The effects and neural correlates of visuomotor scaling of finger movements in virtual reality

(1)

The effects and neural correlates of

visuomotor scaling of finger movements

in virtual reality

Research report of second master project performed at the Institute of

Neuroinformatics (ETH Zurich and University of Zurich) in Zurich,

Switzerland

Name: Romy Bakker Student number: 5744385

Date: August 27th, 2012

Master Brain and Cognitive Sciences: Cognitive Neuroscience Supervisor: Johannes Brand

(2)

Abstract

The effects of adaptation to visuomotor feedback distortion have been extensively studied in arm reaching movements. However, fewer studies have been done in visuomotor scaling of fine finger movements. In this study we examine the behavioral effects and neural correlates of visuomotor scaling of finger movements in a target reaching task. The experiment was conducted in a behavioral set-‐up (18 subjects) and using Functional Magnetic Resonance Imaging (9 subjects). In both experiments subjects had to use their right index finger to move a cursor to a target, in which the target position and the feedback gain varied. Results show that subjects were in a persistent state of adaptation to the visuomotor gains, presumably because of the randomization of adaptation and control blocks. They were slower and had greater movement errors on trials with increasing distorted gain factors. Brain areas related with adaptation to the gain factors included left precentral gyrus and bilateral cerebellum, thalamus, precuneus, middle occipital gyrus and inferior temporal gyrus.

(3)

Introduction

Humans have a flexible adaptive sensorimotor system, leading to apparently effortless smoothing of their movements. However, unpredictable errors from the environment or from the movements themselves could disturb goal directed behavior. Therefore, signaling of errors and adapting to errors plays an important role in motor control. Adaptation to movement errors seems to be often implicit (Mazzoni and Krakauer, 2006; Krakauer, 2009). Also, intended movements seem to be updated by internal loops (Desmurget and Grafton, 2000).

Until now, studies on visuomotor adaptation mainly focused on inducing a sensory mismatch during arm movements by rotating visual feedback. Fewer studies have investigated adaptation effects to visual feedback scaling and few have done this with fine finger movements. Finger movements differ from hand and wrist movements, because finger movements are optimized for dexterous prehension tasks and hand movements more for postural control.

Diedrichsen and colleagues (Diedrichsen et al., 2005) compared neural correlates of different kinds of perturbations of arm reaching movements. Subjects adapted to visual feedback rotations of a virtual cursor even if the rotation direction was alternated randomly. They found particularly stronger activation in primary motor areas, anterior parietal areas, dorsal premotor cortex and cerebellum during visuomotor rotation tasks than during normal reaching movements. Krakauer and colleagues (Krakauer et al, 2000) compared adaptation to visuomotor rotation and scaling in a behavioral study. They found that subjects were adapting much faster to scaling distortion than to rotations. In 2004, the same group examined whether different brain areas were activated during adaptation to rotation and scaling, using PET (Krakauer et al, 2004). Subjects were adapting fast to new scaling factors, therefore they changed the visuomotor scaling factor after every 16 movements to keep them in an adaptation state. Adaptation to scaling activated the putamen and the contra lateral cerebellum. Additionally, adaptation to rotation recruited the preSMA, the ipsilateral premotor cortex, the posterior parietal cortex and the contralateral cerebellum.

Besides the PET study of Krakauer and colleagues (Krakauer et al, 2004), no other imaging study has investigated adaptation to a visuomotor scaling factor. Therefore, in this study we try to get new insights in the underlying neural mechanisms

(5)

of visuomotor scaling of finger movements. This will be tested in a paradigm that focuses only on visuomotor gain factor amplitude manipulation in a target reaching task, using fMRI. The study addresses the following questions: (1) do subjects adapt to changing visuomotor gains? (2) How do changing visuomotor gains influence the performance of the finger movement task? (3) What are the neural correlates of adaptation to visuomotor feedback scaling of finger movements? We expect that the performance measurements as peak flexing velocity and movement error will be different for visuomotor adaptation gains compared to normal reaching. Additionally, we expect to find activation changes in premotor cortex, primary motor cortex, SMA and inferior parietal lobe during the task. Finally, we expect that changing visuomotor gains will recruit the putamen and contra lateral cerebellum, as well as parietal and motor related areas.

(6)

Methods

Subjects

The study was performed in two separate experiments. In the first experiment subjects were tested solely on the behavioral level. In the second experiment the same behavioral design was used combined with fMRI and a visual observation condition was added. Subjects were only allowed to participate in one of the two experiments, to ensure they were naïve for the task. Twenty adults (Mean age 25.75 years, SD= 7.56) participated in the behavioral part. In the fMRI phase, ten adults (Mean age: 25.7 years, SD=3.97) participated. All were dominant right-‐hand as assessed by the Edinburgh Handedness Inventory (Oldfield, 1971): Behavioral subjects had a mean Laterality Quotient (LQ) of 82.4 (SD = 16.2) and mean Decile of 6.25. (SD= 2.9). The mean LQ of the fMRI subjects was 90.5 (SD=14.5) and mean Decile was 8.1 (SD= 2.8). Both groups were recruited at the University of Zurich and the Swiss Federal Institute of Technology Zurich. Subjects received compensation for their participation in terms of CHF 20, -‐. All had normal or corrected-‐to-‐normal vision and did not have any diseases affecting hand movements in any kind. They all gave their written informed consent for participation by signing a consent form.

Task and procedure behavioral experiment

Subjects were sitting at a desk with their forearms on the table (Figure 1a and 1b). The task, which was displayed in virtual reality, was projected from a LCD screen onto a mirror placed beneath the LCD screen. Subjects were not able to see their own hand, as the mirror hindered direct vision. The right arm was fixed to the table such that arm movements were blocked.

(7)

Figure 1(a). Experimental set-‐up behavioral experiment, (b) picture of the setup, subject is wearing the data glove and the hand is placed around the blue tube. Right arm is fixed to the table. The subject is

watching the task in virtual reality projected onto the mirror.

Subjects performed a target-‐reaching task in virtual reality, using only their right index finger. A 5DT fMRI compatible data glove was used to obtain positions of the index finger. Their right hand was placed around a well fitting tube, to ensure the hand was in a relaxed and comfortable position and that the index finger could move easily from there (Figure 2). There were three different sized tubes that were increasing in both diameter and length, to ensure subjects with various hand shapes could perform the task. The movement range of the task was 90 % of the full finger movement range, which was calibrated. For the calibration finger positions were measured at three different positions: (1) Closed hand position, although relaxed (Fig 2a), (2) extension of the hand, but not completely (Fig 2b), (3) the middle position between these two positions. The other fingers were kept fixed around the tube. The extended position (Fig 2b) was the starting position of the task, and experiment movements were in the flexing direction.

The position of the finger was represented by a blue cursor in virtual reality. The position of the blue cursor and the finger were scaled as 1:1, such that when the finger moved the cursor moved with the same speed on the screen. Task instructions included to move the blue cursor from the starting point (a yellow circle) to one of three differently targets distances (white circles) as fast and accurate as possible (Figure 3). The targets appeared periodically every two seconds and stayed for one second. The

(8)

order of the distances was randomized. The time course of a trial consisted of (1) bringing the blue cursor into the starting position, (2) appearance of the white target, (3) moving the cursor to the target and back to the starting position in one movement.

Figure 2 Schematic overview of the subject’s hand position around the tube. (a) Closed hand position, flexed position. (b) Start position of the task, extended position.

Figure 3. Target presentation and percentage finger extension. Yellow circle = starting position, white circles = target.

The task consisted of four conditions of 15 blocks each, manipulating the mapping of measured index finger input to cursor movement on the screen (visuomotor gain). Each block consisted of 10 movements of 2s to the targets. The first part of the experiment consisted of the Familiarization condition. In this condition only the reference gain was introduced (Fig. 4A). After the Familiarization, the blocks of the other three conditions were randomly shuffled. These conditions consisted of the Downscaling condition where the reference gain was shuffled with the Downscaling gain (Fig. 4B). Second, the Testing condition in which only the reference gain was introduced (Fig. 4C) and the Upscaling condition in which the reference gain was shuffled with the Upscaling gain (Fig. 4D). Hence, two different visuomotor scaling gains were introduced to map between the index finger and the virtual cursor. The trained reference-‐scaling factor was denoted as 1, the Downscaling factor was denoted as 0.5 and the Upscaling factor as 2. For the Downscaling gain the cursor’s speed is twice as slow as during the reference gain, for the Upscaling twice as fast.

(9)

Figure 4. Example blocks for each condition of the experiment: (A) Familiarization. (B) Downscaling, (C) Testing, (D) Upscaling. The gain factors are modulated in the Upscaling and Downscaling conditions

within one block.

Task and procedure fMRI experiment

The same task was performed during the MRI sessions. The task was slightly modified for the scanner. First, an additional condition was introduced: a visual observation-‐only condition, which took place after the familiarization (Fig 5). During this condition subjects had to observe their own movement trajectories from the familiarization condition, which were presented in a randomized order. They were not allowed to make any movements. Hence, the fMRI session was the same in procedure as the behavioral session, only with an additional condition. Second, the color of the cursor was changed from blue to red for the MRI session, for better contrast on the backprojection system used in the MRI scanner.

Figure 5. The design of the fMRI task: Similar to the behavioral design, only the addition of the visual observation condition. In total, 5 conditions: Familiarization, Visual Observation, Upscaling, Testing, and

(10)

Software

The virtual environment of the task in this study was programmed using Unity (Unity technologies, San Francisco, USA).

Equipment

A 5DT fMRI compatible data glove (Fifth Dimension Technologies, Irvine, California, USA) was used in this experiment to measure the flexion of the index finger with a sampling rate of 75 Hz. The signal was smoothed with an average filter of 100ms length (Lag in Games for Comparison: Unreal Tournament 3: ca. 133ms, via OnLive: ca. 150ms). At every time point we inferred the index finger tip position from a Kochanek-‐ Bartels spline constructed from three calibration points at 0%, 50% and 100% index finger extension at the beginning of the experiment.

Data acquisition

Scans were acquired using a Philips Achieva 3.0 Tesla MR scanner (Philips Medical Systems, Best, The Netherlands) with an 8-‐element head coil. Functional BOLD sensitive images were obtained using a single-‐shot gradient echo EPI pulse sequence (slices = 30, repetition time = 2s, echo time = 30 ms, flip angle = 77-‐80˚, field of view= 220, voxel size = 3x3x3) Using a sensitivity encoding (SENSE) with a reduction factor of 2, the influence of susceptibility artifacts was minimized. Additionally, using SENSE the possible number of slices acquired with one TR could be maximized. Following the functional scans a high-‐resolution anatomical scan was acquired, using a 3D T1-‐ weighted gradient echo sequence (TE/TR= 2.3/20 ms, FOV = 220x 220 mm2, matrix = 256x 256, slices = 180, slice thickness = 0.75 mm). All images were acquired in whole brain and in an oblique axial orientation.

Statistical analysis of behavioral experiment

The acquired movement traces were aligned to movement onset. Features as movement error, peak flexing velocity, time to target, reaction time, time to target were calculated and statistical outliers were removed using MATLAB (MathWorks, Natick, Massachusetts, USA). The effect of the visuomotor scaling gain, target distance and condition of the experiment were examined. . All data was first checked for normality and analyzed using Statistical Package for the Social Sciences (SPSS, Chicago, Illinois, USA): Repeated measures ANOVA were performed to normal distribute data to compare the means and also the standard deviations. To data that was not normally distributed,

(11)

non-‐parametric tests as the Friedman test were performed. All results were post-‐hoc corrected for multiple comparisons, using Bonferroni Correction.

FMRI preprocessing

All fMRI data were preprocessed and analyzed using Statistical Parametric Mapping (SPM8) (Welcome Trust Centre for Neuroimaging, London). Images were realigned to the mean of the functional scans, to correct for head motion. Additionally, the scans were wrapped in the y direction. Subsequently, the anatomical scan was coregistered to the realigned functional scans. Segmentation was conducted to the coregistered anatomical scans, using unified segmentation. The functional and anatomical scans were spatially normalized to the Montreal Neurological Institute (MNI) standard brain. At last, images were smoothed with a 6-‐mm full width at half maximum (FWHM) Gaussian kernel.

Statistical analysis of fMRI responses

Statistical fMRI analysis at group level was performed using the general linear model (GLM), as implemented in SPM8. A model was created defining five conditions: Familiarization, Visual Observation, Upscaling, Testing and Downscaling. As described earlier, a block-‐design was used for this experiment. Additionally, a high-‐pass filter of 128 seconds was applied to the data. The model was convolved with the standardized hemodynamic response function of SPM8, to disclose typical delays in fMRI responses.

In this report only the group results of the second level analyses will be reported. The Second level analysis was performed on the whole brain level. One-‐sample t-‐tests were performed to the contrasts. Statistics are reported at p<0.05 significance level, corrected for multiple comparisons using the False Discovery Rate (FDR). Results are reported in MNI coordinate system. For the anatomical labeling of the nearest grey matter, coordinates were first converted to Talairach coordinates, using a non-‐linear transformation. Anatomical label of nearest grey matter was determined using Talairach Daemon atlas in the Talairach Client (Research Imaging Center, University of Texas, Health Science Center, San Antonio)

(12)

Results

Behavioral experiment

Performance

All subjects felt comfortable during the task and accomplished the task. However, two subjects (one male and one female) were excluded from the analysis due to incorrect movement traces. These incorrect traces were due to a fault in calibration and due to not correct execution of the task.

Movement curves

Figure 7 shows average curves over subjects of the mean finger extension during the experiment. During the Familiarization condition, the average subject’s finger movement minimum extension lies within the range of the target area (Fig. 5A). We found a difference in behavior during the testing condition, which was like the Familiarization condition, also conducted with the reference gain (Fig. 5B.). The slopes of the movement traces were less steep for the testing condition compared to the Familiarization condition. Additionally, the movement extent towards targets 2 and 3 was decreased. Furthermore, traces with reference gain factor from movements in the Upscaling and Downscaling condition (Adaptation condition) were compared to movement traces of the Testing condition (Fig. 5C). The slopes of the movement traces during adaptation were less steep to target 2 and 3 compared to the testing condition. The comparisons for figure 5A to 5C were all with the reference gain, the only difference was the introduction of other flankering gains in the latter two cases. The flankering gain movements, which were the Upscaling and Downscaling gains, were not included in the previous comparisons, but showed to have influence on the gain 1 traces.

(13)

Figure 6. The average curves over subjects of the mean finger extension during the experiment. The surface areas represent +/-‐ 1 standard error. Horizontal black lines mark the edges of the 3 target

positions. The width of the virtual cursor is not shown, but corresponds to the width of the targets.

Inspection of the movement traces with the Upscaling gain factor revealed an overshoot in the movements to the target (Fig. 5D,E). Downscaling resulted in less steepness of the slope and an undershoot to the target (5E,F). Additionally, corrections in undershoot were visible for the downscaling factor; however subjects could not entirely correct for the initial undershoot.

Effect on conditions with reference gain factor

In this analysis we compared features of the reference gain movements from the Familiarization condition, the Testing condition and the two adaptation (Upscaling and Downscaling) conditions. A non-‐parametric Friedman test revealed significant differences between the three experimental conditions to target 2 (P<0.05) and target 3 (P<0.001) for peak flexing velocity. After applying a Bonferroni correction effects were found in peak flexing velocity to target 3 between familiarization and adaptation condition (p<0.01) and between testing and adaptation condition (p<0.001). Additionally, there was a significant effect found in movement error for target 1 (p<0.01) and target 3(p<0.01) both between the three conditions. Bonferroni correction

(14)

revealed significant effects between familiarization and adaptation to target 1 (p<0.01) and target 3 (p<0.01). Additionally, an effect was found to target 3 between testing and adaptation (p<0.01). All significant effects, including also some effects in movement error, are summarized in table 1A and 1B.

1A Movement Error

T1 T2 T3

Condition Mean STD Mean STD Mean STD

fam-test-adap P<0.01 P<0.001* n.s. P<0.001* P<0.01 n.s. fam-test n.s. n.s. n.s. n.s. n.s. n.s. fam-adap n.s. P<0.001* n.s. n.s. P<0.01 n.s. test-adap P<0.01 P<0.001* n.s. P<0.001* P<0.01 n.s.

1B Peak Flexing Velocity

T1 T2 T3

fam-test-adap n.s. P<0.01 p<0.05 n.s. P<0.001 n.s. fam-test n.s. n.s. n.s. n.s. P<0.001 n.s. fam-adap n.s. n.s. n.s. n.s. P<0.001 n.s. test-adap n.s. P<0.01 n.s. P<0.05 P<0.01 p<0.05

Table 1. Statistical differences in Means and Standard Deviations of (A) Movement error (ME) and (B) Peak Flexing Velocity (PFV) to the three different targets (T1, T2, T3) from the Familiarization (fam), Testing (test) and Adaptation (adap). Non parametric tests were performed, corrected for multiple comparison. * indicate the data was normal distributed, and repeated measures ANOVA were performed. Fam-‐test-‐adap indicates difference between 3 conditions (( -‐ ) indicates difference between). n.s. = no significant difference

(15)

Figure 7. Boxplots of the Peak Flexing Velocity of movements to target 3 from the Familiarization, the Testing and the Adaptation conditions. The Asterisks denote significant differences, obtained from a non-‐

parametric Friedman test, corrected for multiple comparisons.

Effect on different gains

To test the differences between visuomotor gains, data was compared during gain 1 from the Testing condition to data with gain 2 of the Upscaling condition and to gain 0.5 of the Downscaling conditions. As shown in figure 5 and 7 significant differences were found between the three gains to target 2 in the Adaptation block. Repeated measures ANOVA revealed a significant difference in error movement between the three gains (P<0.001). Post hoc analysis with Bonferroni correction was applied, resulting in a significance level at P<0.05 between gain 0.5 and gain 1, P<0.001 between gain 1 and 2 and P<0.001 between gain 0.5 and gain 2 (Fig. 7A). Additionally, a non-‐parametric Friedman test showed a significantly difference in peak flexing velocity between gains (P< 0.001). Post hoc Wilcoxon Signed-‐rank tests showed an effect between gain 0.5 and gain 1 (P<0.01), between gain 1 and 2 (P< 0.001) and between gain 0.5 and gain 2 (P< 0.001) (Fig.7B). The results are summarized in Table 2. There were also significant differences found in movement error and peak flexing velocity between the gains to target 1 and 3.

(16)

Figure 8. Boxplots of the Movement Error and Peak Flexing Velocity of movements to target 2 with the Downscaling, reference and Upscaling gain. The Asterisks denote significant differences obtained from (a)

repeated measures ANOVA, corrected for multiple comparisons (b) non-‐parametric Friedman test, corrected for multiple comparisons.

T1 _T2 _T3

Gain Mean STD Mean STD Mean STD

0.5-1-2 x x p<0.001* p<0.001* x x 0.5 -1 x x p<0.05* n.s. P<0.001 n.s. 0.5 -2 x x p<0.001* p<0.001* x x

1 - 2 p<0.001 p<0.001* p<0.001* p<0.001* x x

T1 T2 T3

0.5-1-2 x x p<0.001 p<0.001 x x 0.5 -1 x x p<0.01 n.s. P<0.001 n.s. 0.5 -2 x x p<0.001 p<0.001 x x

1 - 2 p<0.001* n.s. p<0.001 p<0.001 x x

Table 2. Statistical differences in Means and Standard Deviations of (A) Movement error (ME) and (B) Peak Flexing Velocity (PFV) to the three different targets (T1, T2, T3) from the three different gains Upscaling (2), Testing (1) and Downscaling (0.5). Non parametric tests were performed, corrected for multiple comparison. * indicate the data was normal distributed, and therefore repeated measures ANOVA were performed. 0.5-‐1-‐2 indicates difference between 3 gains (( -‐ ) indicates difference between). n.s. = no significant difference. X indicates not a possible combination.

(17)

Effect of different targets

Movement error significantly differed between proprioceptive targets 1,2,3 while the reference gain was used (p<0.01). Correction for multiple comparisons showed a significant difference between target 1 and 2 (p<0.01) and between target 2 and 3 (P<0.01). There were no significant differences found in movement error during the manipulated gains. Peak flexing velocity differed significantly between the 3 targets (P<0.001) during the reference gain. Bonferroni correction showed effects between target 1 and 2 (P<0.001), target 2 and 3 (P<0.001) and between target 1 and 3 (P<0.001). Furthermore, during the downscaling gain peak flexing velocity differed significant between target 2 and 3 (P<0.001). Additionally, during the Upscaling gain a significant effect was found in peak flexing velocity between target 1 and 2 (P<0.001). An overview of the effects are shown in Table 3.

Gain 0.5 Gain 1 Gain 2

Target Mean STD Mean STD Mean STD

T1- T2 -T3 x p<0.001* p<0.01 p<0.001 x x T1 -T2 x x p<0.01 n.s. P<0.001 n.s. T1 - T3 x x n.s. p<0.001 x x T2 - T3 n.s. p<0.001* p<0.01 p<0.001 x x

Gain 0.5 Gain 1 Gain 2

Target Mean STD Mean STD Mean STD

T1- T2 - T3 x x p<0.001 p<0.001 x x T1 -T2 x x p<0.001 P<0.01 P<0.001 p<0.01 T1 - T3 x x p<0.001 p<0.001 x x T2 - T3 p<0.001 p<0.001 p<0.001 p<0.01 x x

Table 3. Statistical differences in Means and Standard Deviations of (A) Movement error (ME) and B) Peak Flexing Velocity (PFV) with three different gains (gain 0.5, gain 1, gain 2) to the three different targets (T1, T2, T3). Non parametric tests were performed, corrected for multiple comparison. * indicate the data was normal distributed, and therefore repeated measures ANOVA were performed. T1-‐T2-‐T3 indicates difference between 3 targets (( -‐ ) indicates difference between). n.s. = no significant difference

(18)

fMRI experiment

Behavioral performance

None of the subjects reported discomfort and except for one, all subjects accomplished the sessions. One female was excluded from the analysis due to not entirely finishing of all the tasks.

The movement traces and the behavioral performance of the task showed no abnormalities and seemed correct after a quick examination. The analyses of the behavioral data of the fMRI sessions are still ongoing and therefore not discussed in this report.

Brain activation

The neural correlates that are examined are based on second level analysis of in total nine subjects. The results will be discussed per contrast. We studied various contrasts and the most interesting contrasts are discussed here and the significant results on remaining contrasts could be found in the appendix. Additionally, all coordinates of the peak activations and activated areas could also be found in the appendix.

Familiarization versus rest

The first analysis examined brain activation during the familiarization condition versus rest (Fig. 8). Activation patterns were found in bilateral parietal cortex (BA 7/40), as well as in the occipital lobe (BA 19/37). Dominant activation in the left hemisphere was found in inferior parietal lobe (IPL), precentral gyrus (BA 6) and precuneus (BA 7). Furthermore, right dominant areas as the inferior temporal gyrus (ITG), precuneus (BA7) and middle occipital gyrus (MOG) were recruited.

Figure 9. Familirisation versus rest (FDR corrected p< 0.05) T-‐contrast vector = [ 1, 0, 0 , 0, 0]

(19)

Adaptation versus rest

The analysis of all conditions in adaptation (Testing, Upscaling and Downscaling) revealed strong activation in the left hemisphere as the precentral gyrus (BA 6), middle frontal gyrus (MFG), Inferior frontal gyrus (IFG), Precuneus (BA 7), lentiform nucleus (putamen), thalamus, insula (BA 13) and IPL. Additionally, in the right hemisphere activation was found in the IFG, MTG, insula, putamen and inferior semi-‐lunar lobule. Results are summarized in Figure 10.

Figure 10. Adaptation versus rest (FDR corrected 0.05 ) t-‐contrast vector [0, 0, 0.33, 0.33, 0.33]

Upscaling & Downscaling versus rest

To examine the effect of the Up-‐ and Downscaling visuomotor gains the same analysis was performed, but without the testing condition versus rest. An activation pattern was visible in several similar areas such as the left precentral gyrus (BA 6), precunues (BA 7), inferior semi-‐lunar lobule. Additionally, activation was found in the right culmen (cerebellum), MTG, and inferior semi-‐lunar lobule. Activation patterns are shown in Figure 11.

Figure 11. Upscaling & Downscaling versus rest (FDR corrected 0.05) t-‐contrast vector [0, 0, 0.5, 0, 0.5]

(20)

Upscaling & Downscaling versus Testing

This analysis compared the Up-‐ and Downscaling conditions versus Testing (Figure 12). There was little difference found. However, after changing the significance level to uncorrected p<0.05 with an extended voxel treshold of 133 a pattern of frontal and temporoparietal areas became visible. This pattern included areas left hemisphere areas as the medial frontal gyrus (BA 8), superior frontal gyrus (BA 8/10), middle frontal gyrus (BA 10), insula (BA 13), superior temporal gyrus (BA 22). Right hemisphere activations included superior frontal gyrus (BA 8), cingulate gyrus and postcentral gyrus. It has to be investigated whether additional subject data will also show these effects with FDR correction enabled.

Figure 12. Up & Down versus testing (uncorrected p= 0.05 ext voxel treshold of 133) t-‐contrast vector [0, 0, 0.5, -‐1, 0.5 ]

Familiarization versus Visual Observation

When contrasting the Familiarization versus the Visual Observation, which is a replay of the Familiarization movement traces, activation as found in right cerebellum (culmen, inferior semi-‐lunar lobule),inferior frontal gyrus (BA 9/44), insula lentiform gyrus, postcentral gyrus. And recruited areas in the left hemisphere were the precentral gyrus, inferior parietal lobule (BA 40), inferior frontal gyrus (BA 9/44), insula, lentiform gyrus and cingulate gyrus.

(21)

Adaptation versus Visual Observation

Substracting activation of the visual control condition from the adaptation conditions (testing, Upscaling and Downscaling) revealed activation in precentral gyrus, middle frontal gyrus (BA 6), Culmen, inferior semi-‐lunar lobule, cingulate gyrus and putamen (Figure 13).

Figure 14. Adaptation _ visual observation (FDR corrected 0.05) t-‐contrast vector [ 0, -‐1, 0.33, 0.33, 0.33]

Visual Observation condition versus rest

Finally, during the visual observation condition in which no movements were done, dominant activation was shown in bilateral temporal lobe (BA 21/ 22) and occipital lobe (BA 18/ 37). High significantly activated areas were middle temporal gyrus (MTG), ITG, MOG, precentral gyrus, precuneus, middle and inferior frontal gyrus and inferior parietal lobe. Activation patterns are visualized in Figure 9.

(22)

Discussion

This was to our knowledge the first imaging study to investigate the effects of Up-‐ and Downscaling of feedback of finger movements. The results revealed that during the Familiarization condition, subjects learned to move the cursor to the target location. However, introduction of Up-‐ and Downscaling gains around the reference gain influenced the movements. Movements became slower as revealed by the decrease of peak flexing velocity over conditions. Additionally, movement errors increased over conditions, as more flankering Up-‐ and Downscaling gains were introduced. These findings show that humans react to the introduction of changing visual feedback by adapting their behavior to move more cautiously.

Furthermore, movements performed during the Upcaling and Downscaling conditions showed undershoots and overshoots in movement extend respectively. Analyses between the different gains pointed out that movement error was increased for the Upscaling and Downscaling gains versus the reference gain. Peak flexing velocity increased with increasing visuomotor gain factor. Differences in movement error between target distances were only visible during the reference gain. There are no differences found in movement errors between targets while the visuomotor scaling gains were used. Hence, the performance of movement extent during visuomotor scaling gains was impaired regardless which target the subjects had to reach for. However, peak flexing velocity did differ between the targets for all scaled gains. The described findings pinpoint the fact that visuomotor feedback scaling only can manipulate finger movements. Even when subjects tried to correct for the undershoot errors they made in the Downscaling condition, they were not able to adapt entirely during the trials and also not during the entire experiment. Concluding, due to the inability to adapt to the errors in the movements, there was no learning effect of the scaled gains found during the experiment. These results are in line with the study of Krakauer and colleagues (Krakauer et al., 2003) who also showed greater movement extend and peak flexing velocity for adaptation to gain factors. Additionally, their subjects could not adapt completely to the new gain factor, because of the changing of the gain factor every 16 movements. However, they showed a small rapid adaptation however not completely, whereas in our study subjects showed no adaptation at all.

(23)

With respect to the first data of the fMRI part of this research project, activation patterns of the Familiarization condition showed as expected activity in the motor regions, left cerebellum, parietal areas, basal ganglia as well as visual areas activation. This is in line with earlier research on simple finger movements (Boecker et al, 1994; Ueno et al, 2010). However, activations were mostly bilateral instead of contra laterally to the movement, also in motor regions. It may be argued that the network activated was more due to cognition rather than just the effect of the motor task in the Familiarization condition. In the visual observation condition, which was a control condition without movements, visual areas such as the V4, middle temporal and occipital gyrus were recruited. In addition to the visual areas, some frontal activation and parietal areas were also observed. The activated network included some motor imagery areas, such as the inferior frontal cortex and somatosensory cortex (Ueno et al, 2010).

The activation in the cerebellum and motor areas found in the Upscaling, Downscaling and Testing conditions, were as expected stronger activated contra laterally to the movement. Additionally, other areas as the putamen, insula and thalamus, also recruited during these conditions, were activated bilaterally. This was also found by Krakauer and colleagues (Krakauer et al; 2003) who reported putamen and contralateral cerebellum activation during visuomotor scaling. Nonetheless, activation only during the Upscaling and Downscaling did not differ much from the contrast with the Testing block. After contrasting the Upscaling and Downscaling conditions versus Testing activation, almost no activation was left. After changing the significance level to uncorrected an interesting network was exposed, including the insula, frontal and temporoparietal areas. This network resembles the salient distinguishing network described by Menon and Uddin (Menon and Uddin, 2010). However, as it was not corrected for multiple comparisons, this finding should be regarded cautiously. Finally, after contrasting the Familiarization condition with the visual observation control condition, more cerebellum and inferior frontal gyrus activation was visible than in during only the Familiarization condition.

Concluding, this study showed that a change of the visuomotor gain factor by visual feedback scaling has a great impact on the performance of a finger movement task. Continuously switching back to the reference gain factor and randomizing gains as well as the conditions, kept subjects in a persistent state of adaptation. Therefore, we conclude that our design is suitable for examining brain mechanisms of adaptation to

(24)

new visuomotor gain factors. Neural correlates of visuomotor gains show as expected motor areas, visual areas and even areas of salient distinguish functioning.

Limitations and suggestions for further research

For this report several things could not be taken into account yet, because the analyses are still ongoing. For the experiment more subjects were already scanned and some more might be following, which could increase the power of our results. Also, electromyography (EMG) was measured during the experiment and is currently analyzed. With these data we can control for muscle movements in the visual control condition. Furthermore, the behavioral data of the fMRI session and the activation from the first level analysis can be correlated to get more insights about the neural correlations of the changes in behavior we observed in this experiment.

(25)

Acknowledgements

I would like to thank a couple of people who have been a great help during my project. First of all I would like to thank Kynan Eng for introducing me and letting me taking part in this project and the rehabilitation research group. I would like to thank Kynan, Marie-‐Claude Hepp-‐Reymond and Daniel Kiper for the great help, valuable feedback, tips and knowledge during this project. I would like to thank Lars Michels for the great help during planning, scanning and analyses of the fMRI experiment. Most of all I would like to thank Johannes Brand: It was really great working together. Thank you for the valuable feedback and knowledge you brought me. I would like to thank the rest of the rehabilitation group for the advices during meetings. Additionally, I would like to thank the Institute for Neuroinformatics in general for its great hospitable , the good atmosphere, what made my stay there great. Finally, of course I would like to thank all my other co-‐students who made my stay in Zurich wonderful.

(26)

References

Boecker, H., Kleinschmidt, A., Requardt, M., Hanicke, W., Merboldt, K.D. and Frahm, J. (1994). Functional cooperativity of human cortical motor areas during self-‐ placed simple finger movements. A high-‐resolution MRI study. Brain, 117, 6, 1231-‐1239.

Desmurget, M. and Grafton, S. (2000). Forward modeling allows feedback control for fast reaching movements. Trends Cogn Sci, 4, 423-‐431.

Diedrichsen, J., Hashambhoy, Y., Rane, T. and Shadmehr, R. (2005). Neural Correlates of Reach Errors. The Journal of Neuroscience, 2005, 25, 9919 -‐9931.

Krakauer, J.W. (2009). Motor learning and consolidation: the case of visuomotor rotation. Adv Exp Med Biol, 629, 405-‐421.

Krakauer, J.W., Pine, Z.M., Ghilardi, M-‐F. and Chez, C. (2000). Learning of visuomotor transformations for vectorial planning of reaching trajectories. The Journal of

neuroscience : the official journal of the Society for Neuroscience, 20, 8916-‐24.

Krakauer, J.W., Ghilardi, M-‐F., Mentis, M., Barnes, A., Veytsman, M., Eidelberg, D., and Ghez, C. (2004). Differential cortical and subcortical activations in learning rotations and gains for reaching: a PET study. Journal of neurophysiology, 91, 924-‐933.

Mazzoni, P. and J.W. Krakauer (2006). An Implicit Plan Overrides an Explicit Strategy during Visuomotor Adaptation. Journal of Neurophysiology, 26, 3642-‐3645.

Menon, V. and Udding, L.Q. (2010). Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct, 214, 655-‐667.

Oldfield, R.C. (1971). The assessment and analysis of handedness: the Edinburghinventory. Neuropsychologia, 9, 97–113.

(27)

Pearson, T.S., J.W. Krakauer, and P. Mazzoni (2010). Learning Not to Generalize : Modular

Adaptation of Visuomotor Gain. New York: 2938-‐295

Ueno, T., Inoue, M., Matsuoka, T., Abe, T., Maeda, H. and Morita, K. (2010). Comparison between real sequential finger and imagery movements: An fMRI study revisited.

(28)

APPENDIX

Additional results behavioral experiment

Table A1. Statistical differences in Means and Standard Deviations of (A) Movement error (ME) and (B) Peak Flexing Velocity (PFV) to the three different targets (T1, T2, T3) from the Familirization (fam), Testing (test) and Adaptation (adap). Non parametric tests were performed, corrected for multiple comparison. * indicate the data was normal distributed, and repeated measures ANOVA were performed. Fam-‐test-‐adap indicates difference between 3 conditions (( -‐ ) indicates difference between). n.s. = no significant difference

A1.A Time to Target

T1 T2 T3

fam-‐test-‐ adap n.s. p<0.05 p<0.05* n.s. p<0.01 p<0.05 fam-‐test n.s. p<0.01 n.s. n.s. p<0.01 n.s. fam-‐adap n.s n.s. n.s. n.s. n.s. n.s. test-‐adap n.s. n.s. p<0.01* n.s. n.s p<0.01

A1.B Reaction Time

T1 T2 T3

fam-‐test-‐ adap n.s. p<0.01 n.s. p<0.001 n.s. p<0.01 fam-‐test n.s. p<0.001 n.s. p<0.01 n.s. p<0.01 fam-‐adap n.s p<0.001 n.s. p<0.001 n.s. n.s. test-‐adap n.s. p<0.05 p<0.01 p<0.05 p<0.01 p<0.01

A1.C Time to Minimum

T1 T2 T3

fam-‐test-‐ adap n.s. p<0.01 p<0.001 p<0.05 p<0.001 p<0.01 fam-‐test n.s. n.s. p< 0.05 n.s. p<0.001 n.s. fam-‐adap n.s n.s. p<0.01 p<0.01 p<0.001 n.s. test-‐adap n.s. n.s. p<0.01 n.s. p<0.001 p<0.01

(29)

A1.D Mean Flexing Velocity

T1 T2 T3

fam-‐test-‐

adap n.s. n.s. P<0.01 n.s. p<0.001 p<0.05

fam-‐test n.s. n.s. n.s. n.s. p<0.001 n.s.

fam-‐adap n.s n.s. P<0.01 n.s. p<0.001 n.s.

(30)

Table A2. Statistical differences in Means and Standard Deviations of (A) Time to Target, (B) Reaction Time, (C) Time to Minimum and (D) Mean Flexing velocity to the three different targets (T1, T2, T3) from the three different gains Upscaling (2), Testing (1) and Downscaling (0.5). Non parametric tests were performed, corrected for multiple comparison. * indicate the data was normal distributed, and therefore repeated measures ANOVA were performed. 0.5-‐1-‐2 indicates difference between 3 gains (( -‐ ) indicates difference between). n.s. = no significant difference. X indicates not a possible combination.

A2.A Time to Target

_T1 _T2 _T3

0.5-‐1-‐2 x x P< 0.001 n.s. x x

0.5-‐ 1 x x x n.s. P< 0.01 P<0.01

0.5 – 2 x x P< 0.01 n.s x x

1-‐2 P< 0.001 P<0.01 P< 0.001 n.s. x x

A2.B Reaction Time

T1 T2 T3

0.5-‐1-‐2 x x P<0.05 n.s. x x

0.5-‐ 1 x x n.s. n.s. n.s. n.s.

0.5 – 2 x x n.s. n.s. x x

1-‐2 n.s. n.s. P<0.017 n.s. x x

A2.C Time to Minimum

T1 T2 T3

0.5-‐1-‐2 x x p<0.001 p<0.001 x x

0.5-‐ 1 x x p<0.001 n.s. p<0.001 n.s.

0.5 – 2 x x p<0.001 p<0.01 x x

1-‐2 n.s. p<0.001 n.s. p<0.001 x x

A2.D Mean Flexing Velocity

T1 T2 T3

0.5-‐1-‐2 x x p<0.001 p<0.05 x x

0.5-‐ 1 x x p<0.001 n.s. p<0.001 n.s.

0.5 – 2 x x p<0.001 n.s. x x

1-‐2 n.s. p<0.01 n.s. p<0.01 x x

The effects and neural correlates of visuomotor scaling of finger movements in virtual reality