This issue of what kind of data and how much is something we plan to investigate in detail for MSMAll (and the cortical areal classifier).
Matt. From: "Harms, Michael" <[email protected]> Date: Thursday, May 9, 2019 at 4:41 PM To: "Glasser, Matthew" <[email protected]>, Maria Sison <[email protected]>, Steve Smith <[email protected]> Cc: HCP 讨论组 <[email protected]> Subject: Re: [HCP-Users] MSMAll vs. MSMSulc reliability in our data While I’m not surprised that the ICCs would be lower for an anatomical-based measure for MSMAll than MSMSulc, I am surprised by the magnitude of the change (from 0.9 to 0.65), especially for a parcellated analysis, since only changes in the precise border of the parcellations should be affecting the results. Your results imply that the test and retest MSMAll registrations are very different from each other. Wouldn’t the Strain maps for MSMSulc vs MSMAll be informative here? You might also want to examine some sort of measure of the distortion between the two MSMAll registrations directly. Cheers, -MH -- Michael Harms, Ph.D. ----------------------------------------------------------- Associate Professor of Psychiatry Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO 63110 Email: [email protected] From: <[email protected]> on behalf of "Glasser, Matthew" <[email protected]> Date: Thursday, May 9, 2019 at 4:23 PM To: Maria Sison <[email protected]>, Stephen Smith <[email protected]> Cc: HCP 讨论组 <[email protected]> Subject: Re: [HCP-Users] MSMAll vs. MSMSulc reliability in our data Not running sICA+FIX might well be a part of the problem. The TR is quite long as Steve says, which will limit the accuracy of sICA+FIX cleanup some also. Also, surface area and thickness might prefer MSMSulc due to correlations with folding patterns. Myelin, task, and resting state fMRI will more tend to correlate with MSMAll. Matt. From: <[email protected]> on behalf of Maria Sison <[email protected]> Date: Thursday, May 9, 2019 at 3:40 PM To: Steve Smith <[email protected]> Cc: HCP 讨论组 <[email protected]> Subject: Re: [HCP-Users] MSMAll vs. MSMSulc reliability in our data Thank you so much, this is very helpful and interesting to think about. We concatenated rest and tasks and regressed out tasks to get around 1000 TRs of pseudo-rest which we then used for MSMAll. Still not nearly as much as HCP, but I would be interested to hear what a ballpark minimum data requirement for MSMALL would be. Best, Maria ________________________________ From: Steve Smith <[email protected]> Sent: Thursday, May 9, 2019 4:19:24 PM To: Maria Sison Cc: HCP 讨论组 Subject: Re: [HCP-Users] MSMAll vs. MSMSulc reliability in our data Hi - probably the single primary thing is number of timepoints - though things like TR and spatial resolution will also affect this. My guess is still that probably you don't have enough timepoints here to get decent single-subject RSN maps (decen enough for MSMALL that is). Emma or Matt might have more direct insight into the minimum amount of data you need to get MSMALL working well. Unless you can combine more of your datasets together (even if just for the purposes of MSM) then you might be better off with MSMSULC. Cheers. ps with this setup I would definitely push multiband at least as high as 6 if not 8. On 9 May 2019, at 15:12, Maria Sison <[email protected]<mailto:[email protected]>> wrote: Hello, Here’s our rfMRI protocol: each participant was scanned using a Siemens Skyra 3T scanner equipped with a 64-channel head/neck coil. A series of 72 interleaved axial T2-weighted functional slices were acquired using a 3-fold multi-band accelerated echo planar imaging sequence with the following parameters: TR = 2000 ms, TE = 27 msec, flip angle = 90°, field-of-view = 200 mm, voxel size = 2 mm isotropic, slice thickness = 2 mm without gap. Total scan length is 496 s. Out of curiosity, which parameters would be most important for MSMAll? Thank you, Maria ________________________________ From: Steve Smith <[email protected]<mailto:[email protected]>> Sent: Thursday, May 9, 2019 3:56:49 PM To: Maria Sison Cc: HCP 讨论组 Subject: Re: [HCP-Users] MSMAll vs. MSMSulc reliability in our data Hi - what is your rfMRI protocol? It might be that you're right that the difference is in the preprop - but my first guess might be that - if the rfMRI data is not as high quality as HCP rfMRI data - it might not be good enough to reliably drive MSMALL? Cheers. On 9 May 2019, at 14:45, Maria Sison <[email protected]<mailto:[email protected]>> wrote: Dear experts, We have run the HCP minimal preprocessing pipelines on our data (1 mm isotropic T1w and FLAIR + rest and 4 tasks) and compared test-retest reliability for MSMSulc and MSMAll in 20 subjects. Specifically, we looked at intraclass correlations for parcellated cortical thickness and surface area and found that they were much lower for MSMAll compared to MSMSulc in our test-retest sample (MSMSulc on average above 0.9 and for MSMAll around 0.65 on average). When we looked in HCP retest data, the ICCs for MSMAll were more similar to those for MSMSulc (both above 0.9), but still slightly lower. There are a few major differences in how we ran the pipeline. We skipped sICA+FIX and ran our own preprocessing on task and rest fMRI after fMRIVolume but before fMRISulc (bandpass filtering, motion correction, censoring, CompCorr, and regressed out tasks). We thought our processing would be ok for cleaning task fMRI, but I see that sICA+FIX is highly recommended before running MSMAll (https://www.mail-archive.com/[email protected]/msg06876.html<https://urldefense.proofpoint.com/v2/url?u=https-3A__www.mail-2Darchive.com_hcp-2Dusers-40humanconnectome.org_msg06876.html&d=DwMFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=jYoYbUrDg7XpYyosGa9JK-bzhv9_497jTGvkuIGUvC4&m=L2sucJXfwykaL3uBceFg9ItkjPLvmDK9C4reZWm3I3w&s=owCDH48LPtiA_trirzvrgF5FPS8Ba-Yz0KyPbwHJ6Mc&e=>), so I’m planning to try to rerun with sICA+FIX. Do you think that MSMAll is so dependent on sICA+FIX that it could be causing these problems in our data or do you have any other ideas about why we're getting such a large drop in ICCs for MSMAll? In other words, what are the minimal preprocessing requirements to effectively use MSMAll in non-HCP data? Any comments would be appreciated! Thank you, Maria _______________________________________________ HCP-Users mailing list [email protected]<mailto:[email protected]> http://lists.humanconnectome.org/mailman/listinfo/hcp-users<https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.humanconnectome.org_mailman_listinfo_hcp-2Dusers&d=DwMFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=jYoYbUrDg7XpYyosGa9JK-bzhv9_497jTGvkuIGUvC4&m=L2sucJXfwykaL3uBceFg9ItkjPLvmDK9C4reZWm3I3w&s=5J6HdZ4Fo_r3ENfbGzf9A5NOjDXu5D5DCwt8ZS8HkVg&e=> --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Head of Analysis, WIN (FMRIB) Oxford FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 610470 [email protected]<mailto:[email protected]> http://www.fmrib.ox.ac.uk/~steve<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.fmrib.ox.ac.uk_-7Esteve&d=DwMFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=jYoYbUrDg7XpYyosGa9JK-bzhv9_497jTGvkuIGUvC4&m=L2sucJXfwykaL3uBceFg9ItkjPLvmDK9C4reZWm3I3w&s=flsdWLpWURamBFUowb_0CPPOb_320ohP4cEZWX9OfmQ&e=> --------------------------------------------------------------------------- Stop the cultural destruction of Tibet<https://urldefense.proofpoint.com/v2/url?u=http-3A__smithinks.net&d=DwMFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=jYoYbUrDg7XpYyosGa9JK-bzhv9_497jTGvkuIGUvC4&m=L2sucJXfwykaL3uBceFg9ItkjPLvmDK9C4reZWm3I3w&s=YgiVdNVcsd8Q5XdU8ruCngYSzvnnxWMf0PaRuPdVRRI&e=> --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Head of Analysis, WIN (FMRIB) Oxford FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 610470 [email protected]<mailto:[email protected]> http://www.fmrib.ox.ac.uk/~steve<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.fmrib.ox.ac.uk_-7Esteve&d=DwMFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=jYoYbUrDg7XpYyosGa9JK-bzhv9_497jTGvkuIGUvC4&m=Um-g-12CMk5HZD2CpGPIwsIjPs-quCZ4nII9-pkJ0QQ&s=N3eSUhJbLj8I-zAqkRZr3NTpkhsenLBPF-9TWq7L-tc&e=> --------------------------------------------------------------------------- Stop the cultural destruction of Tibet<https://urldefense.proofpoint.com/v2/url?u=http-3A__smithinks.net&d=DwMFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=jYoYbUrDg7XpYyosGa9JK-bzhv9_497jTGvkuIGUvC4&m=Um-g-12CMk5HZD2CpGPIwsIjPs-quCZ4nII9-pkJ0QQ&s=OGvyl_gxKmsk4JjyQXTnBrWPX7Qm6bqgYHZ6BRA2IpY&e=> _______________________________________________ HCP-Users mailing list [email protected] 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. _______________________________________________ HCP-Users mailing list [email protected] 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. _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
