Efficient and accurate diagnosis of otomycosis using an ensemble deep-learning model

Area of Science:

• Otolaryngology
• Medical Mycology
• Artificial Intelligence in Medicine

Background:

• Otomycosis, a fungal external ear infection, accounts for over 15% of external otitis cases globally.
• Common in humid regions and cultures with ear-cleaning practices, it is primarily caused by Aspergillus and Candida species.
• Accurate diagnosis relies on early endoscopic and microbiological examinations.

Purpose of the Study:

• To introduce a novel machine-learning model for the accurate and rapid diagnosis of otomycosis.
• To develop a computer-aided decision-making system utilizing deep learning for otomycosis detection.
• To differentiate otomycosis from impacted cerumen, external otitis, and normal ear canal conditions.

Main Methods:

• A deep-learning model was developed, comprising a Java web application and an image classification subsystem.
• The system was trained using otoendoscopic images and confirmed fungal pathogens (Aspergillus, Candida) from clinical samples.
• The EfficientNetB6 model was selected after evaluating ResNet, SENet, and EfficientNet variants, using 2,182 training and 475 validation images.

Main Results:

• The developed deep-learning system achieved an average accuracy of 92.42% on the validation dataset.
• The EfficientNetB6 model demonstrated optimal performance in terms of accuracy and operational speed.
• The system outputs probability values for four categories: otomycosis, impacted cerumen, external otitis, and normal cases.

Conclusions:

• The proposed deep-learning system offers a quick and accurate diagnostic reference for otomycosis.
• This AI-powered tool can assist general practitioners in improving the diagnostic accuracy of fungal ear infections.
• The study highlights the potential of machine learning in enhancing the clinical management of otomycosis.