Classification of Otitis Media Infections using Image Processing and Convolutional Neural Network

Developing countries still to this day suffer from misdiagnosis of otitis media infections. Various studies to solve this issue with various success rates. This study explores a different variation of convolutional neural network (CNN), YOLO V3, or Version 3 of You Only Look Once. This algorithm detects particular objects in various forms of media, and one of them is images. Considering it is designed to detect specific objects, it was the perfect candidate to test on detecting Acute Otitis Media (AOM) and Chronic Suppurative Otitis Media (CSOM). These two variations have an object to look for whenever a doctor is diagnosing a case. Inflammation of the middle ear or otitis media (OM) are separate disease entities but may overlap. Hence, it may be confusing for a newly trained doctor to diagnose it correctly. This study achieved an accuracy rate of 75% when 20 images of AOM, CSOM, and normal tympanic membrane were tested. This result can be improved by adding more images into the training datasets using the same camera used in testing. Another appealing feature of YOLOV3 is the low cost of development and the availability of documentation on using and improving it.