X hits on this document

PDF document

Activation of the cerebellum in co-ordinated eye and hand tracking movements: an fMRI study - page 2 / 12





2 / 12


sor onto a target, appear to be particularly sensitive to cerebellar dysfunction as they provide challenges for co- ordination within the limb (i.e. between different limb segments), for co-ordination between the eye and hand and for the integration of visual feedback signals into on- going movement. The interaction between eye and hand is likely to be bi-directional. It is known that visual- tracking performance is improved if both eye and hand track the same target motion (Brown et al. 1993; Van Donkelaar 1997; Vercher and Gauthier 1988; Vercher et al. 1996), and it seems that an important con- tribution to control of the hand motion comes from the ocular-motor system (Blouin et al. 1996; Van Donkelaar et al. 1994; Van Donkelaar et al. 1996). Likewise, oculo- motion is more accurate, with higher smooth-pursuit gain and reduced lag, if the hand is also used (Abrams et al. 1990; Biguer et al. 1984; Koken and Erkelens 1992; Steinbach and Held 1968; Vercher et al. 1994). Thus, at some level, information about motion of either the eye or hand is available to both control systems. It is already known that the cerebellum is important for this interaction: Vercher and Gauthier (1988) demonstrated that the short-latency ocular following of a handcon- trolled cursor is lost after dentate lesions; Van Donkelaar and Lee (1994) showed that cerebellar patients were more disabled in combined hand and eye tasks than when performing without conjoint eye and hand move- ment.

In this paper, we report functional magnetic-resonance imaging experiments that address these two roles of the cerebellum, looking at the cerebellar activation during co- ordinated eye and hand movement in a tracking task. There have been several reports of cerebral activation in visually guided or visually driven movement (Culham et al. 1998; Ellerman et al. 1998; Flament et al. 1996; Grafton et al. 1992; Jueptner et al. 1996; Tamada et al. 1999; Turner et al. 1998), but only a few have concentrat- ed upon, or even fully imaged, the cerebellum. Hence, we have localised those areas of the human cerebellar cortex that are activated during movements of eyes or the hand, or both, to follow movement of a visual target.

MRI Siemens Magnetom Vision scanner. A standard whole-head coil and an echo-planar imaging (EPI) booster was used for all im- age acquisitions. Functional images with apparent transverse re- laxation time (T*2) were obtained with an EPI sequence (repeti- tion time TR =4.4 s; echo time TE =66 ms; flip angle FA =90°). Ten axial slices (7 mm thick, 2.1 mm inter-slice separation; 128128 pixels with a field of view of 240240 mm) were ac- quired. Voxel sizes were thus 1.881.889.1 mm. A total of 128 multi-slice images were acquired during performance of the track- ing task.

A T1-weighted structural image was next acquired (ten axial slices co-registered with the functional slices, using a conventional sequence; TR =350 ms, TE =6 ms, FA =90°, FoV =240240 mm). Slices were 256256 voxels of 0.980.989.1 mm. Immediately following the structural sequence, a second sequence of 128 func- tional images were collected, using the same EPI parameters de- scribed above, but with the slices shifted axially by 4.6 mm to cover the inter-slice separation.

Data analysis

Data were analysed with SPM-96 statistical parametric mapping software from the Wellcome Dept. of Cognitive Neurology, Lon- don, UK; (Friston et al. 1995a); implemented in Matlab (Math- works, Sherborn Mass., USA).

Spatial realignment and normalisation

The scans from each session were realigned to correct for motion using the first as a reference (Woods et al. 1992). The six parame- ters of this rigid body transformation were estimated using a least- squares approach (Friston et al. 1995b), based on an approximate linear relationship between the images and their partial derivatives with respect to parameters of the transformation. Following realign- ment, all images from each subject were transformed into a standard space. This normalising spatial transformation matched each scan (in a least-squares sense) to a template image using 12 parameter affine (linear) and quadratic (non-linear) three-dimensional transfor- mations. Attempts to register the functional or structural images to the spatial co-ordinates of the standard SPM brain templates (Talairach and Tournoux 1988) were unsuccessful; while some whole-brain images could be registered, images centred on the cere- bellum could not. Hence, the functional images from six sessions (three subjects, each repeated twice) were normalised to a template generated from the structural image from one subject smoothed with a 5 mm FWHM Gaussian filter. Again the parameters were estimat- ed using standard least squares after linearising the problem. As a fi- nal pre-processing step, the images were smoothed using an (5 mm FWHM) isotropic Gaussian kernel.

Imaging paradigm

Statistical inference

Axial slices were taken spanning the vertical extent of the cerebel- lum, but excluding the cerebral sensory and motor areas. Func- tional images were collected using an EPI sequence on a 1.5T

The SPM{t} were transformed to the unit normal distribution (SPM{Z}) with a threshold set at P=0.001, uncorrected. The re- sulting foci were then characterised in terms of spatial extent (k) and peak height (u). The significance of each region was estimated

Materials and methods

Statistical analysis

We tested subjects in two related visual-movement experiments. Three subjects were used in each experiment. All subjects gave their informed consent in accordance with the declaration of Hel- sinki, and the experiments were approved by the local ethical committee. One of the authors (RCM) was a subject; all other sub- jects were na to the tasks, but were given instructions before scan- ning and several minutes practice at each task once in position on the scanner bed. RCM performed both experiments; two other subjects performed experiment 1 only, and two subjects took part in experiment 2 only.

The condition, subject and co-variate effects were estimated ac- cording to the general linear model at each and every voxel (see Friston et al. 1995a). The design matrix included global activity as a confounding co-variate, and this analysis can therefore be re- garded as an ANCOVA (Friston et al. 1990). To test hypotheses about regionally specific condition or co-variate effects, the esti- mates were compared using linear compounds or contrasts. The resulting set of voxel values for each contrast constitute a statisti- cal parametric map of the t statistic, SPM{t}.

Document info
Document views13
Page views13
Page last viewedMon Oct 24 10:56:50 UTC 2016