Related Experiment Video
Updated: Jun 3, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
16.1K
E2E-TM: Dual-Way Feature Extraction and End-to-End Transformer Based Parkinson's Disease Diagnosis Using Integrated
Sundaram Mohanapriya1, Kamalraj Subramaniam1
1Department of Computer Science and Engineering, Coimbatore, India.
Developmental Neurobiology
|September 28, 2025
Summary
This study introduces the E2E-TM model for accurate Parkinson's disease (PD) diagnosis using MRI and EEG data. The novel transformer module achieves superior performance over existing methods, improving early detection capabilities.
Area of Science:
- Neurology
- Artificial Intelligence
- Medical Imaging
Background:
- Parkinson's disease (PD) is a neurodegenerative disorder affecting motor function.
- Current machine learning (ML) and deep learning (DL) diagnostic tools face challenges with data diversity and model generalizability.
- Accurate and early diagnosis of PD is crucial for effective patient management.
Purpose of the Study:
- To develop an advanced end-to-end transformer module (E2E-TM) for precise Parkinson's disease diagnosis.
- To overcome limitations of existing ML/DL models in PD diagnosis, particularly regarding data representativeness and overfitting.
- To enhance the accuracy and reliability of diagnostic tools for diverse patient populations.
Main Methods:
- Acquired and preprocessed magnetic resonance imaging (MRI) and electroencephalography (EEG) data.
- Applied noise reduction techniques (bilateral filter, wavelet decomposition) and Super U-Net for MRI segmentation/reconstruction.
- Developed a novel E2E-TM incorporating a transformer encoder module (TEM) with multi-scale trunk convolution (Multi-TC), dual-way trunk convolutional (DW-TC), and dual-parallel attention network (DPANet), followed by a transformer decoder module (TDM).
Main Results:
- The E2E-TM model successfully integrated and processed multimodal MRI and EEG data.
- Feature extraction and dimensionality reduction were efficiently handled by the TEM components (Multi-TC, DW-TC, DPANet).
- The proposed E2E-TM model demonstrated superior diagnostic performance compared to established baseline approaches.
Conclusions:
- The E2E-TM model offers a robust and effective solution for the precise diagnosis of Parkinson's disease.
- This approach shows promise in improving the generalizability and accuracy of diagnostic tools for PD across diverse demographics.
- The study highlights the potential of advanced deep learning architectures in neurodegenerative disease diagnostics.
