Machine learning for automated otoscopic tympanic membrane assessment

Objectives: To develop and evaluate a deep learning model for automated classification of tympanic membrane and middle ear conditions using otoscopic images. Study Design: Retrospective. Materials and Methods: This study trained an EfficientNet-based convolutional neural network (CNN) on 618 labelled images across five categories: normal tympanic membrane (TM), acute otitis media (AOM), otitis media with effusion (OME), middle ear atelectasis (MEA), and TM perforation. The dataset was split into training (70%), validation (20%), and test (10%) sets. Model performance was assessed using accuracy, sensitivity, specificity, F1 score, and ROC-AUC. Results: On the test set, the model achieved an overall accuracy of 88.98%, sensitivity of 91.65%, specificity of 96.37%, F1 score of 0.89, and ROC-AUC of 0.975. Grad-CAM visualisations confirmed the model’s focus on clinically relevant areas. Conclusions: This study demonstrates that a CNN-based model can accurately classify common middle ear pathologies and has potential as a diagnostic support tool in telemedicine and primary care settings.