A Deep Learning Approach for Detecting Otitis Media From Wideband Tympanometry Measurements

Abstract

Objective: In this study, we propose an automatic diagnostic algorithm for detecting otitis media based on wideband tympanometry measurements. Methods: We develop a convolutional neural network for classification of otitis media based on the analysis of the wideband tympanogram. Saliency maps are computed to gain insight into the decision process of the convolutional neural network. Finally, we attempt to distinguish between otitis media with effusion and acute otitis media, a clinical subclassification important for the choice of treatment. Results: The approach shows high performance on the overall otitis media detection with an accuracy of 92.6%. However, the approach is not able to distinguish between specific types of otitis media. Conclusion: Out approach can detect otitis media with high accuracy and the wideband tympanogram holds more diagnostic information than the commonly used techniques wideband absorbance measurements and simple tympanograms. Significance: This study shows how advanced deep learning methods enable automatic diagnosis of otitis media based on wideband tympanometry measurements, which could become a valuable diagnostic tool.

Three‐dimensional wideband absorbance immittance findings in young adults with large vestibular aqueduct syndrome

“… 38 Keefe 39 also observed variance in reflectance above 4000 Hz using a causal constraint procedure to measure acoustic reflectance. Sundgaard et al 40 thought that WBT results above 4000 Hz were susceptible to noise, which may explain this unreliability. These results indicate a degree of uncertainty with the high‐frequency acoustic reflectance results.…”

Multi‐modal deep learning for joint prediction of otitis media and diagnostic difficulty

“…Automatic diagnosis of otitis media has been tackled in various ways. Previous studies have employed datasets of otoscopy images, 1–4 tympanometry measurements 5–7 optical coherence tomography, 8 or computed tomography 9 . The approaches have utilized a variety of machine learning algorithms for the data analysis and classification task, progressing from simpler methods such as Random Forest 10 and Support Vector Machines, 11 to advanced deep neural networks 1,5,7,12,13 .…”

“…Previous studies have employed datasets of otoscopy images, 1–4 tympanometry measurements 5–7 optical coherence tomography, 8 or computed tomography 9 . The approaches have utilized a variety of machine learning algorithms for the data analysis and classification task, progressing from simpler methods such as Random Forest 10 and Support Vector Machines, 11 to advanced deep neural networks 1,5,7,12,13 . When a doctor examines a patient, the diagnostic decision is rarely based solely on one modality from the clinical examination.…”

“…Wideband tympanometry (WBT) has shown to be more efficient in evaluating the condition of the middle ear, and it provides more detailed information on the mechanical and acoustic status of the middle ear than the standard 226‐Hz tympanogram 15 . Furthermore, higher classification accuracy can be achieved using WBT measurements for the detection of otitis media, compared to both ambient absorbance and standard tympanograms 7 . We propose the use of WBT measurements in combination with otoscopy images for the diagnosis of otitis media in the diagnostic groups: otitis media with effusion (OME), acute otitis media (AOM), and no effusion (NOE).…”

“…An important aspect of this step is allowing the user of a clinical tool to learn more from the model than just the diagnosis. Several studies have employed saliency maps to allow the user to learn about the decision process of the model by identifying the most important features of the input data 7,17 . Another valuable output would be an estimate of the difficulty of the input case.…”