Related Experiment Video
Updated: Aug 15, 2025
07:05
A Protocol for the Administration of Real-Time fMRI Neurofeedback Training
Published on: August 24, 2017
11.1K
Multi-tasking deep network for tinnitus classification and severity prediction from multimodal structural MR images
Chieh-Te Lin1, Sanjay Ghosh1, Leighton B Hinkley1
1Department of Radiology and Biomedical Imaging, University of California San Francisco, 513 Parnassus Ave, San Francisco, CA 94143, United States of America.
Journal of Neural Engineering
|January 3, 2023
Summary
This study introduces a novel deep learning framework using magnetic resonance imaging (MRI) to accurately classify tinnitus and predict its severity. The multimodal approach enhances diagnostic capabilities for this phantom auditory disorder.
Area of Science:
- Neuroimaging
- Machine Learning
- Auditory Disorders
Background:
- Subjective tinnitus is a phantom auditory disorder lacking objective biomarkers.
- Current diagnostic methods for tinnitus lack speed and efficiency.
- Subtle brain changes in magnetic resonance images (MRIs) may offer diagnostic insights.
Purpose of the Study:
- To develop and evaluate a data-driven machine learning framework for tinnitus classification and severity prediction.
- To investigate the utility of multimodal structural MRI (sMRI) data for tinnitus diagnosis.
- To identify key brain components influencing diagnostic performance.
Main Methods:
- A deep, multi-task, multimodal framework was developed using structural MRI (sMRI) data.
- Two MRI modalities, T1-weighted (T1w) and T2-weighted (T2w), were integrated.
- sMRI images were segmented into cerebrospinal fluid, grey matter, and white matter to assess component contributions.
Main Results:
- The multimodal framework effectively utilized information from both T1w and T2w MRI modalities.
- The model demonstrated superior performance in joint tinnitus classification and severity prediction.
- The framework achieved high accuracy, sensitivity, specificity, and negative predictive value.
Conclusions:
- The proposed deep learning framework offers a promising, data-driven approach for tinnitus diagnosis and severity assessment.
- Multimodal sMRI analysis, leveraging both T1w and T2w images, enhances diagnostic accuracy for tinnitus.
- This method has the potential to significantly advance clinical practice in managing tinnitus.
