Related Experiment Video
Updated: Sep 3, 2025
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
725
EMD-based data augmentation method applied to handwriting data for the diagnosis of Essential Tremor using LSTM
José Fernando Adrán Otero1,2, Karmele López-de-Ipina3,4, Oscar Solans Caballer1,2
1Faculty of Computer Science, Multimedia and Telecommunications, Open University of Catalonia, Barcelona, 08080, Spain.
Scientific Reports
|July 27, 2022
Summary
This study introduces a novel data augmentation method to enhance deep learning for essential tremor diagnosis. The technique significantly improves diagnostic accuracy, especially with limited patient data.
Area of Science:
- Medical technology
- Computer science
- Signal processing
Background:
- Deep learning models require substantial data for accurate disease diagnosis.
- Essential tremor diagnosis can be challenging, necessitating advanced analytical methods.
- Current deep learning approaches for temporal signal analysis face data limitations.
Purpose of the Study:
- To propose and evaluate a novel data augmentation technique for improving essential tremor diagnosis using Long Short-Term Memory (LSTM) networks.
- To address the challenge of limited data availability in training deep learning models for medical diagnoses.
- To enhance the accuracy and reliability of AI-driven diagnostic tools for neurological disorders.
Main Methods:
- Utilized multivariate Empirical Mode Decomposition (EMD) to decompose temporal signals from essential tremor patients and control subjects.
- Generated artificial time-series samples by randomly shuffling and combining decomposed signal components.
- Trained an LSTM network using a combination of augmented and real data, with remaining real data used for testing.
Main Results:
- The proposed data augmentation method significantly outperformed 10 other augmentation techniques.
- Achieved a notable increase in classification accuracy from 83.20% to nearly 93% in the best-case scenario.
- Demonstrated the effectiveness of the method in improving diagnostic performance with limited datasets.
Conclusions:
- The novel data augmentation technique is highly effective in enhancing deep learning model accuracy for essential tremor diagnosis.
- This approach offers a promising solution for developing robust diagnostic tools when dealing with small medical databases.
- The method shows potential for broader application in AI-assisted medical diagnosis across various conditions.
