Monday 03 March 2025
The quest for a more efficient way to decode brain signals has been a long-standing challenge in the field of neuroscience. The ability to read and interpret brain activity could have significant implications for treating neurological disorders, enhancing cognitive abilities, and even developing new forms of communication.
Researchers have made progress in recent years by combining electroencephalography (EEG) with eye movement signals to improve decoding accuracy. However, existing methods often rely on simple concatenation or averaging techniques, which can be inefficient and prone to noise.
A new study has introduced a more sophisticated approach by developing a neural network that fuses EEG and eye movement signals in a hierarchical manner. The system, known as MTREE-Net, utilizes a dual-stream feature extractor to extract multi-scale EEG features and compress the features of eye movement components. This allows for more accurate classification of brain activity and improved robustness against noise.
The researchers designed three independent multi-class target RSVP tasks and created an open-source dataset that includes both EEG and eye movement signals. They then tested MTREE-Net on this dataset, achieving significantly higher decoding accuracy compared to existing methods.
One of the key innovations of MTREE-Net is its ability to selectively focus on relevant features from both modalities. This is achieved through a dual-complementary module that simultaneously improves the differentiation of uni-modal features across classes. Additionally, a hierarchical self-distillation module is used to reduce misclassification of non-target samples by transferring knowledge from a binary classifier to guide a triplet classifier.
The study demonstrates the potential of MTREE-Net for real-world applications, including brain-computer interfaces and neurological disorder diagnosis. By combining the strengths of EEG and eye movement signals, this system could provide a more accurate and robust means of decoding brain activity.
The researchers acknowledge that there are still challenges to be overcome before MTREE-Net can be used in clinical settings. However, their approach represents an important step forward in developing more effective brain-computer interfaces and has the potential to revolutionize our understanding of the human brain.
Cite this article: “Decoding Brain Signals with Enhanced Accuracy Using MTREE-Net”, The Science Archive, 2025.
Neuroscience, Brain Signals, Eeg, Eye Movement, Neural Network, Hierarchical Feature Extraction, Noise Reduction, Classification Accuracy, Brain-Computer Interfaces, Neurological Disorders.
