A Soft Computing Approach for Efficient Diagnosis of Otitis Media Infection by Mucosal Disease Early Detection and Referrals

Otitis Media (OM) represents a prevalent and potentially severe middle ear condition, primarily affecting children. Timely and accurate OM diagnosis is crucial for effective intervention. Often it happens due to mucosal diseases affected by mucosal membranes. This study explores the application of deep learning techniques to tympanic membrane (TM) images for precise OM detection. the dataset, collected and verified from AIIMS Raipur Hospital, comprises four distinct image categories: Chronic Otitis Media (COM), Earwax plug (EP), Myringosclerosis (MS), and normal Tympanic Membrane (TM). Each category encompasses 40 meticulously selected images for training and 40 for testing. Binary and multi-class classification 3-layer models were developed and rigorously assessed using standard performance metrics, including accuracy, sensitivity, specificity, precision, and Fl-score. The binary classification models consistently demonstrated impressive accuracy, surpassing 95%, in distinguishing various OM types from normal TMs. Notably, the COM vs. Normal model achieved an accuracy of 97.03%. Furthermore, the multi-class classification model exhibited an overall accuracy of 89.45% for categorizing TMs into COM, EP, MS, and normal classes.