Goodevening. 

The more "Traditional" way and software to localize areas in fMRI is SPM it's main drawback is that needs the regressors (time courses of the experiments condacted) and you have to assume an HRF function which may not be correct.

Other tools which I consider (at least at my point of view) more modern and  more accurate since less assumptions are needed are FSL and GIFT (they mainly use different approaches of ICA).

If you will perform multi subject analysis maybe it will be helpful to use tensor based methods such as TPICA, CPD, BTD or IVA (which is not tensor based but still oriented towards multi subject cases)

Thanks Noham, this is helpful!

I am trying to grasp if there is any sort of consensus, or something that resembles it, when considering what is the best task / best mask / best scan for localization of these structures.

@ Christos - thanks for the detailed response. In my question I was not interested in ways to analyze the data, but rather optimal data collecting strategies.

Hi Mikhail,

Thanks for the reply!

Why do you prefer the across-subject definition rather than subject-based localizers?

I was under the impression that as some of these regions are fairly small, slight anatomical differences between subjects can make the across-subjects coordinates less reliable.

Hi Mikhail,

Thanks for the detailed response!

Indeed I want to capture the VTA and NAcc (in addition to other regions), as well as the other regions. Is the MID task reliable for the VTA and NAcc as well?

Having done some research in neuroscience of memory, I know that people's hippocampi differ a lot in terms of size and shape. Using individually-defined (anatomical) masks of the hippocampus can lead to different results compared with group/atlas-based definition of the hippocampus. The normalization procedure does not eliminate the between-participants variability, at least for this region (in both SPM and FSL).

Drawing on that experience, I suspect that the VTA and NAcc might be susceptible to the same type of variability, so I am looking for ways to overcome that.

I prefer to avoid drawing these regions for each participant. But if that needs to be the case - that I do need to draw - I would like to have the best anatomical scan for that. From what I've gathered so far, it seems like the standard MPRAGE/SPGR is not optimal for differentiating these regions, so I am looking for the best protocol for anatomical scans as well.

Hi Mikhail,

Thanks again!

A common practice in the human Hippocampus literature is to extract data from each subject based on their individual mask and use these data. In recent years there is a developing standartization of the manual segmentation protocols to make the procedure as transparent and reproducible as possible.

There are also automatic tools for that now - FSL's and FreeSurfer's segmentation algorithms are pretty good most of the time. IMHO, using data from the actual hippocampus of participants (as defined by individual masks) is better than using data which originates in the hippocampus for only some of them. This is especially true when you get down to the resolution of sub-parts of the structure.

If you want to use custom anatomical ROIs (e.g. in patients), you may want to acquire proton density (PD)-weighted images. They have very good grey/white matter contrast in subcortical structures, and you can simply segment the ventral striatum manually. You can't see the VTA directly, but you can readily recognize the substantia nigra (SN), and the VTA is located just anterior-medially to the SN. And be aware that the SN/VTA distinction is not that clear-cut in humans, and particularly the medial SN also activates in tasks that would typically engage the mesolimbic rather than the nigro-striatal dopamine system (see, e.g., the review by Düzel et al., TINS, 2009)

Thanks a lot for the helpful info!