Using Artificial Intelligence to Diagnose 'Glue Ear' in Children


Fei Zhao, Professor in Hearing Science Audiology, Centre for Speech and Language Therapy and Hearing Science, Cardiff School of Sport and Health Sciences, Cardiff Metropolitan University. He has been awarded an Artificial Intelligence in Health and Care Award, through the National Institute for Health Research (NIHR), Sêr Cymru III Enhancing Competitiveness Infrastructure Award (MA/KW/5554/19), Great Britain Sasakawa Foundation (5826), Cardiff Metropolitan University Research Innovation Award and The Global Academies Research and Innovation Development Fund, Cardiff Metropolitan University Research Innovation Impact Fund. Professor Zhao also won Global Academies and Santandar 2021 Fellowship Award. This funding enables him to collaborate with a world-leading scholar in this field, Professor De Wet Swanpoel from the University of Pretoria in South Africa. Further development and implementation of effective and non-specialist Artificial Intelligence (AI) tools will improve accessibility and quality of global hearing healthcare for children living with hearing impairment, particularly in Low- and Mid- Incomes Countries (LMICs).

Otitis Media with Effusion (OME), commonly known as “Glue Ear” is one of the most common causes of childhood hearing impairment and disability. It is estimated that more than 80% of children will have an episode of otitis media before the age of 10.  This high prevalence places a significant cost burden on the NHS with approximately 200,000 children with OME seen annually in primary care. Because of lacking clear infectious symptoms (e.g., earache or fever) and difficulties in assessing the hearing status of young children (particularly as there are no audiology specific diagnostic facilities in the primary care setting), glue ear becomes one of the biggest challenges for non-ENT/Audiology professionals. Poor diagnostic accuracy leads to delayed diagnosis and inappropriate intervention decisions, such as re-assurance without appropriate treatment or referral to ENT unnecessarily by GPs. Furthermore and concerning is that delayed diagnosis and poor management can result in severe and persistent OME with surgical treatment becoming the only management option, leading to long waiting times and excessively high costs for the NHS.

Professor Zhao is currently leading a research team to carry out interdisciplinary work for development and implementation of AI tools in hearing healthcare services as a clinical decision support system. Their recent research into AI has made great progress in demonstrating its potential for the accurate diagnosis of glue ear (Grais et al., 2021). This novel research measures energy absorbance of the middle ear at varying frequencies and pressures in normal and glue ears. With this information the AI tools will provide an accurate automated diagnosis which can support the clinician in their decision making process. The recent results proves the capability of AI tools to diagnose glue ear automatically with an accuracy of 95%. This work will have significant and direct impact on clinical assessment and diagnostic concepts for glue ear in professional communities, and thus NICE guidelines for glue ear treatment. Furthermore, the important research outputs of this research will provide long-term benefits to the wellbeing of children with glue ear and their families. In the meantime, low cost AI clinical decision support system for automated diagnosis of glue ear will provide solutions to the challenges of shortage of specialist training, unaffordable equipment, and thereby establish a cost-effective and sustainable hearing healthcare service model, which could be used globally, particularly in LMICs. 

Relevant publications

Grais, E.M., Wang, X.Y., Wang, J., Zhao, F., Jiang, W., Cai, Y.X., Zhang, L.F., Lin, Q.W, Yang, H.D., 2021, Analysing wideband absorbance immittance in normal and ears with otitis media with effusion using machine learning. Scientific Reports, 11:10643.

Cao ZW,Chen FF, Grais EM, Yue FJ, Cai YX, Swanepoe DW, Zhao F. Machine Learning in Diagnosing Middle Ear Disorders Using Tympanic Membrane Images: A Meta-Analysis. Laryngoscope (in press) doi: 10.1002/LARY.30291

Zeng JB, Kang WB, Chen SJ, Lin Y, Deng WT, Wang YJ, Chen GS, Ma K, Zhao F, Zheng YF, Liang, MJ, Zeng LQ, Ye WJ, Li P, Chen YB, Chen GP, Gao JL, Wu MJ, Su YJ, Zheng YQ, Cai YX. A Deep Learning Approach to Predict Conductive Hearing Loss in Otitis Media with Effusion Using Otoscopic Images. JAMA Otolaryngology (in press). doi:10.1001/jamaoto.2022.0900