[pymvpa] Accounting for the haemodynamic lag

[Nick Oosterhof]

On Jan 20, 2014, at 10:55 AM, David Watson wrote:

> Dear All,
> 
> I was wondering if anyone could give me some advice on how best to account for the haemodynamic lag of the BOLD signal when performing a pattern analysis on 4D fMRI data?  [...] I get the impression that there are two main ways that people tend to do this:
> 1. Offset the timeseries or the sample labels by a suitable number of TRs. [...] But I am a little concerned as to how accurate this is likely to be, e.g. the lag is unlikely to be precisely 6 seconds in all subjects.
> 2. Convolve my model regressors with an HRF. 

Indeed, these are to my knowledge the two main options. The first one is model-free: no assumption is made about the shape of the HR. The second option assumes a particular shape of the HR. If the HR matches the used HRF well then this would be preferable, but if it deviates much then estimates of the signal get worse. 

At the same time many people (including myself) use HRFs because their use is quite straightforward, even though we know that that HRs vary across participants, brain regions, and by paradigm. (Most HRFs used are based on data in visual cortex).

One potentially nice approach is here:

https://pypi.python.org/pypi/hrf_estimation

Yarik has done some work with this package but I don't know how well it worked.  To my knowledge it is not part of PyMVPA (yet).

> [The HRF] option seems like it might be preferable, and I can easily enough derive a gamma HRF (e.g. the nitime package seems to provide one), but I'm not sure how I would then apply this to a given model within PyMVPA. 

Do you have to? Alternatively you could get HRF-based response estimates using an external package (e.g. AFNI, SPM), then load these estimates in PyMVPA.