Deep Learning Techniques for Ear Diseases Based on Segmentation of the Normal Tympanic Membrane

Abstract

Background: Otitis media is a common infection affecting people worldwide. Owing to the limited number of ear specialists and rapid development in telemedicine, several trials have been conducted for developing novel diagnostic strategies to improve the diagnostic accuracy and screening of patients afflicted with otologic diseases based on abnormal otoscopic findings. Although these strategies have demonstrated a high diagnostic accuracy for the tympanic membrane (TM), the insufficient explainability of such techniques limits their deployment in clinical practices. Methods: Herein, we used a deep convolutional neural network (CNN) model based on the segmentation of a normal TM into five substructures (malleus, umbo, cone of light, pars flaccida, and annulus) to identify abnormalities in otoscopic ear images. The Mask R-CNN algorithm learned the labeled images. Subsequently, the values were combined according to the five substructures using a three-layer fully connected neural network to determine whether an ear disease was present. Results: We obtained the receiver operating characteristic (ROC) curves of the optimal condition for the presence or absence of eardrum diseases according to each substructure or a combination of substructures. The results indicated that the highest area under the curve ranged from a 0.911 true-positive rate in the ROC curve combined with malleus, cone of light, and umbo, compared with the corresponding range of 0.737–0.873 in each substructure. Thus, an algorithm using these five important normal anatomical structures could prove to be explainable and effective in screening abnormal TMs.Conclusion: This automated algorithm can improve the diagnostic accuracy to discriminate between normal and abnormal TMs and facilitate appropriate and timely referral consultations to improve the quality of life in public healthcare.

Feasibility of Multimodal Artificial Intelligence Using GPT-4 Vision for the Classification of Middle Ear Disease: Qualitative Study and Validation

“…The emergence of artificial intelligence (AI) has altered the landscape of medical technology, particularly in diagnosis, which leverages the identification of features based on imaging and physiological data [1][2][3]. In the field of otolaryngology, AI and deep learning models are being used for imaging; ongoing efforts focus on classifying diseases based on tympanic membrane images of middle ear disease [4][5][6]. Technological advancements, including deep learning and transfer learning using pre-trained models, have resulted in an accuracy range of 70-90% in models for analyzing otoscopic images [7].…”

Artificial Intelligence With Deep Learning Based Automated Ear Infection Detection

“…Due to bacterial infection, AOM may occur in the middle of the ear, bringing about the build-up of fluid. OME leads to fluid build-up in the middle of the ear because of inflammation, which is much more severe than AOM [3]. The standard diagnosis depends on pneumatic otoscopy, the benchmark for distinguishing AOM from OME [4].…”

Artificial Intelligence With Deep Learning Based Automated Ear Infection Detection

“…AOM, or middle ear mass, can result from bacterial infection and cause fluid accumulation. OME causes a buildup of fluid in the middle of the ear due to inflammation, which is significantly worse than AOM [3]. Pneumatic otoscopy, the gold standard for differentiating between AOM and OME, is required for the conventional diagnosis [4].…”