Forthcoming tool uses AI to help diagnose ear infections
Click Here to Manage Email Alerts
Worldwide, most children encounter at least one episode of ear infection before reaching 7 years of age. In the United States, health care costs related to the condition may be as high as $5 billion annually, data in Molecules show.
The authors of the Molecules article also wrote that two common types of ear infections — acute otitis media and otitis media with effusion — are caused by a bacterial infection in the middle ear that leads to a buildup of fluid. Patients with either infection are also subject to eardrum swelling, making it difficult to tell these infections apart.
OtoPhoto — a device that uses images of the inner ear and machine learning to determine whether an infection exists — may make it easier for physicians to diagnose ear infections, according to a Johns Hopkins University press release.
Healio Primary Care spoke with James Clark, MB, BCh, BAO, assistant professor of otolaryngology at the Johns Hopkins University School of Medicine and co-inventor of OtoPhoto, to learn more about how the device works, its potential for use via telehealth and more.
Q: How does OtoPhoto work?
A: Our Otoscope integrates with an AI algorithm that has been taught to analyze perceptual data from a video recording of the patient’s eardrum. From this real-time analysis, the algorithm predicts the presence or absence of ear infections with the sensitivity and specificity of a physician with expertise in the ear.
Q: How could this technology improve the diagnosis of ear infections in the primary care setting?
A: The diagnosis of an ear infection is made clinically. The variability in presentation and the poor correlation of ear symptoms with actual ear infections present a significant diagnostic challenge and contribute to misuse and overuse of antibiotics. Correct diagnosis is therefore reliant on visualization of characteristic signs of infection such as inflammation, color change and/or bulging of the eardrum. The subtlety of signs and complexity of the anatomy can make such pattern recognition problematic.
Studies have shown that pediatricians and general practitioners have only been reported to have a diagnostic accuracy little better than a coin toss. OtoPhoto will enable users, irrespective of level of training or experience, the ability to identify and treat ear infections in real time from an image with the level of precision of an ear specialist or otologist.
Q: What resources would a primary care physician need to implement OtoPhoto in their practice? How much will the tool and other potential resources cost?
A: OtoPhoto is currently under development. However, we seek to provide a quality noninvasive tool with an associated cost that enables widespread usage and availability of our product.
Q: How can OtoPhoto be used in a telehealth setting?
A: We envision OtoPhoto to be a tool that patients will be able to use at home to help determine if there is an ear infection or not. In the case of an ear infection, the patient will be able to send captured images to their health care provider who can then decide to prescribe an antibiotic. In the current setting, it is hard to correctly diagnose ear infection via telehealth as it is hard for clinicians to get a view of the eardrum and, as mentioned above, there is poor correlation between ear symptoms and ear infections.
Q: How are you testing the protype? When do you predict OtoPhoto will be available for real-world use?
A: We will be conducting rigorous scientific validation of our product comparing diagnoses made by our tool to real-time expert opinions and testing intra- and interobserver variability, sensitivity and specificity to develop OtoPhoto as a screening tool. We hope to have a device ready for the market within the next 24 months.