Parkinson's Disease Spatial Covariance Pattern: Noninvasive Quantification with Perfusion MRI

Abstract

Parkinson's disease (PD) is associated with elevated expression of a specific disease-related spatial covariance pattern (PDRP) in radiotracer scans of cerebral blood flow and metabolism. In this study, we scanned nine early-stage patients with PD and nine healthy controls using continuous arterial spin labeling (CASL) perfusion magnetic resonance imaging (pMRI). Parkinson's disease-related metabolic pattern expression in CASL pMRI scans was compared with the corresponding ¹⁸F-fluorodeoxyglucose positron emission tomography values. The PDRP expression was abnormally elevated (P<0.01) in patients scanned with either modality, and the two values were highly intercorrelated (P<0.0001). Perfusion MRI methods can be used for accurate quantification of disease-related covariance patterns.

Introduction

Spatial covariance analysis of ¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography (PET) images has been used extensively in the study of Parkinson's disease (PD) and related disorders (Post-on and Eidelberg, 2009). Idiopathic PD is associated with the expression of a specific disease-related covariance pattern (PDRP) characterized by pallidothalamic and pontine hypermetabolism, along with hypometabolism in the premotor cortex, supplementary motor area, and parietal association regions (Ma et al, 2007). PDRP scores increase linearly with disease progression (Huang et al, 2007) and are modulated by effective treatment of motor symptoms (Asanuma et al, 2006; Feigin et al, 2007). As cerebral blood flow (CBF) and glucose metabolism are tightly coupled in PD patients scanned off medication, PDRP expression can also be quantified in resting CBF scans acquired using H₂¹⁵O PET or ^99mTc-ECD (ethyl cysteinate dimer) SPECT (single photon emission computed tomography) imaging (Eckert et al, 2007; Ma et al, 2007).

Arterial spin labeling (ASL) perfusion-weighted magnetic resonance imaging (pMRI) is another possible method for the quantification of disease-related spatial patterns, because it uses magnetically labeled arterial blood water as an endogenous tracer and can be repeated multiple times within session to improve signal-to-noise ratio. In this regard, continuous ASL (CASL) is particularly useful because of its superior image quality in multislice acquisition (Floyd et al, 2003). Cerebral blood flow measurements with CASL pMRI are stable over intervals ranging from minutes to weeks (Parkes et al, 2004) and are used in functional studies of healthy aging. Recently, CASL pMRI has been used with spatial covariance analysis to identify and validate a significant regional pattern associated with Alzheimer's disease (Asllani et al, 2008). In this pilot study, we examined the feasibility of CASL pMRI for quantifying PDRP expression in patients and healthy control subjects. The resulting PDRP scores were compared with those obtained using FDG PET in the same subjects.

Materials and methods

Subjects

A total of 9 patients (6 men and 3 women; age 64.7 ± 9.2 years (mean ± s.d.)) with early stage PD (Hoehn and Yahr Stages 1 to 2; off-state UPDRS (Unified Parkinson's Disease Rating Scale) motor ratings 22.8 ± 8.8) underwent CASL pMRI. All patients exhibited a clear-cut response to levodopa and/or dopamine agonist medications ( > 20% change in motor UPDRS ratings). Overall, 9 healthy volunteers (5 men and 4 women; age 54.1 ± 5.4 years) served as controls. Patients with PD and 7 of the controls also underwent FDG PET imaging. Ethical permission for this study was obtained from the Institutional Review Board of the North Shore University Hospital. Written consent was obtained from each subject after a detailed explanation of the procedures.

Imaging Protocol

Anti-parkinsonian medications were withheld for at least 12 h before MRI/PET imaging. Perfusion MRI was performed using a GE Signa 3-T MRI scanner (General Electric, Waukesha, WI, USA) using the product head coil. The scanner had a max gradient strength of 40 mT/m and a max gradient slew rate of 150 T/m per sec. Subjects were instructed to lie still in the scanner; their ears were plugged. A reference proton-density-weighted anatomic image with the same slice placement and thickness was acquired immediately after ASL acquisition. A high-resolution T1-weighted anatomic image was also acquired using a spoiled gradient recalled sequence (repetition time/echo time = 30/8 msecs, flip angle = 45°, field of view = 24 cm, 150 contiguous coronal slices, thickness = 1.5mm, matrix = 256 × 256, in-plane resolution = 0.94 × 0.94 mm²).

Continuous ASL pMRI was performed at the cervicomedullary junction using velocity-driven adiabatic inversion with an amplitude-modulated control labeling pulse (Wang et al, 2005). This procedure permits volumetric imaging of the whole brain in the same session by reproducing the frequency-dependent off-resonance effects of the labeling pulse. The effects of CASL pMRI were measured using single-shot, gradient echo and echo-planar images (EPIs) with a field of view of 24 cm along the frequency-encoding direction and 18 cm along the phase direction. Image acquisition was performed continuously across 48 slices (matrix size = 64 × 48, in-plane resolution = 3.75 × 3.75 mm², slice thickness = 3.75 mm). Raw echo amplitudes from EPIs were saved on a workstation (Sun Microsystems, Mountainview, CA, USA) for reconstruction and processing using software written in Interactive Data Language (IDL) (Research Systems, Boulder, CO, USA). After correcting for noise and motion artifacts, quantitative CBF images were calculated as described previously (Alsop and Detre, 1998), and smoothed in-plane by using a Gaussian kernel with a FWHM (full-width at half-maximum) of 1.88 mm.

¹⁸F-fluorodeoxyglucose PET imaging was performed in three-dimensional mode using a GE Advance tomograph (General Electric) within 8 weeks of MRI. The FDG PET scans were conducted under conditions similar to those described above for ASL pMRI. A 10-min static frame was acquired in a quiet and dimly lit room starting 35 mins after radiotracer injection. Details of this procedure are described elsewhere (Ma et al, 2007). The PET images were reconstructed with a three-dimensional image resolution of 8-mm FWHM with a 6-mm Hanning filter (matrix = 128 × 128 × 35; voxel size = 2.43 × 2.43 × 4.25 mm³) after correcting for photon attenuation, scatter, randoms, and dead-time effects.

Data Processing

Continuous ASL pMRI and FDG PET images were processed separately using SPM software (Institute of Neurology, London, UK) running on Matlab 6.5 (Mathworks, Natick, MA, USA). The CASL pMRI scans were spatially normalized to a Talairach space by coregistration with in-plane proton-density images and high-resolution spoiled gradient recalled sequence scans. The normalized scans were smoothed using a Gaussian kernel at FWHM = 12 mm. The FDG PET scans were spatially normalized to a Talairach-based PET brain template and then smoothed using a Gaussian kernel at FWHM = 10 mm.

Parkinson's Disease-Related Metabolic Pattern Quantification

We computed PDRP scores for the individual CASL pMRI and FDG PET scans on a prospective single case basis using an automated blinded algorithm (Ma et al, 2007; Spetsieris et al, 2009) (software available in the Scaled Subprofile Model/Principal Component Analysis (SSM/PCA) toolbox; http://www.fil.ion.ucl.ac.uk/spm/ext/). These computations were performed blind to subject class (PD, controls) and imaging modality (CASL, PET) using a PDRP topography (Figure 1A), which was previously identified and validated with FDG PET (Ma et al, 2007). Raw PDRP scores were z-transformed on the basis of mean and s.d. of the whole sample. The resulting values for each imaging modality were offset so that the control mean was equal to zero.

Figure 1

(A) Parkinson's disease (PD)-related metabolic pattern (PDRP) obtained by spatial covariance analysis of FDG PET scans from a combined sample of 33 PD patients and 33 age-matched controls, unrelated to the participants in the current study cohort (Ma et al, 2007). This abnormal regional covariance pattern was characterized by pallidal, thalamic, pontine, and cerebellar hypermetabolism associated with metabolic decrements in the lateral premotor and posterior parietal areas. (Voxel weights of the pattern overlaid on a standard MRI brain template. The displayed voxels were shown to contribute significantly to the pattern (P = 0.001) and to be reliable (P < 0.001) by bootstrap estimation. Voxels with positive region weights (metabolic increases) are color coded red, whereas those with negative region weights (metabolic decreases) are color coded blue). (B) PDRP expression measured prospectively in CASL pMRI (left) and FDG PET (right) scans. Pattern expression (z-scored) measured with either imaging method was increased in PD patients relative to normal controls (NCs) (P < 0.01). (The larger open triangles represent the two healthy control subjects who underwent CASL pMRI but not FDG PET imaging. Error bars are mean ± s.d.). (C) Correlation between PDRP expression measured with CASL pMRI and FDG PET (see text). The correlation between these measures was significant (r = 0.86; P < 0.0001). (PD patients are denoted by filled circles and NCs by open circles).

Statistical Analysis

Differences in PDRP scores between groups (PD versus controls) were compared separately for the two imaging measures using two-sample Student's t-tests. Differences between the two imaging modalities (CASL versus PET) for each subject group were assessed using paired Student's t-tests. In addition, PDRP scores were correlated across modalities by computing Pearson's correlation coefficients for each group separately and for the combined cohort.

Owing to the significant age difference between the patient and control groups (P = 0.01), we repeated the same analyses on PDRP data after adding an ‘age-appropriate’ correction to the raw value for each modality. This age-correction was performed using a linear equation that related PDRP expression to subject age in an independent FDG PET data set obtained from 58 normal volunteers (age range: 20 to 80 years). All statistical analyses were conducted using SPSS version 9.0 on PCs (SPSS, Chicago, IL, USA) and were considered significant at P < 0.05.

Results

PDRP scores obtained from the PD and control groups are presented in Table 1 (top). PDRP expression in the PD group was abnormally elevated (P < 0.01) in both the CASL pMRI and the FDG PET data sets (Figure 1B). The mean PDRP expression did not differ for the two imaging modalities (P > 0.6). Nonetheless, on an individual basis, pattern scores for the two groups overlapped more in the CASL pMRI data than in the FDG PET data. PDRP scores measured with CASL pMRI correlated closely with those from FDG PET (r = 0.86, P < 0.0001; Figure 1C). This correlation reflected a significant association between CASL and PET measurements in the patient group (r = 0.82, P = 0.0006) but not in controls (r = 0.36, P = 0.43). There was no significant correlation between UPDRS motor ratings and PDRP scores measured using the two imaging techniques (CASL pMRI: r = 0.31, P = 0.41; FDG PET: r = 0.48, P = 0.19). Age-corrected PDRP scores for the two groups are presented in Table 1 (bottom). Group comparisons and between-modality correlations were similar to the findings without age correction.

Table 1

PDRP scores measured with ASL pMRI and FDG PET

Uncorrected	PD	NC
ASL pMRI	1.22 ± 0.34 a	0.00 ± 0.16*
FDG PET	1.20 ± 0.35	0.00 ± 0.10*
Correlation	0.82 b	0.36
Age corrected
ASL pMRI	1.20 ± 0.34	0.00 ± 0.17*
FDG PET	1.12 ± 0.36	0.00 ± 0.10**
Correlation	0.79	0.49

ASL, arterial spin labeling; FDG, ¹⁸F-fluorodeoxyglucose; NC, normal control; PD, Parkinson's disease; PDRP; PD-related metabolic pattern; PET, positron emission tomography; pMRI, perfusion magnetic resonance imaging.

*

P < 0.01

**

P < 0.02; Student's t-tests for PD > NC.

a

z-scored values: mean ± s.e. in PD and NCs.

b

Correlations between PDRP scores measured with ASL pMRI and FDG PET (P < 0.01). The values reflect Pearson's correlation coefficients.

Discussion

Spatial covariance analysis has been used in conjunction with radionuclide imaging of cerebral glucose metabolism and blood flow to identify and measure the expression of disease-related regional patterns in PD and other neurodegenerative disorders (Habeck et al, 2008; Poston and Eidelberg, 2009). In this study, we found that PDRP expression measured with CASL pMRI were comparable with values quantified by FDG PET in the same subjects. Indeed, PDRP scores obtained using CASL pMRI differentiated PD from controls with accuracy similar to that of FDG PET. This accords with previous PDRP findings in resting CBF scans acquired with H₂¹⁵O PET (Ma et al, 2007) and ^99mTc-ECD SPECT imaging (Eckert et al, 2007).

We found a significant correlation between PDRP scores computed using CASL pMRI and FDG PET methods. The magnitude of this correlation (R² ∼ 0.7) in the combined cohort of PD patients and healthy subjects was comparable with that between PDRP scores computed from H₂¹⁵O PET and FDG PET scans in the same subjects (Ma et al, 2007). These results are also consistent with correlations reported between regional CBF measures obtained with CASL pMRI and H₂¹⁵O PET scans in healthy subjects. Between-modality correlations were significant in PD patients but not in healthy volunteers. This is reasonable because the dispersion of PDRP scores in healthy subjects is typically smaller than that for PD patients (cf. Ma et al, 2007). Similarly, we found that UPDRS motor ratings did not correlate with PDRP scores computed using either imaging modality, due mostly to the relatively mild symptoms and limited range of clinical scores present in these patients. Finally, one cannot exclude the possibility that between-modality disparities in PDRP expression result from subtle differences in PET and MRI resting conditions. The precise influence of the rest state on PDRP quantification is a topic for future study.

We have previously observed that PDRP scores computed in the FDG PET scans of normal subjects exhibit a modest (R² ∼ 0.2), but significant, correlation with age (Moeller et al, 1996). Given the significant difference in age between patients and controls, we explicitly corrected for age in both CASL pMRI and FDG PET data. As expected, the age-related effects were subtle and did not influence the accuracy of group separation or the significance of the between-modality correlation in PDRP scores.

This study underscores the use of PDRP quantification with CASL pMRI as a potential screening tool and perhaps also for differential diagnosis of Parkinsonism (Spetsieris et al, 2009). Although relatively new, ASL pMRI can be easily implemented on conventional 1.5-T MRI platforms without additional hardware. The current method is based on gradient echo EPI, which generally has higher signal-to-noise than spin echo EPI, but is more susceptible to spatial artifacts from large vessels. In this regard, spin echo EPI may prove to be a better choice at 3 T. It is also important to note that dopaminergic therapy can lead to a significant dissociation of regional CBF and metabolism at critical PDRP nodes in PD. This results in a treatment-mediated increase in PDRP expression measured in CBF images, as opposed to the reduction observed with FDG PET (Hirano et al, 2008). Hence, caution may be appropriate in the interpretation of ASL-based pattern scores in medicated patients. Additional studies in a larger sample will be required to characterize and validate specific ASL-based PDRP topographies for prospective use with this imaging technique. Moreover, further refinement of CASL pMRI methodology may serve to reduce between-subject variability, thereby improving the accuracy of group separation with this approach.