Resources for the paper titled "EEG-GCNN: Augmenting Electroencephalogram-based Neurological Disease Diagnosis using a Domain-guided Graph Convolutional Neural Network". Accepted for publication (with an oral spotlight!) at ML4H Workshop, NeurIPS 2020.
In the code of this repository, there are many files that have not been found, such as "standard_1010.tsv.txt", "master_metadata_index.csv", "psd_features_data_X", "labels_y", "spec_coh_values.npy", can you complete these files, thank you very much!
Hi, neerajwagh!
The EEG-GCNN is a great job, thank you for sharing the related code!
May I ask why you use geodesic distance to build the spatial adjacency matrix, is it to extract spatial features?
If so, since the position of the electrodes on the EEG cap is fixed, is the spatial adjacency matrix the same for each sample?
Hello author,
The work you post is very good. We want to take this as a base for our model implementation. Currently, with this code, we are facing the following issues:
Dataset - we have tried to download the dataset from MP-Lemon as we r unable to register with the TUH EEG set. The problem while downloading the MP Lemon is we tried to download Raw data for 1 subject, and it showed 3 formats under that .edf format is not available. We request you, please add the dataset in .edf format in the repository.
Can you connect with me to better understand your model implementation?
Hi,
When I try the test run with the pre processed data for the deep gcnn and the shallow gcnn i get the following error
and I'm having trouble figuring it out, do you have any suggestion?
I am trying to find the file "standard_1010.tsv.txt". You use it in the EEGGraphDataset class in the method get_sensor_distance. I am not able to find it in the repo, neither in the TUH dataset server or in your figshare. Is is possible to obtain this file?
Hello author, your code has taught me a lot, but I have a question I would like to ask you.
In your thesis, you used 1385 subjects in TUAB and marked them as diseased, but I checked the content on the TUAB official website and found that these 1385 are theoretically normal EEG signals. Is there a mistake in my understanding, thank you for your guidance.
In your paper, the adjacency matrix is stated to be a weighed combination of structural and functional connectivity A_ij = 0.5 * (A_ij^s + A_ij^f). However, in your code the adjacency matrix is binary. Are the presented results from a binary or weighted adjacency matrix?
Hello, I am trying to adapt this code for a model that predicts future illnesses related with old age, and would like to kindly ask you for the training script. I have preprocessed the data and am missing a script to train the deep graph-convolutional neural network. Thank you for your help
Hello author, thank you again for your sharing, it has benefited me a lot. I have a question to discuss with you, why did you only consider using PSD to extract features, but did you consider using time-frequency features, which are theoretically more appropriate for epileptic signals? Best wishes~
Hello,
I'm noticing that the file to prepare other eeg and the eeg pipeline the oldest ones, and when I use them to produce my csv file it has some differences from the one on Figshare such as: the presence of the numeric label and the generic order of the columns.
Is that procedure sill valid to process my own eeg or there is an updated one?