Journals / 21. Yüzyılda Fen ve Teknik / 2025 / Cilt: 12 Sayı: 24

Detection Of Tinnitus From Audiogram Images Using Machine Learning Methods

Makine Öğrenimi Yöntemlerini Kullanarak Odyogram Görüntülerinden Kulak Çınlamasının Tespiti

Remzi Gürfidan(ISPARTA UYGULAMALI BİLİMLER ÜNİVERSİTESİ)0000-0002-4899-2219
Yasin Tatlı(AVRASYA ÜNİVERSİTESİ)0009-0003-4936-0007
Pages
76–86
Publication Date
2025-12-31
DOI
—
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

Considering the models above, the best overall results were obtained using the Xception model with 0.9387 validation accuracy and 91.67% test accuracy, together with rather balanced precision, recall, and F1-score values of 0.928.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Tinnitus is an important symptom of hearing problems, the diagnosis of which is generally carried out through subjective clinical interviews and audiological tests; this may lead to variability in its diagnosis and follow-up.
📊
Key Findings & Data:A conventional convolutional neural network and four transfer-learning–based deep architectures, namely InceptionV3, DenseNet121, VGG19, and Xception, were trained and tested.
💡
Conclusion & Impact:Considering the models above, the best overall results were obtained using the Xception model with 0.9387 validation accuracy and 91.67% test accuracy, together with rather balanced precision, recall, and F1-score values of 0.928.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Kulak çınlaması · odyogram · derin öğrenme

Abstract

Tinnitus is an important symptom of hearing problems, the diagnosis of which is generally carried out through subjective clinical interviews and audiological tests; this may lead to variability in its diagnosis and follow-up. The goal of this study is to develop an objective decision-support approach for tinnitus detection by classifying audiogram images with state-of-the-art machine learning methods. This study used a publicly accessible audiogram dataset with standard clinical hearing threshold information and tinnitus labels. These records were converted into image representations and split into training, validation, and test sets. A conventional convolutional neural network and four transfer-learning–based deep architectures, namely InceptionV3, DenseNet121, VGG19, and Xception, were trained and tested. Model performance was evaluated based on metrics such as validation accuracy, validation loss, test accuracy, precision, recall, and F1-score. Considering the models above, the best overall results were obtained using the Xception model with 0.9387 validation accuracy and 91.67% test accuracy, together with rather balanced precision, recall, and F1-score values of 0.928. It was followed by models InceptionV3 and DenseNet121, which attained 0.9202 validation accuracy and over 90% test accuracies, while the CNN baseline resulted in 74.02% test accuracy. VGG19 showed markedly lower performance, indicating that not all deep architectures are equally suitable for this task. The findings demonstrate that modern deep learning architectures, particularly Xception, can successfully extract discriminative patterns from audiogram images and provide reliable, objective support for tinnitus detection. These results suggest that integrating such models into audiology workflows could contribute to faster, more standardised, and reproducible clinical decision-making.

Özet

Kulak çınlaması, işitme sorunlarının önemli bir belirtisidir ve tanısı genellikle öznel klinik görüşmeler ve odyolojik testler ile konur; bu da tanısı ve takibinde farklılıklara yol açabilir. Bu çalışmanın amacı, en son teknoloji makine öğrenimi yöntemleri ile odyogram görüntülerini sınıflandırarak kulak çınlamasının tespiti için objektif bir karar destek yaklaşımı geliştirmektir. Bu çalışmada, standart klinik işitme eşiği bilgileri ve kulak çınlaması etiketleri içeren, kamuya açık bir odyogram veri seti kullanılmıştır. Bu kayıtlar görüntü temsillerine dönüştürülmüş ve eğitim, doğrulama ve test setlerine ayrılmıştır. Geleneksel bir evrişimli sinir ağı ve dört transfer öğrenme tabanlı derin mimari, yani InceptionV3, DenseNet121, VGG19 ve Xception eğitilmiş ve test edilmiştir. Model performansı, doğrulama doğruluğu, doğrulama kaybı, test doğruluğu, hassasiyet, geri çağırma ve F1 puanı gibi metriklere göre değerlendirilmiştir. Yukarıdaki modeller göz önüne alındığında, en iyi genel sonuçlar, 0,9387 doğrulama doğruluğu ve %91,67 test doğruluğu ile birlikte 0,928'lik oldukça dengeli kesinlik, geri çağırma ve F1 puanı değerlerine sahip Xception modeli kullanılarak elde edildi. Bunu, 0,9202 doğrulama doğruluğu ve %90'ın üzerinde test doğruluğu elde eden InceptionV3 ve DenseNet121 modelleri izlerken, CNN temel modeli %74,02 test doğruluğu ile sonuçlandı. VGG19 belirgin şekilde daha düşük performans gösterdi, bu da tüm derin mimarilerin bu görev için eşit derecede uygun olmadığını gösterdi. Bulgular, modern derin öğrenme mimarilerinin, özellikle Xception'ın, odyogram görüntülerinden ayırt edici kalıpları başarıyla çıkarabildiğini ve kulak çınlaması tespiti için güvenilir, objektif destek sağlayabildiğini göstermektedir. Bu sonuçlar, bu tür modellerin odyoloji iş akışlarına entegre edilmesinin daha hızlı, daha standart ve tekrarlanabilir klinik karar verme sürecine katkıda bulunabileceğini göstermektedir.

🔖

Cite This Article

Gürfidan, R., Tatlı, Y. (2025). Detection Of Tinnitus From Audiogram Images Using Machine Learning Methods. 21. Yüzyılda Fen ve Teknik, Cilt: 12 Sayı: 24, 76–86. https://ajindex.com/articles/1628557

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (19)

  1. Allgaier, J., Neff, P., Schlee, W., Schoisswohl, S., & Pryss, R. (2021). Deep Learning End-to-End Approach for the Prediction of Tinnitus based on EEG Data. In Proceedings of the 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC)
  2. Awais, M., Kassoul, K., El Omri, A. E., Aboumarzouk, O. M., Abdulhadi, K., & Brahim Belhaouari, S. (2024). Graph-Based Electroencephalography Analysis in Tinnitus Therapy. Biomedicines, 12(7), 1404
  3. Baguley, D., McFerran, D., & Hall, D. (2013). Tinnitus. The Lancet, 382(9904), 1600-1607.
  4. Couth, S., Mazlan, N., Moore, D. R., Munro, K. J., & Dawes, P. (2019). Hearing difficulties and tinnitus in construction, agricultural, music, and finance industries: contributions of demographic, health, and lifestyle factors. Trends in hearing, 23, 2331216519885571.
  5. Hong, E.-S., Kim, H.-S., Hong, S. K., Pantazis, D., & Min, B.-K. (2023). Deep learning-based electroencephalic diagnosis of tinnitus symptom. Frontiers in Human Neuroscience, 17, 1126938
  6. Huang, L., Chen, M., Fan, S., Tursun, N., He, X., Li, W., & Li, S. (2025). Machine learning identification of tinnitus-related features in auditory peripheral spontaneous activity in a guinea pig noise-induced tinnitus model. Hearing Research, 465, 109371
  7. Kaggle, 2024. Audiogram Analysis for Tinnitus Detection. https://www.kaggle.com/datasets/kavehmcsd/audiogram-analysis-for-tinnitus-detection/data. Accessed: 25-11-16.
  8. Kleinjung, T., Peter, N., Schecklmann, M., & Langguth, B. (2024). The current state of tinnitus diagnosis and treatment: a multidisciplinary expert perspective. Journal of the Association for Research in Otolaryngology, 25(5), 413-425.
  9. Mai, J., Wang, X., Li, Z., Liu, J., Zhang, Z., & Fu, H. (2023). EEG signal classification of tinnitus based on SVM and sample entropy. Computer Methods in Biomechanics and Biomedical Engineering, 26(5), 580–594.
  10. Møller, A. R. (2007). Tinnitus: presence and future. Progress in brain research, 166, 3-16.
  11. Ogawa, K., Sato, H., Takahashi, M., Wada, T., Naito, Y., Kawase, T.,. & Kanzaki, S. (2020). Clinical practice guidelines for diagnosis and treatment of chronic tinnitus in Japan. Auris Nasus Larynx, 47(1), 1-6.
  12. Pryce, H., Dauman, N., & Burns-O’Connell, G. (2023). What is the burden of tinnitus?. Frontiers in Psychology, 13, 981776.
  13. Savage, J., & Waddell, A. (2012). Tinnitus. BMJ clinical evidence, 2012, 0506.
  14. Simoes, J. P., Daoud, E., Shabbir, M., Amanat, S., Assouly, K., Biswas, R.,. & Genitsaridi, E. (2021). Multidisciplinary tinnitus research: Challenges and future directions from the perspective of early-stage researchers. Frontiers in aging neuroscience, 13, 647285.
  15. Tang, D., Li, H., & Chen, L. (2019). Advances in understanding, diagnosis, and treatment of tinnitus. Hearing loss: Mechanisms, prevention and cure, 109-128.
  16. Watts, E. J., Fackrell, K., Smith, S., Sheldrake, J., Haider, H., & Hoare, D. J. (2018). Why is tinnitus a problem? A qualitative analysis of problems reported by tinnitus patients. Trends in hearing, 22, 2331216518812250.
  17. Zhang, C., Wang, F., Ding, Z., Liu, P., Xue, X., Wang, L., Jiang, Y., Zhang, Z., Guo, X., Lu, Q., Liu, J., Peng, X., Ma, Y., Chen, J., Shen, W., & Yang, S. (2026). XTinnitusNet: Multi-view robust model ensemble for EEG-based tinnitus diagnosis with noisy labels. Biomedical Signal Processing and Control, 112, 108434
  18. Ogawa, K., Sato, H., Takahashi, M., Wada, T., Naito, Y., Kawase, T.,. & Kanzaki, S. (2020). Clinical practice guidelines for diagnosis and treatment of chronic tinnitus in Japan. Auris Nasus Larynx, 47(1), 1-6.
  19. Simoes, J. P., Daoud, E., Shabbir, M., Amanat, S., Assouly, K., Biswas, R.,. & Genitsaridi, E. (2021). Multidisciplinary tinnitus research: Challenges and future directions from the perspective of early-stage researchers. Frontiers in aging neuroscience, 13, 647285.