Application of Artificial Intelligence in Tinnitus Diagnosis and Treatment: A Pilot Study
Tinnitus is the perception of phantom sound when there is no sound source and is often described as ringing in the ear. Symptoms of tinnitus and severity vary significantly from person to person, largely be attributed to an individual’s perception to the condition and increasing amount of...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10966921/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Tinnitus is the perception of phantom sound when there is no sound source and is often described as ringing in the ear. Symptoms of tinnitus and severity vary significantly from person to person, largely be attributed to an individual’s perception to the condition and increasing amount of people seek medical intervention. Unlike other diseases, tinnitus pathophysiology is complex and sometimes inexplicable. Currently, there are no universally accepted treatment options for this condition. Moreover, there is no well-established correlation between tinnitus features and projection of treatment. In practice, the treatments provided by practitioners are not based on defined and regulated rules or expected patient outcomes. Instead, they are at a certain level and differ significantly among clinicians and across regions. The complexity of tinnitus features and lack of well-adapted prognostic treatments present an excellent opportunity for Artificial Intelligence (AI) and Machine Learning (ML). AI models can learn intricate patterns between tinnitus features and treatments, as suggested by experts. In this study, we trained an AI model with an expert system to predict tinnitus treatment based on tinnitus symptoms. We describe the curation of the input data, the algorithm (CTGan) used to extend the dataset, and the ML model (Random Forest) to predict each of the suggested treatments. The average accuracy score of the trained model is currently between 0.81 and 0.96 for most treatment predictions, based on a small input dataset size of 300 samples. The generalizability of these results to broader clinical settings requires further validation using larger real-world datasets. |
|---|---|
| ISSN: | 2169-3536 |