patients
and test images from 20 patients, totaling 720 images across 4 classes. In future studies, it is planned to work
on new models by increasing the disease classes.