It depends on the temporal and spatial resolution you need to answer your hypothesis. EEG for instance has millisecond temporal resolution but very poor spatial resolution. fMRI on the other hand has good spatial resolution (a few millimeters on a high T machine) but poor temporal resolution (over 2 seconds).  

It's become common practice to include at least two of these techniques into an experimental paradigm to cover both temporal and spatial resolutions.

Dear Osama,

The term cognitive state is rather ambiguous term used in the literature of neuroscience. Even there are researchers suggesting that it is untrue to use this term in the context of neuroimaging because these techniques (those you mentioned above) measure changes in the function of brain and it is difficult (if not impossible) to infer state from these functions. So, it really depends on your exact question to choose one of these techniques. I think Nevah and Micah already gave best advice, I just can suggest you to visit IPM, IRICSS and institute for brain and cognition at Shahid Beheshti Uni to get to know people who use these techniques in their research. 

Dear Ali,

I am agree that cognitive state is an ambiguous term, but can you tell me how do we study cognitive state without neuroimaging techniques?

Dears Micah and Neva

I think the human connectome project is the  choice of my works http://www.humanconnectome.org/

I need to apply some data fusion method to get more complement information on specific cognitive state.

Dear Osama,

I agree with all that the cognitive state is an ambiguous term, but in order to study any brain state in vivo I would suggest to use a multimodal approach, given that every single neuroimaging modality gives you different information about the same process, normally a more global view of a single process will improve our understanding and give a more general answer instead of just a part of it.

The connectome would be one option, but you must always consider about the neurobiological meanning of the research technique ;)

Cheers

In my opinion there is no such a thing as the most efficient neuroimage modality. Compromise between temporal and spatial resolution is certainly important but leaves out other considerations; whether you are looking for a structural or functional image, whether portability, subject motion and ecological validity are important, whether purpose is for research, bedside monitoring or a clinical test, invasiveness, whether your goal is to establish activity loci following a certain stimulus (segregational study), information processing (causal profiles), connectivity (of whatever kind -integrational study) or perhaps understanding the neurovascular coupling, etc. Also, many of the neuroimages modalities do have submodalities which do "see" different things; e.g. MRI has FLAIRS, EPI-BOLD, T1-weighted, DTI, etc, or similarly NIRS has CW, frequency-domain, TRS, etc. Multi-modal neuroimaging looks good in publications, but is not always necessary.

Dear Gabriel,

I think that you hit the right thing, "multimodal approaches", but were can i find data repositories that contain multimodal functional neuroimages. I am agree with you in the every study the neurobiological meaning would not be neglected. I think this is maybe the big challenge.

Well it will depend on your research question, there are several data repositories, some of them dedicated to specific research questions (eg. resting-state, mental disorders: Pain, AD, PD), some of the most used data repositories:

http://www.incf.org/resources/research-tools/resources_listing_view?getTopic=Neuroimaging&getResource_type=Data%20repository

also, the adni database contains many other biomarkes for AD

some useful tools:

https://github.com/LTS5/

But there are a few more.

Bests

Hybrid fNIR-EEG is the best choice. I am working with it now. Dear, Osama Hourani, if you need we can discuss about it based on my capacity.

Regards

Md. Asadur Rahman