AI-Based Point-of-Care Diagnosis Using Otoscopy Images

This paper presents an AI-based point-of-care diagnostic system that leverages otoscopy images to facilitate the early detection of ear-related pathologies. The proposed approach addresses the limitations posed by small clinical datasets by incorporating a comprehensive data augmentation strategy, which significantly increases the effective training sample size. The convolutional neural network (CNN) architecture comprised multiple layers, enabling the extraction of image features. The proposed model achieved a high classification test accuracy of 98.43%, with strong performance metrics including high precision, recall, and F1-scores. Although results are promising, the study recognizes that reliance on synthetic augmentation underscores the necessity for larger datasets to improve the model’s robustness and generalizability in real-world scenarios. The findings underscore the importance of deep learning methodologies in transforming point-of-care diagnostics and provide a foundation for future research focused on refinement of automated otoscopic image analysis.