Phase Locking Value - Dynamic connectivity

I'm currently investigating connectivity patterns between the anterior and posterior insula and specific pain-related brain regions, including the anterior, mid, and posterior cingulate cortex. Specifically, I aim to discern these patterns in individuals with chronic pain during electrical stimulation.

Having undergone Mike Cohen's course and delved into several papers authored by him and others, I've surmised that the Phase Locking Value (PLV) is the most suitable metric for my study. This is because my research question is driven by a specific hypothesis rather than being exploratory in nature.

Moreover, I'm keen to explore the impact of the stimulus on brain connectivity among participants. This has led me to consider a dynamic connectivity analysis. While I've noted that Brainstorm may not directly support a sliding window approach for PLV, I'm contemplating a manual implementation of the same:

  1. Segment the continuous data based on a predefined window position. For instance, in a 2-second continuous recording with a 500 ms window and 100 ms step, the data segments would be from 0-500 ms, 100-600 ms, 200-700 ms, etc.
  2. Compute the PLV for each segment using Brainstorm's connectivity tools.
  3. Record the PLV value and proceed by moving the window according to the step size, repeating the process.

I'd love to hear your insights on this approach. Alternatively, would you recommend any other time-resolved phase-based metric such as correlation or coherence?

On a related note, if you're looking to self-educate on connectivity, I highly recommend Mike Cohen's YouTube lectures. You can access them

Thank you for your time and input!

1 Like

I now found this option with PLV:

Keep time information: Computes PLV across trials instead of across time: the outcome is a PLV time series for each frequency band.

This is actually what I need, but it says that it requires many trials. What can be seen as many trials?

I have 66 stimuli per participant.

From the connectivity tutorial

PLV values tend to be overestimated when derived from few (<50) time samples. As a consequence, when comparing PLV values between conditions, please make sure that they are derived from about the same number of samples in each condition.

Above 50 trials reduced the chance of getting spurious PLV results. You can also check the Intertrial Phase Clustering chapter in:

Cohen MX. Analyzing Neural Time Series Data: Theory and Practice. MIT Press; 2014