Really clever reasoning in this idea. It considers the hemodynamic response across the cortex as a spatial-temporal phenomenon. (Like 4D space-time!) If you use or are interested in ME-ICA for fMRI, you should read this paper. - ThreadSky

practicalfmri.bsky.social • 59 days ago

Really clever reasoning in this idea. It considers the hemodynamic response across the cortex as a spatial-temporal phenomenon. (Like 4D space-time!) If you use or are interested in ME-ICA for fMRI, you should read this paper.

Reposted from bioRxiv Neuroscience

Differentiating BOLD and non-BOLD signals in fMRI time series using cross-cortical depth delay patterns https://www.biorxiv.org/content/10.1101/2024.12.26.628516v1

Comments

practicalfmri.bsky.social•59 days ago

One Q I have: I don't see why the layer immediately above white matter (D1) should be faster than D2 if D2 includes input cortical layer IV (Fig 1). Perhaps D1 is largely irrelevant. That D3 is later than D2 by an appropriate lag (Figs 2 & 3) is probably key.

danielhandwerker.bsky.social•59 days ago

One critique of ICA approaches I got from @blaisefrederick.bsky.social is that vascular artifacts will be dominated by inter-regional delays that standard ICA might not capture as well. I'll need to read the preprint more carefully to see how they deal with this. 1/2

practicalfmri.bsky.social•59 days ago

I don't think ME-ICA or this new method can deal with true BOLD vascular artifacts, e.g. non-stationary blood gases. Gladly, they make this point in the Discussion. But I do wonder how this new method (and ME-ICA) would fare post RapidTiDe lag correction.

danielhandwerker.bsky.social•59 days ago

We can include temporal info into these methods that can detail with regional lags, like RapidTIDe does. I merely need time to actually work on this! As for terminology. I'm fairly comfortable saying non-BOLD==bad (unlikely to be neural) and I try to be clear that BOLD does NOT mean good/neural.

practicalfmri.bsky.social•59 days ago

But inflow weighting (via short effective T1) on arterial side is non-BOLD and ironically likely to be closer to activating neurons than draining veins :-) Running these approaches with fMRI collected at very small vs Ernst flip angles would be interesting.

danielhandwerker.bsky.social•59 days ago

That said, one reason I've been excited about multi-echo denoising is that it reopened discussions on using lower-level properties of fMRI signals for better noise removal. This preprint is a great and promising example of that. 2/2

practicalfmri.bsky.social•59 days ago

Seconded! Anything which shines a spotlight on (so-called) denoising is good. Even the very terminology we use (confounds, regressors of no interest, physiologic noise) is biased against the holistic nature of brain-blood-body circuitry. BOLD = good, non-BOLD = bad is already a false boundary.

Comments

Posting Rules

Reply