Multi-modality NDE fusion using encoder-decoder networks for identify multiple neurological disorders from EEG...

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Diagnosing neurological disorders like epilepsy, Parkinson's, schizophrenia, stroke, and Alzheimer's is challenging due to complex brain activity.
• Traditional methods struggle with integrated analysis of diverse data sources.
• Multi-modal data and neural networks offer potential for improved diagnostic accuracy.

Purpose of the Study:

• To propose a novel approach integrating Nondestructive Evaluation (NDE) data with electroencephalography (EEG) signals.
• To leverage advanced neural networks for enhanced diagnosis of neurological disorders.
• To identify and correlate shared latent features across heterogeneous NDE datasets.

Main Methods:

• EEG signals were transformed into 2D scalogram images using wavelet signal processing.
• An encoder-decoder neural network was employed to extract and correlate features from EEG and NDE data.
• The method was evaluated on datasets containing EEG and NDE data for neurological disorder identification.

Main Results:

• Significant improvements in diagnostic accuracy and efficiency were observed.
• The encoder-decoder network successfully identified shared latent features across heterogeneous NDE datasets.
• The fusion of multi-modal NDE data with EEG signals enabled robust automatic identification of multiple neurological disorders.

Conclusions:

• This approach represents a substantial advancement in neurological disorder diagnosis.
• Integrating diverse NDE data with EEG signals enhances accuracy and efficiency for multiple neurological conditions.
• This multi-modal data fusion method has the potential to revolutionize neurological diagnostic practices.