English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

In brain signal processing, deep learning (DL) models have become commonlyused. However, the performance gain from using end-to-end DL models compared toconventional ML approaches is usually significant but moderate, typically atthe cost of increased computational load and deteriorated explainability. Thecore idea behind deep learning approaches is scaling the performance withbigger datasets. However, brain signals are temporal data with a lowsignal-to-noise ratio, uncertain labels, and nonstationary data in time. Thosefactors may influence the training process and slow down the models'performance improvement. These factors' influence may differ for end-to-end DLmodel and one using hand-crafted features. As not studied before, this papercompares models that use raw ECoG signal and time-frequency features for BCImotor imagery decoding. We investigate whether the current dataset size is astronger limitation for any models. Finally, obtained filters were compared toidentify differences between hand-crafted features and optimized withbackpropagation. To compare the effectiveness of both strategies, we used amultilayer perceptron and a mix of convolutional and LSTM layers that werealready proved effective in this task. The analysis was performed on thelong-term clinical trial database (almost 600 minutes of recordings) of atetraplegic patient executing motor imagery tasks for 3D hand translation. Fora given dataset, the results showed that end-to-end training might not besignificantly better than the hand-crafted features-based model. Theperformance gap is reduced with bigger datasets, but considering the increasedcomputational load, end-to-end training may not be profitable for thisapplication.

Deep learning for ECoG brain-computer interface: end-to-end vs. hand-crafted features