Hi Ana.

It is not surprising that your SNR is decreasing drastically when you go from 2.5mm isotropic to 2mm isotropic as your resolution (cubic) is roughly half (i.e. 2 x 2 x 2 = 8mm3, and 2.5 x 2.5 x 2.5 = 15.625mm3). Acquiring 2 x 2 x 2 would be great, but 2.5 x 2.5 x 2 .5 would also be acceptable as long as you have good SNR. What sort of brain coil are you using?

I have done a lot of multishell MRI... Instead of bvalues of 400, 700 and 2800, I suggest you use 700, 1000, and 2800, with number of directions (for example) as 20 (b=700), 45 (b=1000), 75 (b=2800). This will create data suitable for HARDI and NODDI.

I think interleaved b=0 every 10 directions may not be necessary. Instead, interleaving them every 20 volumes should be enough.

Are you using multiband? If so, try using a factor of 3 or 4. In terms of acceleration, I suggest you use a factor of 2 (using a higher acceleration factor will result in substantially reduced SNR, and possibly artifact in the midline structures). Does your sequence have CAIPI? It makes a huge difference to the acquired data.

Are you also acquiring a quick (6 direction) blipped (reverse phase-encoding) sequence? Then you can correct for distortions in the typical regions of EPI distortion (frontal and temporal) using FSL's topup module and then feed the output into eddy.

In terms of TR/TE etc, that is something you will have to work out as they will be different for every sequence and scanner.

Regards,

Jerome 

Hi Ana.

It is not surprising that your SNR is decreasing drastically when you go from 2.5mm isotropic to 2mm isotropic as your resolution (cubic) is roughly half (i.e. 2 x 2 x 2 = 8mm3, and 2.5 x 2.5 x 2.5 = 15.625mm3). Acquiring 2 x 2 x 2 would be great, but 2.5 x 2.5 x 2 .5 would also be acceptable as long as you have good SNR. What sort of brain coil are you using?

I have done a lot of multishell MRI... Instead of bvalues of 400, 700 and 2800, I suggest you use 700, 1000, and 2800, with number of directions (for example) as 20 (b=700), 45 (b=1000), 75 (b=2800). This will create data suitable for HARDI and NODDI.

I think interleaved b=0 every 10 directions may not be necessary. Instead, interleaving them every 20 volumes should be enough.

Are you using multiband? If so, try using a factor of 3 or 4. In terms of acceleration, I suggest you use a factor of 2 (using a higher acceleration factor will result in substantially reduced SNR, and possibly artifact in the midline structures). Does your sequence have CAIPI? It makes a huge difference to the acquired data.

Are you also acquiring a quick (6 direction) blipped (reverse phase-encoding) sequence? Then you can correct for distortions in the typical regions of EPI distortion (frontal and temporal) using FSL's topup module and then feed the output into eddy.

In terms of TR/TE etc, that is something you will have to work out as they will be different for every sequence and scanner.

Regards,

Jerome 

Hi Jerome,

Thank you very much for your answer. Sorry for taking time to answer, I did not see it. I will take your advice a I will use 700, 1000, and 2800 b-values. We do not use multiband because as I understand it is not fully optimized in philips resonances.

I have never hear about CAIPI but I investigate about it. Is it a option in the resonance? I read that it is for the aliasing. 

We have tried to reconstruction our own raw data but our machine return files with extension .data and .list, but the reconstruction that I know (gadgetron) only do with files with extension .*.lab, *.raw and *.sin. Could you give me some advice about this?

Thank you very much.

Cheers,

Ana.

Hi Ana,

CAIPI is Controlled Ailiasing In Parallel Imaging. It has to be incorporated into the pulse sequence to be used. I have found it extremely useful as the data is much better than without using CAIPI, but only for multiband. Its full name is CAIPIRINHA (https://www.ncbi.nlm.nih.gov/pubmed/?term=caipirinha+setmompop).

Without multiband, those number of directions will take a very long time to acquire (I am guessing around 20 to 30 minutes)...

I think the data extensions .data and .list are proprietary formats of Philips, so I don't know how to utilise those formats. I think you should ask your local Philips applications specialist about that as they are likely to have their own reconstruction pipelines that call those files with those particular extensions.

Regards,

Jerome 

Hi Jerome,

thank you very much for you answer. What resonance do you use that it has multi band? (Siemens, Philips,...) or the multi-band is it a patch that you did it?

Regards,

Ana.

Hi Ana,

Different vendors call it different names. GE calls it HyperBand, and Siemens calls it Simultaneous Multi Slice (SMS).

regards,

Jerome