Ear-Keeper - A cross-platform artiﬁcial intelligence system for rapid and accurate ear disease diagnosis

Sep 29, 2026

Objective: Early and accurate detection of ear diseases is essential for preventing hearing impairment and improving population health. Methods: A lightweight, high-performance real-time deep learning model was built for otoscopic image classification and deployed in a cross-platform diagnostic system. Results: Best-EarNet achieved 95.23% accuracy on the internal test set and 92.14% on the external set, processing at 80 FPS on a standard CPU. Grad-CAM improved interpretability. Conclusion: Ear-Keeper supports real-time screening across smartphones, tablets, and PCs for clinical and community use.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages20

Reading Time50 min

AccessFree to download

Abstract

Objective

Methods

Results

Conclusion

Introduction

Ear-Keeper: A cross-platform artiﬁcial intelligence system for rapid and accurate ear disease diagnosis DIGITAL HEALTH Volume 12: 1–20 © The Author(s) 2026 Article reuse guidelines: [sagepub.com/journals-permissions](sagepub.com/journals-permissions)[ ](sagepub.com/journals-permissions)[DOI: 10.1177/20552076251412635](DOI: 10.1177/20552076251412635)[ ](DOI: 10.1177/20552076251412635)[journals.sagepub.com/home/dhj](journals.sagepub.com/home/dhj) Feiyan Lu 1,*, Yubiao Yue2,* , Zhenzhang Li3 , Meiping Zhang4, Wen Luo5, Fan Zhang6, Tong Liu 1, Jingyong Shi7, Guang Wang8 and Xinyu Zeng 1 Abstract Objective: Early and accurate detection of ear diseases is essential for preventing hearing impairment and improving population health. This study aimed to develop a lightweight, high-performance, and real-time deep learning model for otoscopic image classiﬁcation and to deploy it in a cross-platform diagnostic system for clinical and community use. Methods: We constructed a large-scale, multi-center otoscopy dataset covering eight common ear diseases and healthy cases. Based on this dataset, we developed Best-EarNet, an ultrafast lightweight architecture integrating a local-global spatial feature fusion module and a multi-scale supervision strategy to enhance feature representation. Transfer learning was applied to optimize performance. The model was evaluated on internal (22,581 images) and external (1,652 images) test sets, with subgroup analyses by age and gender. Grad-CAM visualizations were used to improve interpretability. Across-platform intelligent diagnostic system, Ear-Keeper, was further developed for deployment on smartphones, tablets, and personal computers. Results: Best-EarNet achieved accuracies of 95.23% on the internal test set and 92. 14% on the external test set, with a model size of 2.94 MB. It processed images at 80 frames per second on a standard CPU. Subgroup analyses demonstrated consistently high performance across age and gender groups. Grad-CAM visualizations highlighted lesion-related regions, and Ear-Keeper enabled real-time video-based ear screening across multiple platforms. Conclusion: Best-EarNet offers an accurate, efﬁcient, and interpretable solution for ear disease classiﬁcation. Its realtime performance and cross-platform deployment through Ear-Keeper support both clinical practice and communitylevel screening, with strong potential for early detection and intervention. Keywords Ear diseases, deep learning, healthcare, intelligent system, otoscopy Received: 22 August 2025; accepted: 16 December 2025 1 Department of Otorhinolaryngology, The Second Afﬁliated Hospital of Shenzhen University, Shenzhen, China 2Department of Biomedical Engineering, School of Life Science and Technology, Xi’an Jiaotong University, Xian, China 3School of Mathematics and Systems Science, Guangdong Polytechnic Normal University, Guangzhou, China 4College of Computer and Cyber Security, Fujian Normal University, Fuzhou, China 5Department of Scientiﬁc Research, Kangmeihuada (KMHD) GeneTech Co., Ltd, Shenzhen, China 6Department of Science and Education, Foshan Sanshui District People’s Hospital, Foshan, China 7Department of Otorhinolaryngology, Zhongshan Gangkou Hospital, Zhongshan, China 8Department of Otorhinolaryngology, Yichang Central People’s Hospital, Yichang, China *These authors contributed equally to this work. Corresponding author: Xinyu Zeng, Department of Otorhinolaryngology, The Second Afﬁliated Hospital of Shenzhen University, No. 118, Longjing 2nd Road, Xin’an Street, Bao’an District, Shenzhen 518101, Guangdong, China. Email: [2432968@qq.com](2432968@qq.com) Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution NonCommercial 4.0 License ([https://creativecommons.org/licenses/by-nc/4.0/](https://creativecommons.org/licenses/by-nc/4.0/)) which permits non-commercial use, reproduction and distribution o

FAQs

What problem does Ear-Keeper address?

Ear-Keeper targets early and accurate detection of ear diseases to prevent hearing impairment and support better population health.

How was the deep learning model developed and optimized?

The study built a large multi-center otoscopy dataset covering eight ear diseases plus healthy cases, then developed Best-EarNet with local-global feature fusion, multi-scale supervision, and transfer learning.

What performance and deployment capability does Ear-Keeper provide?

Best-EarNet reached 95.23% accuracy internally and 92.14% externally with a 2.94 MB model, running at 80 FPS on a standard CPU and enabling real-time video-based screening across multiple platforms.

Sep 29, 2026

Uploaded by Mafia Boss