Normalizing sources of cyclical EEG data (heartbeat-evoked responses)


I would like to estimate group-level sources for EEG data using minimum norm current density map, constrained, with standard anatomy. I'm using Brainstorm version 19-Apr-2024. From the tutorials, I understand it is best to first compute sources per subject, then z-score them using a baseline window, and finally average z-scored sources across subjects. While this approach works well for a dataset of stimulus-evoked responses with a pre-stimulus baseline window, I would like to estimate sources of heartbeat-evoked responses, which are cyclical in nature and thus don't have any baseline to use for z-scoring. Do you have any suggestions on what to use instead for normalization before group-level averaging?
Thank you,
Best regards,

Maybe you could consider z-scoring the source maps over the entire beat-to-beat epoch then? The z-score will show the deviation from the mean activity over an entire cycle. You may want to refer to the MEG work by C Tallon-Baudry and replicate their approach in that respect.

Dear Sylvain, thank you for your swift response. Indeed I considered using this approach. However, to cross-validate it, I tested what happens when I use whole time-window z-scoring on the same data but time-locked to a stimulus instead of the heartbeat.

With baseline z-scoring (-100 to 0 ms before stimulus onset), I was getting reasonable results: activity in the somatosensory cortex for a touch-evoked P50, activity in the auditory cortex for a sound-evoked N100. Using the signal epoch instead for z-scoring (0 to 250 ms after stimulus onset), estimated sources became nonsensical. I thus have a hard time trusting whole time-window z-scoring for heartbeat-evoked response source estimation... Do you have any thoughts on this comparison between ERPs and HERs, do you think it makes sense to try and cross-validate in this way? Could there be any other alternative normalization approach to try out?

Thank you for the reference - I am in fact working under C. Tallon-Baudry's supervision. We are trying to deepen our understanding of source estimation for HERs in particular.