Article Text


Statistical parametric mapping with 18F-dopa PET shows bilaterally reduced striatal and nigral dopaminergic function in early Parkinson’s disease
  1. K Itoa,c,d,
  2. P K Morrisha,
  3. J S Rakshia,
  4. T Uemaa,
  5. J Ashburnerb,
  6. D L Baileya,
  7. K J Fristonb,
  8. D J Brooksa,b
  1. aMRC Cyclotron Unit, Hammersmith Hospital, London, UK, bInstitute of Neurology, Queen Square, London, UK, cDepartment of Radiology, Nagoya University School of Medicine, Nagoya, Japan, dDepartment of Biofunctional Research, National Institute for Longevity Sciences, Obu, Japan
  1. Dr Kengo Ito, Department of Biofunctional Research, National Institute for Longevity Sciences, 36–3 Gengo Morioka-cho, Obu, Aichi Prefecture, 474–8522 Japan. Telephone 0081 562 46 2311, ext 843 or 836; fax 0081 562 44 6596; email kito{at}


OBJECTIVE To apply statistical parametric mapping to 18F-dopa PET data sets, to examine the regional distribution of changes in dopaminergic metabolism in early asymmetric Parkinson’s disease.

METHODS Thirteen normal volunteers (age 57.7 (SD 16.5) years; four women, nine men ) and six patients (age 50.3 (SD 13.5) years; three women, three men) with asymmetric (right sided) Parkinson’s disease were studied. Images from each dynamic dopa PET dataset were aligned and parametric images of18F-dopa influx (Ki) were created for each subject. The Ki images were transformed into standard stereotactic space. The Ki values of the caudate and putamen on spatially normalised images were compared with the Ki values before normalisation. The application of statistical parametric mapping (SPM) allowed statistical comparison of regional Ki values on a voxel by voxel basis between healthy volunteers and patients with Parkinson’s disease.

RESULTS There was a strong correlation between the Ki values before and after spatial normalisation (r=0.898, p=0.0001). Significant decreases in the Ki values were found for the Parkinson’s desease group throughout the entire left putamen (p< 0.001) and focally in the dorsal right putamen (p< 0.001). Decreased Ki values were also shown bilaterally in the substantia nigra (p< 0.01).

CONCLUSION Using (SPM) and 18F-dopa PET, reductions in both striatal and nigral brain dopaminergic function could be demonstrated in early Parkinson’s disease.

  • 18F-dopa
  • positron emission tomography
  • Parkinson’s disease
  • statistical parametric mapping

Statistics from

18F-Dopa positron emission tomography (PET) permits presynaptic dopaminergic terminal function to be studied in vivo in both healthy volunteers and in patients with Parkinson’s disease. In Parkinson’s disease this technique has been used to provide a measure of severity of disease and to identify those with early, and even preclinical, disease.1-4 The technique has also provided a method of monitoring the survival of fetal mesencephalic cells after implantation.5 6

Numerous analytical methods have been proposed to quantify biochemical function from 18F-dopa PET images. The most often used method, the multiple time graphical approach (MTGA), provides rate constants (Ki) for the storage of 18F-dopa within regions of interest (ROI) placed over the striatum.7 8 In most studies such ROIs have been placed manually by visual inspection of an aggregate image with reference to a stereotactic atlas of the brain. This approach has proved useful for the clinical evaluation of Parkinson’s disease, but has significant drawbacks.9 10The positioning of ROIs is observer dependent and may be prone to bias. It may also be difficult to place ROIs accurately in severe Parkinson’s disease where the decreased striatal 18F-dopa uptake means that the aggregate image provides an inaccurate representation of striatal anatomy.10 Additionally, there may be involvement of extrastriatal dopamine terminals to varying extents in different subgroups of idiopathic Parkinson’s disease (tremulous, akinetic-rigid) and other parkinsonian syndromes. Accurate detection of changes in extrastriatal areas, where uptake constants are lower than those in the striatum, is unreliable using conventional ROI analysis and a judgement must be made about the sites of loss of dopaminergic function before analysis. As a consequence the ROI approach is not really suited for use as an exploratory technique with18F-dopa PET.

Statistical parametric mapping is an approach developed to localise statistically significant changes in spatially normalised images on a voxel by voxel basis, without previous assumption of the distribution of differences between the images. In theory it should be applicable to18F-dopa PET data sets allowing assessment of dopaminergic function throughout the whole brain. Using a labour intensive MRI to PET coregistration approach we have previously showed that there are topographical patterns of striatal dopaminergic loss that may be more predictive of impending Parkinson’s disease than whole striatal or putamen Ki values alone.11 In particular, we were able to identify dorsal to ventral and caudal to rostral gradients in putamen Ki similar to those shown for postmortem dopamine concentrations in early disease.12 We were interested in studying whether the statistical parametric mapping (SPM) approach would detect a similar striatal gradient of abnormality in Parkinson’s disease and also localise extrastriatal regions of impaired dopaminergic function.

Subjects and methods


Six patients (age 50.3 (SD 13.5) years; three women, three men) with early Parkinson’s disease were recruited from the Hammersmith Hospital Movement Disorder clinic. The United Kingdom brain bank criteria for prospective diagnosis of Parkinson’s disease13 were fulfilled by all patients. Each patient was assessed clinically by a single observer (PKM) before PET at least 12 hours after stopping medication. Their illness was clinically staged with the unified Parkinson’s disease rating scale (UPDRS),14 and the Hoehn and Yahr15 rating scale. Four patients had pure right hemiparkinsonism and two had asymmetric right predominant parkinsonism. The details of patients are shown in table 1. A separate group of 13 normal volunteers (age 57.7 (SD 16.5) years; four women, nine men) were recruited and scanned over the same time period. All subjects gave written informed consent before PET. Permission to perform these studies was obtained from the sthics committee of the Hammersmith Hospital, London, United Kingdom, and from the Administration of Radioactive Substances Advisory Committee (ARSAC), United Kingdom.

Table 1

Patients with Parkinson’s disease


All subjects were given 100 mg oral carbidopa 1 hour before and 50 mg 5 minutes before scanning. PET was performed using a CTI 931/08/12 tomograph (CTI, Knoxville, TN, USA) yielding 15 simultaneous planes with an axial full width half maximum resolution of 7 mm and an in plane resolution of 8.5 mm×8.5 mm. Correction for tissue attenuation of 511 KeV γ radiation was measured using an external Ge-68 ring.18F-Dopa (80–180 MBq) in normal saline solution was infused intravenously over 30 seconds. Scanning began at the start of the tracer infusion with a protocol of 25 time frames (4×1 min, 3×2 min, 3×3 min, 15×5 min) over 94 minutes.


Analysis of data was performed on SUN workstations (Sun Microsystems, Silicon Valley, CA, USA). To correct for head movements between scans, all images from each subject were aligned using Automated Image Registration (AIR) software.16 After realignment of images conventional region of interest (ROI) analysis was performed for each subject using in house software written in IDL image analysis software (Research Systems, Inc, Boulder, CO, USA) based on the MTGA approach of Patlak and Blasberg. We used the time frames from 25 to 94 minutes postinjection with occipital counts as the input function.17 Occipital ROIs were located on the occipital cortex of each of two planes, avoiding the midline and the ventricles. The ROIs were placed on two contiguous planes containing caudate and putamen, by image inspection with reference to the stereotactic atlas of Talairach and Tournoux.18 One circular ROI of 84.0 mm2 (diameter=10.3 mm) was positioned by inspection on each caudate nucleus. One elliptical ROI of 156.0 mm2 was placed on each putamen. 18F-dopa influx (Ki) was calculated for each of right and left caudate and putamen. The Ki values of the caudate and putamen were compared with the Ki values without realignment to test the effect of alignment in six normal subjects and six patients with Parkinson’s disease.

Parametric images of 18F-dopa influx constants (Ki) were created for each subject using the same software. The Ki images were transformed into a standard stereotactic space.19 The integrated images from 0 to 20 minutes postinjection was used to identify the transformation parameters. Stereotactic normalisation of PET images allows the comparison of scan data in identical voxels across different subjects. After normalisation, a gaussian filter of 10×10×8 mm (full width half maximum in the x, y, z planes, respectively) was applied to remove high frequency noise in the images and to accommodate differences in gyral and functional anatomy between subjects. The Ki values of the caudate and putamen on spatially normalised images were compared with the Ki values before normalisation in six normal subjects and six patients with Parkinson’s disease. The same set of ROIs, defined from the stereotactic atlas of Talairach and Tournoux, was applied to the normalised images from all subjects. Statistical differences in mean regional Ki values between all normal subjects and patients with Parkinson’s disease were characterised with SPM on a voxel by voxel basis. Appropriate contrasts were used to derive the between group (unpaired) tstatistic using the general linear model. The resulting set of voxelt values constitutes an SPM{t} that is used to make inferences about regionally specific changes in Ki.20 The SPM{t}s were transformed to the unit normal distribution (SPM{Z}). When regional effects were predicted in advance, a threshold of p<0.01 uncorrected—that is, Z>2.33 was accepted as significant (anatomically specified hypotheses). For these regions comprising the caudate, putamen, and substantia nigra, the p values associated with other regional effects were corrected for multiple dependent comparisons using the theory of gaussian fields (anatomically exploratory hypotheses).


Caudate and putamen Ki values derived from ROI analysis before and after image realignment with AIR software showed a close correlation (r=0.974, p=0.0001, fig 1). We confirmed by visual inspection that normalisation into Talaraich space worked successfully for all subjects and the ROI analysis showed a high correlation between the Ki values before and after spatial transformation (r=0.898, p=0.0001, fig 2). The SPM showed a significant decrease in mean Ki for the Parkinson’s disease group in voxels throughout the entire left putamen (p<0.001 uncorrected). In the right putamen voxels showing a significant difference between the Parkinson’s disease group and the normal group were confined to the dorsal posterior putamen (p<0.001 uncorrected). Voxels demonstrating a significantly reduced Ki value from the normal mean were also seen showing in the substantia nigra (p<0.01 uncorrected). In the caudate no significant decreases in Ki were seen. The corresponding SPMs and peak coordinates are shown in fig 3 and table 2, respectively. No significant increases in Ki were seen anywhere at a corrected level of 0.05.

Figure 1

Correlation between the caudate and putamen Ki values before and after realignment for six normal subjects and six patients with Parkinson’s disease.

Figure 2

Correlation between individual Ki values before and after spatial normalisation for six normal subjects and six patients with Parkinson’s disease.

Figure 3

(A) SPMs showing the spatial distribution of significant regions (p<0.01 uncorrected—that is, Z > 2.33) where Ki decreased in Parkinson’s disease. Images are shown as integrated projections along sagittal, coronal, and transverse views of the Tarairach and Touroux brain atlas. (B) Significant regions (p<0.01) of subnormal dopamine metabolism (Ki) in hemiparkinsonism, superimposed on a normalised MRI of the brain. Statistical parametric mapping of18F-dopa PET images shows regional change in the striatum and substantia nigra in Parkinson’s disease.

Table 2

Peak coordinates in SPM analysis


We have shown that SPM can be successfully applied to18F-dopa PET images.

In this study we found that a group of early hemiparkinsonian patients showed significantly reduced F-dopa uptake in voxels distributed over the entire putamen contralateral to the most symptomatic limb, and confined to the dorsal caudal putamen in the putamen contralateral to the asymptomatic or least symptomatic limb. These findings compare favourably with our previous study using MRI to PET coregistration,12 in which the greatest loss of dopaminergic function was found in the dorsal caudal putamen. In that study we suggested that in early Parkinson’s disease the ventral and rostral putamen may have Ki values indistinguishable from normal, suggesting that the dopaminergic loss in Parkinson’s disease is initially a focal process targeting specific areas within the striatum. These PET findings are also in line with postmortem studies which have shown greatest dopamine losses from the dorsal posterior putamen in Parkinson’s disease.21 The present study confirms this finding and reinforces our view that the clinical presentation and progression of Parkinson’s disease may be explained largely by progression of the dopaminergic deficit throughout the striatum. By contrast with our work with MRI coregistration,12 we did not identify significant abnormality in the caudate in this study. This may be a reflection of the different technique, or a different patient group. Application of SPM also allowed us to examine loss of dopaminergic function in regions outside the striatum. The substantia nigra of our patients with Parkinson’s disease also showed a reduction in dopaminergic metabolism, confirming our previous report with conventional ROI analysis.22 We think that the application of SPM to Parkinson’s disease may in the future enable us to objectively investigate the progression of Parkinson’s disease in the whole brain in detail. By applying a statistical analysis to compare people suspected to have preclinical Parkinson’s disease with a large group of normal individuals identification of such people should also become more sensitive and specific. In the future it should also be possible to use SPM to compare regional striata and extrastriatal differences in dopaminergic metabolism between idiopathic Parkinson’s disease and atypical parkinsonian syndromes, between tremulous and akinetic rigid Parkinson’s disease, and between mild and severe Parkinson’s disease.

In this provisional study we had relatively few subjects. The ensuing low degrees of freedom reduce the overall power or sensitivity of our analysis. This means that, although we can be confident that the differences shown to be significant do in fact exist, we cannot infer that there are no other differences just because we fail to show them. It should be noted that it is generally the case that the null hypothesis cannot be accepted just because of failure to reject it. However, with greater degrees of freedom (greater numbers of subjects) the power of our analyses would increase and the probability of missing true differences would decrease. Having said this our results suggest that the present study was sufficiently sensitive to identify differences in regional tracer distribution and we can assert that, for example, the contralateral putamen was more significantly affected than the ipsilateral putamen.

There are other advantages in the application of SPM. Sources of error in conventional ROI analyses of 18F-dopa PET include variance in head positioning and ROI placement. Realignment of images and stereotactic normalisation counter errors arising from head positioning. ROI analysis using anatomically normalised images reduces error arising from ROI positioning, particularly in those in which the18F-dopa metabolism is severely impaired. To determine that there was a close correlation between Ki values obtained with ROI analysis before and after image realignment and then spatial normalisation we combined the data for normal subjects and patients into one plot to disclose the overall tendency. It could be argued that correlation coefficients obtained from combined patient and control data are invalid and we would accept this point. It is still clear, however, that the transformation process led to little change in Ki values derived by ROI placement. The image alignment program was originally developed for 15O-H2O activation studies and so the reliability of this alignment program for other tracers, especially for a data set of dynamic images, remains unproved. In this study there was a good correlation between the caudate and putamen Ki values before and after realignment suggesting there were no serious problems in the process of realignment, but further evaluation is necessary to confirm the reliability of applying alignment programs to 18F-dopa PET data sets. During anatomical normalisation cerebral blood flow images are conventionally used as a normalised brain template. Images of 18F-dopa PET are very different from rCBF template images but an add image of the early time frames (0–20 minutes) reflects rCBF and can be used as an anatomical template. Then normalised 18F-dopa Ki images can be generated using the parametric information of transformation obtained from the normalisation of the early add image. The Ki values obtained from the images anatomically normalised in this way closely correlated wth those derived before normalisation. In a previous study we used PET to MRI coregistration to evaluate regional changes in striatal18F-dopa metabolism in Parkinson’s disease. A great advantage of the method that we have shown here is that each patient need only undergo PET to derive essentially similar information. In this study we have simplified the analysis further by using occipital counts as an input function in the generation of parametric images. This approach avoids the need for arterial plasma sampling and is likely to be as reliable as using metabolite corrected plasma counts as an input function.23


The technique of spatial normalisation allows regions of interest (ROIs) to be applied to identical anatomical regions of18F-dopa PET in different people. The approach of SPM to localise significant changes in mean Ki values between groups of normal subjects and patients with early Parkinson’s disease has enabled us to objectively demonstrate reductions in regional dopa metabolism in the patients, without a priori knowledge. In this study we showed significant reduction in dopaminergic function throughout the entire putamen contralateral to the most symptomatic limb whereas the reduction was confined to the dorsal caudal putamen contralateral to the less affected or asymptomatic limb. Although further studies will be necessary to confirm this finding, it is in agreement with the concept that clinical progression of Parkinson’s disease is associated with a loss of dopaminergic metabolism beginning in the dorsocaudal putamen and spreading ventrally. Additionally, this study was able to show a significant reduction in nigral 18F-dopa uptake in early Parkinson’s disease.


We thank Andrew Blyth, Andreana Williams, and Graham Lewington for their help with scanning.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.