J Neurol DOI 10.1007/s00415-011-5915-2
ORIGINAL COMMUNICATION
Diffusion tensor imaging and olfactory identification testing in early-stage Parkinson’s disease Tyler M. Rolheiser • Heather G. Fulton • Kimberley P. Good • John D. Fisk • J. Roger McKelvey • Christophe Scherfler • Naeem M. Khan • Ronald A. Leslie Harold A. Robertson
•
Received: 15 September 2010 / Revised: 11 January 2011 / Accepted: 11 January 2011 Ó Springer-Verlag 2011
Abstract Evidence from imaging, clinical studies, and pathology suggests that Parkinson’s disease is preceded by a prodromal stage that predates clinical diagnosis by several years but there is no established method for detecting this stage. Olfactory impairment, which is common in Parkinson’s disease and often predates clinical diagnosis, may be a useful biomarker for early Parkinson’s. Evidence is emerging that diffusion imaging parameters might be altered in olfactory tract and substantia nigra in the early stages of clinical Parkinson’s disease, possibly reflecting pathological changes. However, no study has examined olfaction and diffusion imaging in olfactory tract and substantia nigra in the same group of patients. The present study compared newly diagnosed Parkinson’s disease patients with a matched control group using both olfactory testing and diffusion tensor imaging of the substantia nigra and anterior olfactory structures. Fourteen patients with stage 1–2 Hoehn & Yahr Parkinson’s disease were matched to a control group by age
and sex. All subjects then completed the University of Pennsylvania Smell Identification Test, as well as a series of MRI scans designed to examine diffusion characteristics of the olfactory tract and the substantia nigra. Olfactory testing revealed significant impairment in the patient group. Diffusion tensor imaging revealed significant group differences in both the substantia nigra and anterior olfactory region, with fractional anisotropy of the olfactory region clearly distinguishing the Parkinson’s subjects from controls. This study suggests that there may be value in combining behavioral (olfaction) and MRI testing to identify early Parkinson’s disease. Since loss of olfaction often precedes the motor symptoms in Parkinson’s disease, the important question raised is ‘‘will the combination of olfactory testing and MRI (DTI) testing identify pre-motor Parkinson’s disease?’’
T. M. Rolheiser H. A. Robertson Department of Pharmacology, Dalhousie University, Halifax, NS, Canada
N. M. Khan Department of Diagnostic Imaging, Dalhousie University, Halifax, NS, Canada
H. G. Fulton Department of Psychology, Dalhousie University, Halifax, NS, Canada
C. Scherfler Department of Neurology, Innsbruck Medical University, Innsbruck, Austria
K. P. Good J. D. Fisk Department of Psychiatry, Dalhousie University, Halifax, NS, Canada
H. A. Robertson (&) Brain Repair Centre and Department of Pharmacology, Faculty of Medicine, Dalhousie University, 5850 College Street, Halifax, NS B3H 1X5, Canada e-mail:
[email protected]
R. A. Leslie Department of Anatomy and Neurobiology, Dalhousie University, Halifax, NS, Canada
Keywords Parkinson’s disease Diffusion tensor imaging Olfactory testing Substantia nigra
J. R. McKelvey Department of Medicine/Neurology, Dalhousie University, Halifax, NS, Canada
123
J Neurol
Introduction Interventions that halt or reverse the progression of Parkinson’s disease (PD) remain a major unmet medical need. Current therapies (e.g., drugs, cell transplantation, deep brain stimulation) are only applied after clinical diagnosis of PD, when damage may be irreparable. Testing of therapies directed at halting progression requires the development of new biomarkers that identify PD in the preclinical stages. It is clear that preclinical PD antedates diagnosis by 4–6 years and involves pathological changes to brain anterior olfactory structures [1–4]. The majority of patients with PD (about 90%) have olfactory deficits in their earliest clinical stages [5–8], and it has been suggested that subtle changes in olfactory function may occur a decade or more before diagnosis [9]. The relationship between olfactory deficits and PD is so close that it has been argued that an absence of olfactory dysfunction is a good reason to question the diagnosis of PD [10]. Studies have suggested that olfactory testing, in the presence of other indicators of risk, might identify persons at risk for developing clinical signs of PD [11–13]. Diffusion weighted magnetic resonance imaging, which provides a measure of the integrity of neural tissue, has confirmed a disruption of diffusion parameters in the orbitofrontal region of the cerebral cortex in early stage PD [14], a finding congruent with early neuropathological changes in anterior olfactory structures [4]. Further, recent studies have identified changes in diffusion parameters in the substantia nigra region in people with early stage PD [15–17]. Such diffusion changes in olfactory structures and substantia nigra may be measurable before motor symptoms emerge. Coupled with a behavioral marker of the disease, these measures could provide a means of identifying a person at risk for developing PD. Despite the fact that the olfactory tracts are located in a region with susceptibility artifacts, recent studies have demonstrated that fiber tracking of the olfactory tracts is feasible using DTI [18]. While each of these potential biomarkers of PD has been examined independently, they have not previously been examined concurrently. In the present study, behavioral olfactory testing, as well as diffusion tensor imaging (DTI) of the anterior olfactory structures (AOS) and the substantia nigra (SN) were examined and compared in a cohort of early stage PD patients and a matched control group.
Methods: subjects Fourteen PD patients at Hoehn & Yahr stage 1 or 2 were recruited from the Movement Disorders Clinic of the Capital District Health Authority (CDHA), Halifax, NS,
123
Canada. Diagnosis was made by a board-certified movement disorders specialist, using the United Kingdom Parkinson’s Disease Society Brain Bank Criteria as a guide [19]. All PD patients were between 50 and 64 years old (Table 1). All except two of the subjects were on drug treatment as follows: ten were on a MAOB inhibitor (seven selegiline, three rasagiline), nine were on levodopa, three were on a dopamine agonist (two ropinerole, one pramipexole). Fourteen healthy control subjects, matched to PD cases on the basis of sex (8 male, 6 female) and selected to provide a group matched in age (mean ± SD = 55.2 ± 6.2), were recruited from the general population via local advertisements. Exclusion criteria for PD patients included signs of dementia, autonomic dysfunction, vertical gaze palsy and/or cerebellar signs on clinical examination by a movement disorder clinic board-certified neurologist. The CDHA Research Ethics Board approved the study in accordance with the Declaration of Helsinki.
Olfactory data collection All participants were administered the University of Pennsylvania Smell Identification Test (UPSIT) on the day of their MRI scan [20, 21]. This standardized and reliable test of olfactory function consists of four booklets of microencapsulated odours. Participants were required to ‘‘scratch and sniff’’ each stimulus and pick one of the four response options that best represented the odour, regardless of whether they could smell an odour or not.
Table 1 Patient characteristics Subject
Age (years)
Sex
Disease durationa
Hoehn and Yahr
S01
62
M
24
2
S02
56
M
24
1.5
S03
54
F
4
1
S04
64
M
48
2
S05
58
M
36
1
S06
59
M
36
2
S07
51
F
72
1.5
S08
50
M
48
1
S09
52
F
24
1
S10
51
M
60
1
S11
62
F
1
1.5
S12
50
F
24
1
S13
56
M
18
1
S14
58
F
4
1
Mean SD
56 4.8
30 21.2
1.3 0.42
a
Months since diagnosis
J Neurol
MRI data collection and dependent measures MRIs were obtained using a GE 1.5-Tesla whole body magnet with an 8-channel phased-array head coil at the IWK Health Centre, Halifax, NS. MRI post-processing and analyses were performed using FSL version 4.1. All off-line statistical operations were performed using Graphpad Prism 5. Each MRI acquisition included a series of four MRI sequences that were obtained during a single scanning session of approximately 37 min duration. Three sequences generated structural images that were used for coregistration with the diffusion images. These included a 3-dimensional T1 SPGR sequence (TE: In phase; prep time: 500 ms; flip angle: 20o; matrix: 256 9 256; slice thickness: 1.5 mm), an axially collected T2 GRE sequence (TR/TE: 600/20 ms; flip angle: 30o; matrix: 256 9 256; slice thickness: 3 mm; signal averaging: 2), and a coronally acquired T2 GRE sequence (TR/TE: 4,750/102 ms; matrix: 256 9 256; slice thickness: 2 mm; signal averaging: 2). The fourth acquisition was a DTI scan from which diffusion variables were extracted. Key data acquisition parameters for this were TR/TE of 12,000/72 ms, b-values of 0 and 900 ms, 31 orthogonal sampling directions collected twice and averaged, FOV of 26 cm, matrix of 128 9 128, and a slice thickness of 3 mm. To improve the quality of the DTI scan, parallel imaging was used during the acquisition of the scans (ASSET). All MRI acquisitions were visually checked for artifacts before processing. Consistent with previously used methods [14, 17], several DTI variables were extracted using a region of interest (ROI) approach. For both the AOS and the SN, the following parameters were used for groupwise comparison: fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD).
MRI data analysis All MRI scans underwent a custom brain data extraction protocol. DTI data were further corrected for eddy current
distortions, as well as motion artifact, before diffusion parameters were calculated. Structural scans were coregistered to the DTI data using a 7 degrees of freedom model. ROI tracing was completed by a single evaluator (TR) and confirmed by a second evaluator (HR), both of whom were blinded to subjects’ behavioral results. ROI tracing of the SN region followed the methods described in detail previously [17]. Briefly, an axial T2-weighted image was used to locate the superior portion of the red nucleus. Once this location was identified, the ROI tracing was completed on the slice immediately inferior to this upper-most slice. A first eigenvector color map was superimposed on the highresolution scan to allow better visualization of the SN, as well as to prevent excessive partial volume effects from the adjacent tissues (see Fig. 1). The extraction of DTI variables from the AOS began by tracing masks of the olfactory nerves on the coronally acquired T2 structural images. The ROI was defined as starting at the anterior pole of the olfactory bulb, and continued until the olfactory nerve became encapsulated in the olfactory sulcus. These masks were then coregistered into a subject’s T1 space. The DTI data were also coregistered into T1 space, where data extraction occurred (see Fig. 2). All extracted DTI data from the AOS and SN ROIs were visually checked for outlier voxels, and inspected to ensure unimodal distribution of DTI values. A final visual check of the data was performed to ensure that the AOS masks were composed of voxels whose primary direction of diffusion occurred in the anterior-posterior plane (determined using a first eigenvector directional lines map). Values were then entered into the statistical program for comparison using independent sample t-tests. Statistics were adjusted for multiple comparisons using a Bonferroni correction. As a control to ensure that there were no systemic diffusion differences that might affect the ROI analysis, a between-group voxelwise comparison was conducted using Tract-Based Spatial Statistics (TBSS) [22]. To facilitate this, each subject’s FA map was aligned into a common
Fig. 1 Region of interest delineation for the substantia nigra. High resolution axially collected T2 image coregistered into diffusion space. Color inlay represents a superimposed V1 map that allowed for greater delineation of ROI tracings, approximated here using black outlines
123
J Neurol
Fig. 2 Region of interest delineation for the anterior olfactory structures. Extraction of DTI parameters of the anterior olfactory structures. a The structures were first defined using a coronal T2weighted image. b The masks were subsequently coregistered to a T1weighted image. c The diffusion weighted images were also
coregistered to the T1-weighted anatomical image, where the data extraction occurred. All data were checked to ensure that the ROI data were extracted from voxels whose primary direction of diffusion was anterior-posterior as defined by a V1 map (bright green section in the inlay)
space (FMRIB_58 1 mm FA map) using a non-linear registration tool (FNIRT) [23, 24]. A group FA skeleton was created representing the centers of all major white matter tracts common to the groups (e.g., the corpus callosum, cerebrospinal tract, superior longitudinal fasciculus, optic radiation). Each subject’s FA values, therefore, were projected onto the group skeleton to allow for statistical voxelwise comparisons. The analysis, however, did not compare FA values of gray matter or cranial nerves.
Results
Fig. 3 Scatter plot of UPSIT scores. *Horizontal lines represent mean and standard deviation
Each subject enrolled in the study completed all testing: no data were removed from the analysis. The groups did not differ in age (control mean 55.2 years; PD mean 56 years) or sex (8 males and 6 females per group).
since diagnosis, sex, and history of smoking. None of these measures was able to account for any significant amount of variability in UPSIT test scores.
UPSIT
MRI: substantia nigra ROI
Results of the UPSIT scores revealed a significant main effect of group [t (26) = 4.83, p \ 0.0001]. The group mean for PD subjects was within the severe microsmia range (mean = 22.0, Fig. 3), while according to normative data, the group mean for control subjects fell into the mild microsmic range with a mean score of 33.7 [21]. A single outlier in the control group contributed to the relatively low group mean. Without this value the control group mean of 34.5 was within the normal range. The outlier was a nonsmoker without any past history of medical problems known to impact the ability to identify odors. As the PD group scores ranged from complete anosmia to mild microsmia, a regression analysis was performed to probe for relations between the UPSIT score and age, time
Results of the SN ROI analyses using independent sample t-tests revealed several large between-group effects (Table 2) but with Bonferroni correction for multiple comparisons, only FA and RD differed between groups. Regardless, within the SN, water diffusion was less constrained in the PD patients, with greater diffusion along the second and third axes of the diffusion ellipsoid.
123
MRI: anterior olfactory structure ROI Results of the AOS ROI analyses using independent samples t-tests also revealed a number of between-group differences, though only FA differed between PD patients and
J Neurol Table 2 Substantia nigra region of interest statistics Control mean (SD) FA
0.450 (0.0270)
MD
0.000793 (0.000379)
AD RD Voxels in ROI
0.00120 (0.0000581)
PD subject mean (SD) p 0.420 (0.0275)
0.0010
0.000840 (0.0000480) 0.0085 0.00124 (0.0000806) 0.1820
0.000591 (0.0000294) 0.000642 (0.0000432) 0.0012 65 (12)
69 (16)
0.421
* Bold values denote significance after correction for familywise error rate
Table 3 Olfactory region of interest statistics Control mean (SD) PD subject mean (SD) p FA
0.207 (0.0079)
0.166 (0.0193)
