No increase in deep-brain grey matter magnetic susceptibility observed over the Parkinson's disease

234 

Abstract Submission 

George Thomas¹, Naomi Hannaway¹, Angeliki Zarkali¹, Karin Shmueli², Rimona Weil^1,3,4

¹Dementia Research Centre, UCL, London, UK, ²Department of Medical Physics and Biomedical Engineering, UCL, London, UK, ³Wellcome Centre for Human Neuroimaging, UCL, London, UK, ⁴Movement Disorders Consortium, UCL, London, UK

George Thomas, Dr  
Dementia Research Centre, UCL
London, UK

Naomi Hannaway  
Dementia Research Centre, UCL
London, UK

Angeliki Zarkali  
Dementia Research Centre, UCL
London, UK

Karin Shmueli, Prof  
Department of Medical Physics and Biomedical Engineering, UCL
London, UK

Rimona Weil  
Dementia Research Centre, UCL|Wellcome Centre for Human Neuroimaging, UCL|Movement Disorders Consortium, UCL
London, UK|London, UK|London, UK

Magnetic susceptibility measured using quantitative susceptibility mapping (QSM) has previously been shown to be sensitive in detecting disease related changes in Parkinson's disease (PD). However, whether QSM can be used to track disease progression in PD is not known. Here, we present a 3-year longitudinal study of voxel-wise magnetic susceptibility in PD.

59 PD participants within 10 years of diagnosis were recruited from October 2017 to December 2018. All subjects were seen again after an average interval of 38.5±4.4 months (mean±SD). Imaging at both timepoints comprised single echo susceptibility-weighted spoiled GRE scans and anatomical MPRAGE scans.

For QSM pre-processing, phase images were unwrapped using a rapid path-based based minimum spanning tree algorithm [Dymerska et al., 2021] and brain masks calculated using BET2. Background field removal was completed with Laplacian boundary value extraction [Zhou et al., 2014] and 3D polynomial residual fitting. Susceptibility maps were estimated using Multi-Scale Dipole Inversion [Acosta-Cabronero et al., 2018]. A study-wise template was created from native space T1 images across both timepoints using a previously optimised routine [Acosta-Cabronero et al., 2017]. QSM images were transformed into this space.

For voxel-wise whole brain analyses, standardised images were spatially smoothed using a 3mm Gaussian kernel, requiring the use of absolute QSM to improve statistical conditioning [Betts et al., 2016]. To investigate changes in susceptibility between visits, single-group paired t-tests in the form of permutation analyses (adjusted for age, sex, and time between scans) were performed using randomise and threshold-free cluster enhancement in FSL. Significant clusters were inferred from 10,000 permutations and reported at family-wise error (FWE)-corrected P<0.05.

ROI analyses using both absolute and signed QSM were carried out to probe the relative contribution of diamagnetic and paramagnetic susceptibility sources to the interactions observed throughout the brain, and to further investigate regions commonly implicated in PD. The following ROIs were segmented from the anatomical template: substantia nigra pars compacta (SNpc) and pars reticulata (SNpr), dentate nucleus, red nucleus, caudate nucleus, putamen, globus pallidus, insular cortex, pars opercularis, middle temporal gyrus, precentral gyrus, and rostral middle frontal cortex. In R, linear mixed models were fitted at each ROI to investigate the effect of follow-up time on magnetic susceptibility (adjusted for age at baseline and sex). ANOVAs determined test-statistics for each model, p-values were FDR adjusted across the 12 ROIs.

Voxel-wise analysis revealed increased absolute magnetic susceptibility at follow-up relative to baseline in the left precentral gyrus, left middle frontal cortex and right middle temporal gyrus in PD (P_FWE<0.05, Fig 1). Post-hoc ROI analyses investigating both signed and absolute susceptibility corroborated the pattern seen at whole brain, with no significant increases observed in the iron-rich deep brain nuclei. Moderate increases in absolute susceptibility were seen in the pars opercularis, middle temporal gyrus, precentral gyrus, and rostral middle frontal cortex (P_FDR<0.05, Fig 2). However, no such relationships were observed for signed susceptibility, suggesting these changes were not driven by gross increases in cortical iron.

We present the first voxel-wise longitudinal study of magnetic susceptibility in PD. We report no increases in magnetic susceptibility over a 3-year period of the iron-rich deep brain nuclei commonly associated with PD. We find sparse changes in cortical magnetic susceptibility over time that are unlikely to be driven by increases in iron. In future, sequences sensitive to other tissue measures, such as multiparameter maps, or amyloid PET-CT, could be used to enrich our interpretation of magnetic susceptibility changes in PD.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Imaging Methods Other ²

Other - Parkinson's disease; Quantitative susceptibility mapping; longitudinal analysis