Skip to main content

Please note: this event has passed


Abstract: 

Background: The mechanism of human neural responses to different stimuli has always been of interest to neuroscientists. In clinical situations, tools to distinguish different diseases or states are required. However, classic classification methods have obvious shortcomings: traditional clinical categorical methods may not be competent for behaviour prediction or brain state classification and traditional machine learning models are improvable in classification accuracy. With the increasing use of convolutional neural networks (CNN) in neuroimaging computer-assisted classification, an ensemble classifier of CNNs might be able to mine hidden patterns from MEG signals.

Methods: To address this brain state classification modelling issue, I used MEG signals from 28 participants viewing 14 image stimuli to train the CNN. The CNN subsequently underwent 10-fold cross-validation to ensure proper classification of MEG. I also extracted the relative power spectrum and provided it to the network. The following main techniques were applied in this research: principal component analysis (PCA), convolutional block spatial and temporal features extracting modules, convolutional block attention module (CBAM) techniques, relative power spectrum (RPS) techniques, and fully connected (FC) techniques.

Results: Among all datasets from 28 participants, the average classification accuracy is 23.07%±7.69%, which is better than the baseline models: LSTM RNN model 15.38% (p = 6.8 × 10 –2) and a simple image classification CNN model 11.53% (p = 5.9 × 10 –2). Relative power spectrum information (mainly beta and delta during this task) successfully informed the model improving its performance.

Biography:

Lei Luo, a data scientist and former cognitive neuroscience master's student in 2022 at King's College London. I committed to applying computational modelling techniques to neuroimaging data to understand the mechanisms of brains.

Click here to join the meeting