Abstract
Cerebral stroke is a common disease across the world, and it is a promising method to recognize the intention of stroke patients with the help of brain–computer interface (BCI). In the field of motor imagery (MI) classification, appropriate filtering is vital for feature extracting of electroencephalogram (EEG) signals and consequently influences the accuracy of MI classification. In this case, a novel two-stage refine filtering method was proposed, inspired by Gradient-weighted Class Activation Mapping (Grad-CAM), which uses the gradients of any target concept flowing into the final convolutional layer to highlight the important part of training data for predicting the concept. In the first stage, MI classification was carried out and then the frequency band to be filtered was calculated according to the Grad-CAM of the MI classification results. In the second stage, EEG was filtered and classified for a higher classification accuracy. To evaluate the filtering effect, this method was applied to the multi-branch neural network proposed in our previous work. Experiment results revealed that the proposed method reached state-of-the-art classification kappa value levels and acquired at least 3% higher kappa values than other methods This study also proposed some promising application scenarios with this filtering method.
Highlights
Cerebral stroke (Albers and Olivot, 2007; Menon and Demchuk, 2011) is one of the most common diseases, and disorder in functions related to language and motor makes it hard for stroke patients to live a normal life
All the motor imagery (MI) classification methods can be generally divided into two categories: Common Spatial Pattern (CSP) (Kang et al, 2009; Lu et al, 2010)-based methods, such as Filter Bank Common Spatial Pattern (FBCSP) (Ang et al, 2012) and Common SpatioSpectral Pattern (CSSP) (Lemm et al, 2005); and Deep learning based methods (Suwicha et al, 2014; Schirrmeister et al, 2017), such as C2CM (Sakhavi et al, 2018), Compact convolutional neural network (Lawhern et al, 2016), and shallow ConvNet (Schirrmeister et al, 2017)
The two-stage refine filtering method can be divided into two stages: In the first stage, raw EEG data were trained in the network and the frequency bands to be filtered were selected according to the Grad-CAM results of the last convolutional layers
Summary
Cerebral stroke (Albers and Olivot, 2007; Menon and Demchuk, 2011) is one of the most common diseases, and disorder in functions related to language and motor makes it hard for stroke patients to live a normal life. Gradient-weighted Class Activation Mapping makes a good visual explanation by highlighting the important regions of predicted images according to the last convolutional layer of the network In this case, we considered whether it is possible to improve the performance of our network by preserving the useful frequency (which means these frequencies contribute a lot to correct classification results) and suppress useless frequency (which means these frequencies contribute nothing to correct results). The two-stage refine filtering method can be divided into two stages: In the first stage, raw EEG data were trained in the network and the frequency bands to be filtered were selected according to the Grad-CAM results of the last convolutional layers. Considering the huge amount of training data, a tenth of the data were sampled for Grad-CAM
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
