Otitis media is one of the most common childhood diseases in developing countries; many of its complications are preventable if middle ear fluid is detected early. We propose an accessible and accurate smartphone-based screening tool that (i) sends a soft acoustic chirp into the ear canal using the smartphone speaker, (ii) detects reflected sound from the eardrum using the smartphone microphone, and (iii) employs a machine learning model to classify these reflections and predict middle ear fluid status in realtime. Given the ubiquity of smartphones and the inaccuracy of visual otoscopy, the system we propose has the potential to be the default screening tool used in developing countries by healthcare providers and caregivers at home.
Public Health Relevance Statement: PROJECT NARRATIVE Otitis media is one of the most common childhood diseases in developing countries affecting over 1.23 billion people in 2013 and can lead to complications such as hearing loss, developmental delay, meningitis, mastoiditis, and death. Many of these complications are preventable if middle ear fluid is detected early. However, the absence of an accurate and accessible method to detect middle ear fluid has led to high misdiagnosis rates. The consequence is associated hearing and speech impairment rates greater than any other pediatric condition and growing microbial resistance as a result of antibiotic over-prescription. Currently, the technique of choice for detecting middle ear fluid by primary care providers is visual otoscopy, which has a diagnostic accuracy as low as 51%. Although more accurate methods like tympanometry and pneumatic otoscopy exist, they require significant expertise and referral to a specialist. Commercial acoustic reflectometers and smartphone-mounted otoscopes require specialized hardware. Thus, there is an urgent, unmet need for an accurate, rapid and easily accessible method for resource-limited healthcare providers and caregivers to detect middle ear fluid. This project aims to demonstrate the feasibility of using the speakers and microphones on existing smartphones to detect middle ear fluid by assessing eardrum mobility. Our proposed system would operate by (i) sending a soft acoustic chirp into the ear canal using the smartphone speaker, (ii) detecting reflected sound from the eardrum using the smartphone microphone, and (iii) employing a machine learning model to classify these reflections and predict middle ear fluid status. No additional attachments would be required beyond a paper funnel, which acts as a speculum to reduce waveform variability and can be constructed with printer paper, scissors, and tape. This technique is the first software-based screening tool for middle ear fluid detection that uses off-the-shelf smartphones which does not require hardware attachments or visual interpretation. Using data from our existing preliminary clinical study we aim to develop signal processing and machine learning algorithms to optimize sensitivity and specificity. We plan to develop a bench testing technique that enables previously unsupported smartphones to to run our test and prospectively validate our optimized algorithm clinically in parallel testing with an acoustic reflectometer. Further we aim to develop a user interface and improved funnel design. These new designs will undergo usability testing in physician and parent populations. Given the ubiquity of smartphones, our app has the potential to be the default screening tool used in developing countries by healthcare providers and caregivers at home.
Project Terms: Acoustics; Affect; Agreement; Algorithms; Antibiotics; base; care providers; Caregivers; Cellular Phone; Cessation of life; Childhood; Clinic; Clinical; clinical care; Clinical Data; clinical practice; Clinical Research; Computer software; Data; design; Detection; Developing Countries; Development; Developmental Delay Disorders; Diagnosis; diagnostic accuracy; Disease; Ear; Earwax; Environment; experience; External auditory canal; FDA approved; Feedback; Future; Galaxy; Health; Health Personnel; Hearing; hearing impairment; Home environment; Impairment; improved; Industry Standard; Lead; Liquid substance; Machine Learning; machine learning algorithm; Mastoiditis; Measures; meetings; Meningitis; Methods; microbial; microphone; middle ear; Modeling; Obstruction; Otitis Media; Otoscopes; Otoscopy; Outcome; Output; Paper; Parents; Patients; Performance; Peripheral; Phase; Physicians; Population; Preparation; prospective; Publishing; Research Personnel; Resistance; Resources; Running; screening; Screening procedure; Sensitivity and Specificity; signal processing; Small Business Innovation Research Grant; sound; Specialist; Specificity; Speculums; Speech; System; Techniques; Technology; telehealth; Testing; tool; Training; Tympanic membrane; Tympanometry; urgent care; usability; Visual
