Machine Learning for Accurate Intraoperative Pediatric Middle Ear Effusion Diagnosis

OBJECTIVES: Misdiagnosis of acute and chronic otitis media in children can result significant consequences from either undertreatment or overtreatment. Our objective was to develop train an artificial intelligence algorithm accurately predict the presence middle ear effusion pediatric patients presenting operating room for myringotomy tube placement. METHODS: We trained a neural network classify images as “ normal” (no effusion) “abnormal” (effusion present) using tympanic membrane taken with intent performing possible placement recurrent effusion. Model performance tested on held-out cases fivefold cross-validation. RESULTS: The mean training time model 76.0 (SD ± 0.01) seconds. approach achieved image classification accuracy 83.8% (95% confidence interval [CI]: 82.7–84.8). In support this accuracy, produced area under receiver characteristic curve 0.93 CI: 0.91–0.94) F1-score 0.80 0.77–0.82). CONCLUSIONS: Artificial intelligence–assisted diagnosis may generate value patients, families, health care system by improving point-of-care diagnostic accuracy. With small data set composed intraoperative obtained at tympanostomy insertion, our accurate predicting cases. This is considerably higher than human-expert otoscopy-based reported previous studies.