Activation of cerebellar hemispheres in spatial memorization of saccadic eye movements: an fMRI study

Abstract

What mechanisms allow us to direct a precise saccade to a remembered target position in space? The cerebellum has been proposed to be involved not only in motor and oculomotor control, but also in perceptual and cognitive functions. We used functional MRI (Echoplanar imaging at 1.5 T) to investigate the role of the cerebellum in the control of externally triggered and internally generated saccadic eye movements of high and low memory impact, in six healthy volunteers. Memory-guided saccades to remembered locations of 3 targets (triple-step saccades) in contrast to either central fixation or to visually guided saccades activated the cerebellar hemispheres predominantly within lobuli VI-crus I. The same areas were activated when an analogous visuospatial working memory task was contrasted to the triple-step saccades. Visually guided saccades activated the posterior vermis and the triple-step saccades, contrasted to the working memory task, activated predominantly the posterior vermis and paravermal regions. Our data confirm the primary involvement of the posterior vermis for visually-triggered saccadic eye movements and present novel evidence for a role of the cerebellar hemispheres in the mnemonic and visuospatial control of memory-guided saccades.

Figure 1

Schematic plots of the tested saccade tasks (visually‐guided saccades, triple‐step saccades, and a test for visuospatial working memory). Upward deflection means rightward deviation, ranging between 0 degrees and 10 degrees of visual eccentricity, downward deflection means leftward deviation, respectively. On the abscissa, time is plotted in seconds. After 6.25 sec, the central fixation point reappears to start the next sequence. During the triple‐step task, the 3 targets had to be memorized while subjects kept on fixating the central fixation point. After a delay of 750 msec (memorization time), the fixation point disappeared as the go‐signal to perform saccades to the memorized locations of the three targets in the presented order. For the visuospatial working memory task, the target had to be acquired by a single visually‐guided saccade only when its location was identical to one of the three targets presented before, which happened randomly in 50% of the trials.

Figure 2

Group analysis data (6 volunteers) demonstrating the activation of the posterior vermal lobules VI–VII by visually guided saccades (VG) in transversal (A) and sagittal projection (B). C: The triple‐step task (TR) predominantly activates the left lobuli VI‐crus I (5 volunteers, one excluded because of movement artefacts). D: Cerebral activation foci by visually‐guided saccades (red) including the network comprising of the frontal eye field, supplementary eye field, parietal cortex, and visual areas. The activations of the cerebellar hemispheres, lobuli VI–VII, during the triple‐step task are accompanied by prefrontal activations (green). The activation pattern by visually‐guided saccades is plotted on top of the triple‐step saccades. The color bars represent the corresponding Z‐values (left for a–c, right including the overlay map in d).

Figure 3

Group analysis data of the differential task design. TR with VG as control (tr‐vg, 6 volunteers) predominantly activated areas within the lobules VI‐crus I and the lower part of the left cerebellar hemisphere (left) corresponding to the lobule VIIB‐VIII. TR with WM as control (tr‐wm) revealed activation within the vermal/left paravermal region (middle), whereas the reverse condition (wm‐tr, 3 volunteers) demonstrated predominant hemispheric activation of areas within lobule crus I (right). The color bar represents the corresponding Z‐values.

Figure 4

Summary of the activation foci of the different tasks after group analysis. The foci are schematically displayed within the Talairach space for the coronal (top left), sagittal (top right), and axial (bottom left) orientation. Bottom right: The foci overlaid onto a flattened surface of the cerebellum. Larsell lobules are marked. The tasks with a dominant oculomotor component (visually‐guided saccades, triple‐step vs. visuospatial working memory; light grey) are preferentially oriented along the midline, whereas the tasks with a cognitive preload localize within the lateral hemispheres (black/dark grey).

Figure 5

Modified schematical drawing of cerebro‐ponto‐cerebellar connections in the monkey demonstrating a subspecification within the cerebellum. A: Numbers represent Brodmann areas. B: Cerebellar regions are matched to the cerebral areas. Reproduced from Colin et al. [2002] with permission of the publisher.