[FieldTrip] Comparing spatio-temporal data with FieldTrip toolbox

[Schoffelen, J.M. (Jan Mathijs)]

Dear Toby,

I think that this is not altogether too difficult. What you would need to do is to put your numeric data into a structure that FieldTrip can work with.

Specifically, if you manage to create a so-called ’timelock’ or ‘freq’ representation of your data, you can use ft_timelockstatistics or ft_freqstatistics for the statistical inference. (as a side note, I think it’s up to you to think whether it makes sense to use the spatial clustering heuristic for family-wise error control when comparing the connectivity matrices; yet, you can still do a permutation test to test the null hypothesis of exchangeability between groups).

Long story short for the 12x5 NMI data, I’d create 2  data structures, let’s call them freq1 (intervention group) and freq2 (controls), with the following fields

freq1.label = {‘thenameofthisisnotrelevant’};
freq1.freq = 1:12
freq1.time = 1:5
freq1.dimord = ‘rpt_chan_freq_time’;
freq1.powspctrm = zeros(number-of-subjects, 1, 12, 5);
for i = 1:nsubj

  freq1.powspctrm(i,1,:,:) = nmi; % this should yield a 1 x 12 x 5 matrix

end

and the same thing for freq2.

Then you can use ft_freqstatistics for statistical inference, with optional clustering for multiple comparison correction. In this case, the clustering will take place across the ‘freq’, and ‘time’ dimensions, which in your case boils down to spatial clustering across adjacent electrodes in the x and y directions, respectively.

For the connectivity matrices, I’d convert the single subject matrices into a vector (using the lower triangular part only), but I’d say: first things first.

Best wishes, Jan-Mathijs



On 11 Nov 2018, at 21:40, Toby Steven Waterstone <twater14 at student.aau.dk<mailto:twater14 at student.aau.dk>> wrote:


Hi,

I am analyzing high density surface electromyography (HD-sEMG) data, where I am calculating the similarities between each channel in the sensor grid. This is done in two conditions to see if there are any effect from a specific intervention, from 12 subjects, where pre and post recordings has been performed for both a control and intervention group. For the similarity measure I have calculated normalised mutual information (NMI), where I have two different outcome measures from my calculations. NMI heatmaps (12x5 matrices) representing the aggregated NMI over the HD-sEMG sensor grid, and connectivity maps (60x60 matrices) representing the similarity between each electrode pair. I have attached examples of the data, so you get the idea.

I am trying to find a way to compare this data, which has a spatio-temporal pattern. The FieldTrip toolbox has a lot of functions to analyse EEG/MEG data and as my data (HD-sEMG) has many of the same characteristics as EEG data, this toolbox might be useful to analyse my data. But in my case, I have already analysed the EMG signals by calculating NMI, and only need to do the statistical comparison now. FieldTrip is provided with a couple of statistical functions, but I'm unsure about how to use these on my data and if it is possible with this toolbox. The approach I'm looking into is FieldTrip cluster-based permutation tests<http://www.fieldtriptoolbox.org/tutorial/cluster_permutation_timelock>, but it cannot directly be transferred to my case, because I am comparing the already analysed data. So my data kind of already have the data structure of the output from the analysis of this tutorial (the plots).

So my question is: How can I use the statistical functions of the FieldTrip toolbox for comparison of NMI (spatio-temporal data)? If this is possible? I want to compare the heatmaps (12x5 vs 12x5 data structures) to each other and connectivity maps (60x60 vs 60x60 data structures) to each other.

Thank you

Best regards
Toby S. Waterstone

<Heatmap.png><Connectivity_map.png>_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
https://doi.org/10.1371/journal.pcbi.1002202

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20181114/037e9550/attachment-0001.html>