ASL-MRI abnormal perfusion patterns in early Parkinson’s

  

AIM

• To identify cerebral blood flow (CBF) perfusion patterns derived from MRI
• To use this potential early biomarker of disease to distinguish recently Parkinson's disease (PD) subjects from healthy controls

INTRODUCTION
Early diagnosis of Parkinson's disease is important for both prognosis and treatment planning. Radiotracer studies by others have identified a specific PD-related metabolic/perfusion pattern characterized by pallidothalamic, pontine, and motor cortex hyperperfusion, associated with reduced perfusion in the lateral premotor, rostral supplementary motor, posterior parietal areas, and prefrontal cortex.1,2,3 ASL is an MR perfusion method used to non-invasively and quantitatively measure cerebral blood flow.4 ASL is an exciting method with which to investigate perfusion abnormalities in PD.

METHODS
Participants
Twelve early, drug-naïve PD subjects (mean age: 63.0±14 years, 10 male, mean disease duration: 2.7± 2 years) and a group of eleven controls matched for mean age, gender, and education (mean age: 65.3±10 years, 9 male) completed neuropsychological tests of global cognitive status (Mini Mental State Examination (MMSE: 28.9±1) and Montreal Cognitive Assessment (MoCA: 26.9±2.4)), motor assessment (Hoehn & Yahr (H& Y: 1.8±0.8), UPDRS (Part 3, motor: 27.5±12) and MRI scans.

Data Acquisition
We used pseudo continuous ASL on a 3T GE HDx scanner to investigate CBF perfusion as a potential biomarker in PD (Figure 1). Whole-brain structural MRI data was also acquired. The MR data were preprocessed using statistical parametric mapping software SPM5 and custom Matlab scripts. The images were realigned to the anterior commissure, the ASL-derived CBF map coregistered to the structural image, then normalized to a probabilistic elderly brain atlas5 to more accurately represent the ageing subject population. We smoothed the images with an isotropic Gaussian kernel of 10mm.

Data Analysis
Principal component analysis (PCA) of the entire data set (PD and controls) resulted in a set of characteristic perfusion covariance patterns represented as principal component images. The expression of the first component in each individual was used to examine differences between the two groups including the effect size of this difference. ROC and correlation analyses with neuropsychological test scores were used to examine the utility and impact of this component.

Figure 1: The black and white images in the top row show sequential axial slices of a raw
perfusion scan acquired using ASL from a healthy control. Regions of increased flow
(brighter) correspond to cortical areas and basal ganglia. In the bottom row, the CBF map
was overlaid onto a high resolution T1 structural scan. Red indicates increased flow. Both
scale bars range from 0 (black or blue) to 70 (white or red) ml/100g tissue/min.

RESULTS
The identified covariance pattern (Figure 2) was characterized by relative hypoperfusion in PD versus controls at bilateral posterior parietal-occipital junction, extending anteriorly to include precentral and postcentral gyri, and superior/middle frontal cortex. Hypoperfusion also occurred in the posterior cingulate gyrus, precuneus, bilateral superior/middle temporal lobe, and bilateral thalamus. A t-test of the expression of the perfusion pattern in each individual shows that this measure successfully distinguished between PD and controls (t17 = 2.2, p = 0.04;effect size:Cohen's d = 0.93) (Figure 3a). ROC analysis revealed an area under the curve of 0.78, substantially better than the chance value of 0.5 (Figure 3b). There was no significant correlation
between the expression of the first component and any of the following measures: MoCA (r = 0.23, p = 0.3), MMSE (r = 0.21, p = 0.3), UPDRS (part 3; r = 0.31, p = 0.3), H&Y (r = 0.08, p = 0.8).

DISCUSSION
The specific, recognizable perfusion changes evident in our group of drug-naïve PD subjects provides a new finding that may be indicative of early disease status. However, the limited range of scores on cognition, reflecting non-advanced age and short disease duration, suggest the need to reassess this network in a larger sample of patients showing a wider spectrum of motor and nonmotor impairments. Radiotracer studies suggest regions of hyperperfusion along with areas of hypoperfusion in advanced PD; at this early stage, we identified only reductions in perfusion using ASL.

Figure 2 : Parkinson's disease-related hypoperfusion pattern (Green = maximum differences) as
identified by Principal Component Analysis of perfusion MRI acquired using arterial spin labeling,
displayed on the healthy elderly template.5 This spatial covariance pattern was identified from 12 early,
drug-naïve PD subjects and 11 matched controls. These decreases are most prominent in temporal,
parietal, and midline cortex; subcortically, bilateral changes occur in the posterior thalamus.

(a)

(b)

Figure 3: (a) In the dot-plot, each point indicates the expression of the first principal component in one
individual. Blue diamonds represent the values of the drug-naïve PD group and the red circles
indicate the Control group; the mean of each group is also displayed. The PD group exhibited a
significantly increased mean expression (t17 = 2.2; p=0.04) of this presumably PD-related perfusion
pattern with a large effect size (d=0.93). (b) Receiver Operating Characteristic (ROC) curve for this
principal component. The area under the curve (0.78) suggests good discrimination between
PD/Control with 72.7% sensitivity and 83.3% specificity.

CONCLUSIONS
This study demonstrated that drug-naïve PD patients differed significantly from controls in terms of an increased expression of a PD-specific perfusion network free from drug and treatment related influences. The large effect size indicated that the two groups constitute separate populations despite only slight cognitive and motor impairment. The ROC analysis showed the expression of the first component to be a capable disease classifier. The ASL-derived perfusion pattern offers a potential early biomarker in PD that may also have value in presymptomatic individuals at risk of developing PD.

ACKNOWLEDGMENTS
The authors wish to recognize funding from Canterbury Medical Research Foundation, Christchurch, New Zealand

REFERENCES
[1] Eckert, T., Tang, C. & Eidelberg, D. (2007), Assessment of the progression of Parkinson's disease: a metabolic network approach, The Lancet Neurology, 6 (10): 926-932. [2] Huang, C., Mattis, P., Tang, C., Perrine, K., Carbon, M., & Eidelberg, D. (2007), Metabolic brain networks associated with cognitive function in Parkinson's disease, NeuroImage, 34 (2): 714-723. [3] Ma, Y., Tang, C., Spetsieris, P. G., Dhawan, V., & Eidelberg, D. (2006), Abnormal metabolic network activity in Parkinson's disease: test-retest reproducibility, J Cereb Blood Flow Metab, 27 (3): 597-605. [4] Dai, W., Garcia, D., de Bazelaire, C., Alsop, D.C. (2008), Continuous flow-driven inversion for arterial spin labeling using pulsed radio frequency and gradient fields, Magnetic Resonance in Medicine, 60 (6): 1488-97. [5] Lemaitre, H., et al. (2005), Age- and sex-related effects on the neuroanatomy of healthy elderly, NeuroImage, 26: 900-911.

Tracy R. Melzer1,2, Richard Watts1,3, Michael R. MacAskill1,2, Ajit Shankaranarayanan4, David C. Alsop5, Ross Keenan6, Charlotte Graham1,2, Leslie Livingston1,2, John C. Dalrymple-Alford1,7, Tim J. Anderson1,2

1Van der Veer Institute for Parkinson's Disease and Brain Research, 2Department of Medicine, University of Otago, Christchurch, 3Department of Physics and Astronomy, University of Canterbury, Christchurch, 4GE Healthcare, Menlo Park, California, USA, 5Beth Israel Deaconess Medical Center, Boston, MA, USA, 6Christchurch Radiology Group, 7Department of Psychology, University of Canterbury, Christchurch, New Zealand