search

brainhack-school2020 / tiawei_MEG_ML

ML for finger movement source-localized MEG data
1 stars 0 forks source link

Connectivity and features #3

Open beapdk opened 4 years ago

beapdk commented 4 years ago

Hi Tania ! It's definitely a good idea to start from simplier approaches as you project suggests! I would be very interested to know what kind of connectivity analyses were performed on your data ? Also do you have an idea of the features of the signal you want to select already, As I understood you want to use features from your time-frequency analyses ?

tiawei commented 4 years ago

Hi Tania ! It's definitely a good idea to start from simplier approaches as you project suggests! I would be very interested to know what kind of connectivity analyses were performed on your data ? Also do you have an idea of the features of the signal you want to select already, As I understood you want to use features from your time-frequency analyses ?

Hi Penelope, I measured phase-coupling based connectivity (e,g. coherence, phase-lag index, and phase slope index), amplitude-coupling based connectivity (e.g, instantaneous amplitude connectivity), and multivariate model-based Granger Causality. Here is a great paper explaining each connectivity measure and summarizing its pitfalls. This paper helped me a lot! https://pubmed.ncbi.nlm.nih.gov/26778976/

Yes, the features would include different ROIs (on left and right motor cortices) and time-frequency points. The power differences in those features should be able to train the classification between left-hand and right-hand movements. :)