AI Tool Shows Promise in Accurately Diagnosing Pediatric Ear Infection

March 7, 2024
Ear infections are incredibly common in children yet difficult to diagnose, but this tool could provide doctors with more clarity

A new study has shown that an artificial intelligence (AI)-based tool may boost accuracy and reduce unnecessary antibiotic prescribing in ear infections in children. Historically, acute otitis media (ear infection) has been hard to diagnose accurately even though it is one of the most common infections in children.

A research team has developed a “medical grade smartphone application that uses a smartphone camera to capture video otoscope of the tympanic membrane through an endoscope or otoscope.” The scientists then collected a library of these videos in children younger than 36 months old over the course of two years in pediatric clinics in Pittsburgh. Then, “two validated otoscopists reviewed the video and assigned a final diagnosis.” Using the videos and this diagnosis information, they created an AI algorithm to evaluate tympanic membrane (TM) features and provide a diagnosis.

Initial results from this study were overwhelmingly positive. The AI tool was found to have a “sensitivity of 93.8%” and a “specificity of 93.3%.” Plus, 80% of parents asked about the AI tool “wanted the doctor to use [it] during future visits.”

The authors of the study concluded that “AI accuracy was better than that of pediatricians, primary care physicians, and advance-practice clinicians.” They are optimistic that use of this tool can “reduce inappropriate use of antimicrobials” for ear infections.

Some possible drawbacks to the study noted in a related commentary include “how training and testing data were selected and that expert otoscopists were the gold standard, rather than myringotomy and tympanocentesis,” two procedures the researchers decided not to use because of their invasive nature. Regardless, the commentators wrote that the “high accuracy of the algorithm…as well as its implementation as a mobile application that could be used in real time can lead to the hope that diagnosis of otitis media could be transformed using such technology.”

Fairness and bias are also important things that need to be studied with regards to this algorithm, as AI models trained on lighter skin tones “might make it less accurate for patients who have darker skin tones.”

CIDRAP’s website has the news release.