Diffusion tensor imaging of sequential neuropathological patterns in progressive supranuclear palsy

Abstract

Background and objective: A neuropathological cerebral staging concept for progressive supranuclear palsy (PSP) has been proposed that tau inclusions in PSP may progress in a sequential regional pattern. The objective was to develop a hypothesis-guided region/tract of interest-based (ROI/TOI) approach to use diffusion tensor imaging (DTI) targeted to analyze in vivo the regions that are prone to be involved at each neuropathological stage of PSP.

Methods: Two data cohorts were analyzed: cohort A of 78 PSP patients [55 Richardson's syndrome (PSP-RS) and 23 PSP with predominant parkinsonism (PSP-P)] and 63 controls, recorded at 3.0T at multiple sites, and a single-site cohort B constituted by 1.5T data of 66 PSP patients (46 PSP-RS and 20 PSP-P) and 44 controls. In cohort A, 21 PSP patients (13 PSP-RS and 8 PSP-P) and 17 controls obtained a follow-up scan after 17 months. Whole brain-based spatial statistics (WBSS) was used to identify the alterations in PSP patients vs. controls. The combined ROI- and TOI-based approach targeted structures that are prone to be involved during the course of PSP.

Results: WBSS demonstrated alterations predominantly in brainstem/midbrain, basal ganglia, and frontal lobe, more pronounced in the longitudinal data. Statistical analyses of the ROIs/TOIs showed a sequential pattern of structures that were assigned to previously defined neuropathological steps.

Conclusion: The combined ROI- and TOI-based DTI approach was able to map the disease stages of PSP in vivo cross-sectionally and longitudinally, lending support to DTI as a technical marker for imaging disease progression according to PSP stages. This approach might be useful as a tool for stratification of PSP patients MRI with respect to its proposed neuropathological progression in future longitudinal and autopsy-controlled studies.

Keywords: diffusion tensor imaging (DTI); neuropathology; progressive supranuclear palsy; sequential pattern; tau protein.

FIGURE 1

Regions-of-interest (ROIs) and tracts-of-interest (TOIs) selected for PSP patterns 1-3 based on previous findings in imaging and neuropathological studies. Sagittal projectional ROI (left panel) and TOI (right panel) representations for the three patterns corresponding to sequential involvement of brain structures during the course of PSP.

FIGURE 2

Scheme for data analysis and categorization of PSP patients. Fractional anisotropy (FA) maps were calculated from DTI data, i.e., 36 parameters were calculated for each subject from combined region-of-interest (ROI) and tract-of-interest (TOI) analyses (see Supplementary information II). Parameters were sorted into 3 patterns and for each pattern (after “z-normalization”) a threshold-based decision of affectation (1 or 0) was performed. The combinations of pattern-specific decisions [the individual imaging-based categorization step (CS)] lead to a categorization into CSα, CSβ, and CSχ.

FIGURE 3

Whole brain-based spatial statistics (WBSS) of fractional anisotropy (FA) maps. (a) Exemplary slicewise representation of WBSS of FA maps of 78 PSP patients vs. 63 controls at baseline (cohort A) and 66 PSP patients vs. 44 controls at baseline (cohort B). A decrease of FA in PSP patients vs. controls is displayed in hot colors, an increase in cold colors. (b) Differences at the group level (illustrated by variation of thresholded t-values) in projectional sagittal views for cohorts A and B. (c) Subdivision of WBSS into the subtypes PSP with Richardson’s syndrome (PSP-RS) and PSP with predominant parkinsonism (PSP-P). MNI, Montreal Neurological Institute coordinate frame.

FIGURE 4

Whole brain-based spatial statistics (WBSS). (a) Slicewise representation of WBSS (p < 0.05, corrected for multiple comparisons) of FA maps of 21 PSP patients vs. 63 controls at baseline and at follow-up. (b) Results at a thresholded amplitude of difference (t-values) in projectional sagittal views for baseline and at follow-up. MNI, Montreal Neurological Institute coordinate frame.

FIGURE 5

Pattern analysis results [z (< FA >)] for cohort A for baseline and follow-up (N = 21 PSP-patients) vs. controls (N = 63). The transparent red interval indicates an error estimation obtained from baseline and follow-up controls’ scans. **p < 0.0001.