AI Tool Enhances Accuracy in Diagnosing Acute Otitis Media in children, reveals JAMA study

This study was published in JAMA Pediatrics by Nader Sheikh and colleagues. Despite being frequently diagnosed in children, the accuracy of AOM diagnosis remains suboptimal. This leads to potential overdiagnosis or underdiagnosis, impacting patient care and healthcare resources. AI-based approaches offer promising solutions to improve diagnostic accuracy and streamline clinical decision-making in AOM.

The study analyzed otoscopic videos of the tympanic membrane captured using smartphones during outpatient clinic visits in Pennsylvania from 2018 to 2023. Children presenting for sick visits or wellness visits were included in the study. Two AI models were developed: a deep residual-recurrent neural network and a decision tree network. The models were trained to predict AOM vs. no AOM based on features of the tympanic membrane observed in the videos. The accuracy of the AI models was compared, and a noise quality filter was trained to assess the adequacy of video segments for diagnostic purposes.

Key Findings:

• The deep residual-recurrent neural network and the decision tree network exhibited almost identical diagnostic accuracy.

• The neural network algorithm classified tympanic membrane videos into AOM vs. no AOM categories with a sensitivity of 93.8% and specificity of 93.5%, while the decision tree model had a sensitivity of 93.7% and specificity of 93.3%.

• Bulging of the tympanic membrane was the feature most closely aligned with the predicted diagnosis of AOM, present in all cases where AOM was predicted in the test set.

The study demonstrates the high accuracy of the AI algorithm in diagnosing AOM based on tympanic membrane features observed in otoscopic videos. This suggests that the AI tool, along with a medical-grade application for image acquisition and quality filtering, could be utilized in primary care or acute care settings to aid in automated diagnosis of AOM and treatment decisions. This advancement has the potential to improve clinical outcomes and optimize healthcare resource utilization in managing AOM in children.

Reference:

Shaikh, N., Conway, S. J., Kovačević, J., Condessa, F., Shope, T. R., Haralam, M. A., Campese, C., Lee, M. C., Larsson, T., Cavdar, Z., & Hoberman, A. Development and validation of an automated classifier to diagnose acute otitis media in children. JAMA Pediatrics,2024. https://doi.org/10.1001/jamapediatrics.2024.0011