Kassjański Michał, Kulawiak Marcin, Przewoźny Tomasz, Tretiakow Dmitry, Molisz Andrzej
Department of Geoinformatics, Faculty of Electronics, Telecommunications and Informatics, Gdansk University of Technology, Gdańsk, Poland.
Department of Otolaryngology, Medical University of Gdańsk, Gdańsk, Poland.
Sci Rep. 2025 Apr 24;15(1):14302. doi: 10.1038/s41598-025-99338-5.
Hearing impairments are typically assessed using pure tone audiometry, a diagnostic method that allows for the identification of the degree, type and configuration of hearing loss. The results of this assessment are generally displayed in the form of an audiogram, which graphically represents the softest sounds perceivable by an individual across a range frequencies. This paper presents a novel Open Source mobile application for the Android operating system that allows users to scan and analyse audiograms using a smartphone camera and subsequently classify the type of hearing loss. The application workflow is divided into three main stages: scanning, digitalization and classification of the audiogram. For this purpose, the application implements several artificial intelligence and image processing techniques, including YOLOv5, Optical Character Recognition (OCR) and Hough Transform. The scanned audiogram is analysed by a clinically validated AI model for classification of audiometric test results, providing clinicians with valuable assistance in formulating a diagnosis. All implemented algorithms and models were optimized for functionality on mobile devices. The application was evaluated on three distinct classes of smartphones across various price points, demonstrating its efficacy and consistent performance. The presented mobile application constitutes an advanced AI-driven decision support system that is readily accessible to general practitioners, otolaryngologists and audiologists. Its integration in medical facilities presents a substantial opportunity to decrease clinical workload, enhance diagnostic accuracy and reduce the likelihood of human error in hearing loss evaluations, which is particularly important in developing countries.
听力障碍通常使用纯音听力测试进行评估，这是一种诊断方法，可用于确定听力损失的程度、类型和形态。该评估结果通常以听力图的形式呈现，听力图以图形方式表示个体在一系列频率中可感知的最轻柔声音。本文介绍了一款适用于安卓操作系统的新型开源移动应用程序，该程序允许用户使用智能手机摄像头扫描和分析听力图，并随后对听力损失类型进行分类。应用程序工作流程分为三个主要阶段：听力图扫描、数字化和分类。为此，该应用程序实施了多种人工智能和图像处理技术，包括YOLOv5、光学字符识别（OCR）和霍夫变换。通过经过临床验证的人工智能模型分析扫描的听力图，以对听力测试结果进行分类，为临床医生在制定诊断时提供有价值的帮助。所有实施的算法和模型都针对移动设备的功能进行了优化。该应用程序在不同价格点的三类不同智能手机上进行了评估，证明了其有效性和一致的性能。所展示的移动应用程序构成了一个先进的人工智能驱动的决策支持系统，普通医生、耳鼻喉科医生和听力学家都可以方便地使用。将其整合到医疗机构中为减少临床工作量、提高诊断准确性以及降低听力损失评估中人为错误的可能性提供了重大机遇，这在发展中国家尤为重要。
