When not to mix results from FreeSurfer and FSL

Nyberg and colleagues (2015) report increased hippocampal volumes in PD (n = 21) versus non-PD peers (n = 20). Additionally, using FSL FIRST for a vertex analysis they found only right nucleus accumbens shape differences.

The article is interesting in part for the comparison of tremor dominant PD, postural instability dominant PD, and non-PD peers. It has, however, a flaw that’s worth writing about. I’m only going to point out this problem because it has relevance to structural MRI processing.

The authors report subcortical volumes from FreeSurfer 4.5. Version 4.5 is reliable but generally not recommended compared to newer versions, which have improved segmentation methods. FreeSurfer 5.3 was released on May 13, 2013, roughly two years before the publication of Nyberg et al. (2015). Reprocessing would have been trivial, requiring no more than 20 days (2 simultaneous processes) with around 10 days a more realistic estimate (10 hours per brain, running 2-3 subjects at a time). Typical FreeSurfer 5.3 processing times on a 2012 or newer iMac range from 6 to 8.5 hours per subject. Given the limited sample sizes (n = 21 and 20) and the difficulties FreeSurfer has with creating accurate hippocampus volumes (e.g., Lehmann et al. 2010), we are limited in making anything of the finding of increased hippocampal volumes in PD. Hippocampal hypertrophy in PD is possible but given the moderate stage of the participants, that’s a surprising finding at odds with the bulk of research to date (Brück et al, 2004; Camicioli et al., 2003; Laakso et al., 1996). Not all samples are the same or represent the same populations but because of the methodological shortcomings of Nyberg et al., the logical conclusion is that the results are spurious. Maybe they’re not but the evidence that it’s a real finding isn’t convincing.

The other problem with this is the authors use FreeSurfer for volumetric analyses but FSL FIRST for the shape (vertex) analyses. Why not just leave FreeSurfer out of it and report FIRST volumes? FIRST takes just minutes to run per brain and provides arguably superior results to FreeSurfer (knowing this I still typically use FreeSurfer volumes). Maybe the FIRST and FreeSurfer volumes were not different but that could at least be reported. It’s possible the authors used FreeSurfer because it offers an estimate of total intracranial volume (eTIV). This estimate (eTIV) has variable accuracy (own data and others: e.g., Malone et al., 2015). Superior methods exist for automatically estimating intracranial volume (own paper in progress and e.g., Malone et al., 2015). In this case FIRST volumes with head size estimated outside of FreeSurfer would have been preferable to a mixed FreeSurfer/FSL method.

I like and use FreeSurfer but a better practice in this case would have been sticking just with FSL’s tools. Intracranial volume can be accurately estimated using FAST (sum of gray, white, and CSF) because it includes much of the CSF outside the brain. Those results could be improved by using a combination of using FLIRT/FNIRT, BET, and FAST if you are willing to troubleshoot potential registration errors.

References

Brück, A., Kurki, T., Kaasinen, V., Vahlberg, T. and Rinne, J.O., 2004. Hippocampal and prefrontal atrophy in patients with early non-demented Parkinson’s disease is related to cognitive impairment. Journal of Neurology, Neurosurgery & Psychiatry75(10), pp.1467-1469.

Camicioli, R., Moore, M.M., Kinney, A., Corbridge, E., Glassberg, K. and Kaye, J.A., 2003. Parkinson’s disease is associated with hippocampal atrophy. Movement Disorders18(7), pp.784-790.

Laakso, M.P., Partanen, K., Riekkinen, P., Lehtovirta, M., Helkala, E.L., Hallikainen, M., Hanninen, T., Vainio, P. and Soininen, H., 1996. Hippocampal volumes in Alzheimer’s disease, Parkinson’s disease with and without dementia, and in vascular dementia An MRI study. Neurology46(3), pp.678-681.

Lehmann, M., Douiri, A., Kim, L.G., Modat, M., Chan, D., Ourselin, S., Barnes, J. and Fox, N.C., 2010. Atrophy patterns in Alzheimer’s disease and semantic dementia: a comparison of FreeSurfer and manual volumetric measurements. Neuroimage49(3), pp.2264-2274.

Malone, I.B., Leung, K.K., Clegg, S., Barnes, J., Whitwell, J.L., Ashburner, J., Fox, N.C. and Ridgway, G.R., 2015. Accurate automatic estimation of total intracranial volume: a nuisance variable with less nuisance. Neuroimage104, pp.366-372.

Nyberg, E.M., Tanabe, J., Honce, J.M., Krmpotich, T., Shelton, E., Hedeman, J. and Berman, B.D., 2015. Morphologic changes in the mesolimbic pathway in Parkinson’s disease motor subtypes. Parkinsonism & related disorders21(5), pp.536-540.

About Jared Tanner

I have a PhD in Clinical and Health Psychology with an emphasis in neuropsychology at the University of Florida. I previously studied at Brigham Young University. I am currently a Research Assistant Professor at the University of Florida. I spend the bulk of my research time dealing with structural magnetic resonance images of the brain. My specialty is with traditional structural MR images, such as T1-weighted and T2-weighted images, as well as diffusion weighted images. I also look at the cognitive and behavioral functioning of individuals with PD and older adults undergoing orthopedic surgery. Funding for the images came from NINDS K23NS060660 (awarded to Catherine Price, University of Florida). Brain images may not be used without my written permission (grant and software requirements).