We’ve looked fairly extensively at this issue, and have learned a few things:
1) The T2 XETA used in the NAMIC was just a preproduction version of the sequence later called CUBE, so there is no real difference (this is per Sylvain Bouix, who was part of the NAMIC data collection). I just emailed him back to get the dicom header for the SPGR used in the NAMIC data set to see if there are any differences between it and our FSPGR. 2) We initially used a 32 channel NOVA coil for data collection, but in an effort to replicate the NAMIC dataset as closely as possible we also tried the GE 8 channel coil; while it reduced the artifact it was still present. (The transmit coil was the body coil in both cases). 3) PURE correction also doesn’t adequately compensate for the artifact. I forget what exactly happened, but when I’m at work tomorrow I’ll dig up the images and send them along. 4) Post-recon inhomogeneity correction also doesn’t do the trick. I used both ANTS and FAST (both with and without brain masking) and wasn’t able to fully deal with the issue. Often what happened was that the artifact “reversed” in that after correction the edge of the brain had artifactually boosted T1/T2 ratio rather than the middle of the brain. In addition we had some ringing artifact in the T2s that we found was caused by the GE scanner’s default behavior of reconstructing the images in a different matrix size than the acquisition, and in the course of our testing found a set of parameters that seemed to increase G/W contrast in the T2. I can pass those parameters along also when I’m back at my desk. There is a mythical GE version of the MPRAGE thats floating around somewhere, but we haven’t been able to get our hands on it yet. If anyone has one we’d be happy to do some testing with it. Unfortunately GE themselves has been less the helpful on this issue (amongst others) but perhaps if there are enough of us knocking on their door they’ll be of more help. At some point I’m going to run a test on a phillips scanner that we have at Columbia where there is an MPRAGE; I’ll let you know how that works out. At this point though I don’t feel confident that the myelin map data we’ve collected so far is worth much; I had been hoping that inhomogeneity correction would help but it hasn’t so far. Again any other insight would be appreciated. __________________________ gaurav patel gauravpa...@gmail.com www.neurofreak.net > On May 31, 2015, at 4:21 PM, Glasser, Matthew <glass...@wusm.wustl.edu> wrote: > > Hi Barbara, > > I agree that the main heavily myelinated areas are found, the issue is with > that lateral cortex (e.g. the difference between STG and STS is a lot higher > than we see). > > Do you know how the PURE correction is computed? We generally avoid these > sorts of automatic corrections (even Siemens pre-scan normalize, for which we > know what it is doing), and the HCP pipelines don’t require receive fields to > be corrected for ahead of time (though they shouldn’t mind if the same > correction is done to both T1w and T2w images). If the PURE correction is > not based on acquiring additional images that can be used to directly measure > the receive field, it could well do something different to the T1w and T2w > images. I don’t know if Gaurav is using the PURE correction though. > > My question about the head coil was more specific than if you used one: Did > it provide the radio frequency transmit or was it just used to receive radio > frequency? > > I’m not sure what you’re doing with the data, so I can’t say whether it is > usuable or not. You wouldn’t be able to directly compare subjects acquired > on Siemens to those acquired on GE, and you’d need to be careful interpreting > the neuroanatomy (strangely the medial surface looks just fine…). Can you > post the myelin map data without the _BC on it so I can see what that looks > like? > > Unfortunately sorting this issue out may require consultation with the GE MR > physicists to understand why their scanners are behaving strangely here. A > spatial map of the flip angles or transmit field produced by this scanner > might also be informative. > > Peace, > > Matt. > > From: Barbara Kreilkamp <bakk....@googlemail.com> > Date: Sunday, May 31, 2015 at 3:01 PM > To: Matt Glasser <glass...@wusm.wustl.edu>, Gaurav Patel > <gauravpa...@gmail.com>, "hcp-users@humanconnectome.org" > <hcp-users@humanconnectome.org> > Subject: Re: [HCP-Users] myelin maps on GE scanner > > Thank you Matt for this expert advice. > I've checked FNIRT registration for this subject, and it seemed fine, to make > sure the high/low intensities in the myelin map were not generated through > poor registration between T1 FSPGR BRAVO and T2 CUBE. > So, which areas are we worried about in particular - the superior temporal > sulcus for example? The patterns in motor cortex and V1 seem to resemble your > images? > We had real problems with inhomogeneity on the 3T GE scanner, but we applied > the PURE intensity corrections and I computed a significant decrease of the > bias field, when comparing those images to T1s without PURE correction. > Yes, we used a head coil. > Do you think we'd have to bid farewell to a myelin mapping approach with this > data or would you say there is a work-around by doing ROI-analysis etc.? > > Thanks, > Best wishes, > Barbara > > > > On 31/05/2015 20:42, Glasser, Matthew wrote: >> Hi Barbara, >> >> Unfortunately those actually look much more like Gaurav’s than the ones in >> in Figure 3 of Glasser and Van Essen 2011 (panels B, E, H, K were from a GE >> scanner). The issue is that the sulci on the lateral surface are brighter >> than the gyri in a noticeable way relative to what we typically see, which >> is not something we saw in the NAMIC GE data or any of our Siemens data over >> the years. I’m quite puzzled by this issue, since it the myelin maps used >> to work on GE and give comparable results to Siemens. Also, I don’t have >> any experience with GE scanners to know what parameters might be creating >> the issue. >> >> Here is the info that was available for the GE NAMIC data (that was acquire >> some time ago now): >> >> The second dataset includes the 10 control subjects (all male, mean age 42 >> 11 years) from the publicly available Brain Multimodality dataset from the >> National Alliance for Medical Image Computing (NAMIC) (provided by the >> Psychiatry Neuroimaging Laboratory and the Surgical Planning Laboratory, >> Brigham and Women’s Hospital). The data were acquired on a 3T General >> Electric (GE) scanner at Brigham and Women’s Hospital in Boston using an >> 8-channel head coil and GE’s parallel imaging technology Array Spatial >> Sensitivity Encoding Techniques (ASSET) was used with a SENSE (SENSitivity >> Encoding) factor of 2. A T1w spoiled gradient recalled sequence (SPGR; TR >> 7.4 ms, TE 3 ms, TI600 ms, 10° flip angle, FOV 256mm256 mm, matrix 256 >> 256, 1 mm slices) and a T2w extended echo train acquisition (XETA; TR2500 >> ms, TE80 ms, FOV 256mm256 mm, matrix 256256, 1 mm slices) were acquired. >> Data were downloaded from the NAMIC MIDAS website: >> http://insight-journal.org/midas/collection/view/190. >> >> I’d actually be more inclined to suspect that the T2w sequence difference >> CUBE vs the XETA might be the cause of the issue. On Siemens we use a T2w >> SPACE sequence which has a lot of contrast for myelin in it (T2w being a bit >> of a misnomer, as the contrast in the SPACE is mostly T1 and MT), if the T2w >> scan were to have less contrast for myelin, that could reduce the quality of >> the results. Another possibility would be if the transmit field is >> substantially more inhomogeneous on this GE scanner (which is probably a >> different model than the one used for the NAMIC data). Is the transmit done >> from a body coil or a head coil for this GE scanner? In the Siemens >> scanners we use the more uniform transmit from the body coil, but if the >> head coil were used for transmit, it could produce a more inhomogeneous >> field. >> >> Peace, >> >> Matt. >> >> From: Barbara Kreilkamp <bakk....@googlemail.com> >> Date: Sunday, May 31, 2015 at 2:25 PM >> To: Gaurav Patel <gauravpa...@gmail.com>, "hcp-users@humanconnectome.org" >> <hcp-users@humanconnectome.org> >> Subject: Re: [HCP-Users] myelin maps on GE scanner >> >> Hi Gaurav, >> >> Could you please send over a screenshot of what the artifacts look like? >> We've acquired 3D FSPGR BRAVO with PURE correction on the GE MR750 Discovery >> 3T together with 3D CUBE with PURE correction (32 channels). >> The images that HCP computed from that for me look very much like the ones >> from Matt's publication, Neuroimage 2014 (single subject), scaling is 4% to >> 96% (attached). >> Please let me know if this helps or if I can help in any other way. >> >> Thanks, >> Best wishes, >> Barbara >> >> >> >> >> On 30/05/2015 17:53, Gaurav Patel wrote: >>> Another problem we have been having is artifacts in myelin maps generated >>> from the T1s and T2s acquired on our GE MR750D 3T magnet. Matt and I >>> looked at this in December, and found that in our maps there tended to be >>> an artifactual bias towards increasing T1/T2 ratio towards the center of >>> the brain, leading to apparent dense myelination depp in sulci and light >>> myelination at the tops of sulci that were not real. We've been trying to >>> get to the bottom of this problem since then with no luck, so I'm turning >>> to the list to see if anyone else has successfully created myelin maps on a >>> GE scanner and would be willing to share their sequences (or at least >>> parameters). We've traced the problem (we think) to the difference between >>> the FSPGR that GE scanners use and the MPRAGE, specifically the inversion >>> pulse that seems to make the FSPGR more sensitive to B1 receive field >>> inhomogeneity than the MPRAGE (and the T2 sequence, hence the bias in the >>> T1/T2 ratio). We' >>> ve compar >>> ed our seq >>> uence to the one used in the NAMIC dataset obtained on an older GE scanner >>> that Matt used in his original myelin mapping paper and haven't come up >>> with any real differences so far. Any help or insight would be >>> appreciated. Thanks! >>> >>> __________________________ >>> gaurav patel >>> >>> gauravpa...@gmail.com >>> >>> >>> www.neurofreak.net >>> >>> >>> >>> >>> >>> >>> _______________________________________________ >>> HCP-Users mailing list >>> >>> HCP-Users@humanconnectome.orghttp://lists.humanconnectome.org/mailman/listinfo/hcp-users >> >> _______________________________________________ >> HCP-Users mailing list >> HCP-Users@humanconnectome.org >> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >> >> The materials in this message are private and may contain Protected >> Healthcare Information or other information of a sensitive nature. If you >> are not the intended recipient, be advised that any unauthorized use, >> disclosure, copying or the taking of any action in reliance on the contents >> of this information is strictly prohibited. If you have received this email >> in error, please immediately notify the sender via telephone or return mail. > > > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you are > not the intended recipient, be advised that any unauthorized use, disclosure, > copying or the taking of any action in reliance on the contents of this > information is strictly prohibited. If you have received this email in error, > please immediately notify the sender via telephone or return mail. _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
