Eardrum Exams Take a Digital Turn

Machine learning analyzes smartphone otoscope images to differentiate between otitis media with effusion and normal tympanic membrane images.

By
Kathryn Wighton
April 10, 2026
3 min

Conexiant

Overview

A supervised machine-learning model utilizing smartphone images demonstrated high accuracy in detecting middle ear effusion in pediatric patients. The model achieved 96% sensitivity and 89% accuracy during training, although performance declined in testing.

Background

Otitis media with effusion (OME) is a common condition in children that often goes misdiagnosed, leading to potential delays in treatment. Accurate diagnosis is crucial as OME can impact hearing and language development. This study explores the use of machine learning and smartphone technology to improve diagnostic accuracy in a clinical setting.

Data Highlights

Metric	Training Set	Testing Set
Sensitivity	96%	87%
Specificity	81%	74%
Accuracy	89%	81%
Balanced Accuracy	-	80.4%
F1 Score	-	82.5%

Key Findings

The machine-learning model was trained on 54 tympanic membrane images.
It achieved 96% sensitivity and 89% accuracy in the training phase.
In the testing phase, sensitivity dropped to 87% and accuracy to 81%.
Ground-truth diagnoses were established by consensus between two otolaryngologists.
Limitations included a small sample size and lack of external validation.

Clinical Implications

The findings suggest that smartphone-based imaging combined with machine learning could enhance the diagnostic process for OME in pediatric patients. Clinicians may consider integrating such technology to improve diagnostic accuracy and reduce misdiagnosis rates.

Conclusion

The study indicates that machine learning can effectively assist in diagnosing middle ear effusion using smartphone technology, although further validation is necessary to confirm its clinical utility.