Determination of Tympanic Membrane Region in the Middle Ear Otoscope Images with Convolutional Neural Network Based YOLO Method

Abstract

Due to inflammation of the middle ear, various deformations occur in the eardrum. In order todiagnose the disease, it is necessary to examine the tympanic membrane in detail with an otoscope.In recent years, deep learning has been applied in many areas including biomedical field and veryeffective results have been achieved. Deep learning based methods are used successfully in automaticobject detection. In this study, a deep learning based object detection method namely You Only LookOnce (YOLO), is used for automatic detection of tympanic membrane in eardrum images obtainedusing otoscope device. To enable automatic detection of tympanic membrane by YOLO, experimentalstudies were conducted with AlexNet, VGGNet, GoogLeNet and ResNet. According to the performanceresults, the most efficient results were obtained with ResNet and VGGNet architectures. Tympanicmembrane region detection with YOLO, was performed with an accuracy rate of 93%.