ABSTRACT Acute infections of the middle ear (acute otitis media - AOM), are the most commonly treated childhood disease. Treatment is fueled by concern for complications and effects on children's cognitive and language development. The financial burden of AOM is estimated at more than $5 billion per year. Because AOM is so common, a major societal problem is the over-diagnosis and over-treatment of this disease, as a result of two factors: First, accurately diagnosing AOM is difficult, even for experienced primary care or ear, nose, and throat (ENT) physicians. Second, with a growing shortage of primary care physicians in the US, more Nurse Practitioners and Physician Assistants serve as first-line clinicians in primary care settings, but lack extensive training in otoscopy (i.e. clinical examination of the eardrum). Consequently, practitioners often err on the side of making a diagnosis of AOM and prescribing oral antibiotics. Over 8 million unnecessary antibiotics are prescribed annually, contributing to the rise of antibiotic-resistant bacteria, and creating the largest number of pediatric medication-related adverse events. Many children with inaccurate diagnoses of AOM are referred to ENTs for surgical placement of ear tubes, and up to 70% of these cases are not indicated. Diagnosing AOM still depends on clinician subjectivity, based on a brief glimpse of the eardrum. This diagnostic subjectivity creates a critical barrier to progress in society's goal of decreasing healthcare costs and reducing over-diagnosis and over-treatment of AOM. According to the American Academy of Pediatrics in 2013, devices are needed to assist in more accurate, consistent, and objective diagnosis of AOM. A simple and objective method of analyzing an image of a patient's ear to diagnose or rule out AOM would drastically reduce over-treatment. This project will fill that gap, by developing computer-assisted image analysis (CAIA) software that provides objective information to a clinician by analyzing eardrum images collected using currently available hardware. Based on previous work in applying similar methods to improve clinician performance in radiology and surgical pathology, our overarching hypothesis is that the incremental implementation of enhanced images, automated identification of abnormalities, and retrieval of similar cases will result in improved clinician diagnostic accuracy. In our preliminary work, we developed software, called Auto-Scope, which labels eardrums as ?normal? versus ?abnormal.? In this study, we propose two Specific Aims to improve diagnostic performance: Specific Aim #1: Create an enhanced composite image of the eardrum. Specific Aim #2: Use machine learning approaches for clinical decision support.

R21DC016972

2018-09-21

2021-08-31