Neuroradiology (2007) 49:73–81 DOI 10.1007/s00234-006-0162-4

FUNCTIONAL NEURORADIOLOGY

Functional MRI of the cervical spinal cord on 1.5 T with fingertapping: to what extent is it feasible? N. Govers & J. Béghin & J. W. M. Van Goethem & J. Michiels & L. van den Hauwe & E. Vandervliet & P. M. Parizel

Received: 1 March 2006 / Accepted: 7 September 2006 / Published online: 21 November 2006 # Springer-Verlag 2006

Abstract Introduction Until recently, functional magnetic resonance imaging (fMRI) with blood oxygen level-dependent (BOLD) contrast, was mainly used to study brain physiology. The activation signal measured with fMRI is based upon the changes in the concentration of deoxyhaemoglobin that arise from an increase in blood flow in the vicinity of neuronal firing. Technical limitations have impeded such research in the human cervical spinal cord. The purpose of this investigation was to determine whether a reliable fMRI signal can be elicited from the cervical spinal cord during fingertapping, a complex motor activity. Furthermore, we wanted to determine whether the fMRI signal could be spatially localized to the particular neuroanatomical location specific for this task. Methods A group of 12 right-handed healthy volunteers performed the complex motor task of fingertapping with their right hand. T2*-weighted gradient-echo echo-planar imaging on a 1.5-T clinical unit was used to image the cervical spinal cord. Motion correction was applied. Cord

activation was measured in the transverse imaging plane, between the spinal cord levels C5 and T1. Results In all subjects spinal cord responses were found, and in most of them on the left and the right side. The distribution of the activation response showed important variations between the subjects. While regions of activation were distributed throughout the spinal cord, concentrated activity was found at the anatomical location of expected motor innervation, namely nerve root C8, in 6 of the 12 subjects. Conclusion fMRI of the human cervical spinal cord on an 1.5-T unit detects neuronal activity related to a complex motor task. The location of the neuronal activation (spinal cord segment C5 through T1 with a peak on C8) corresponds to the craniocaudal anatomical location of the neurons that activate the muscles in use.

N. Govers : J. Béghin University of Antwerp, Universiteitsplein 1, 2610 Wilrijk, Belgium

Introduction

N. Govers : J. Béghin : J. W. M. Van Goethem (*) : L. van den Hauwe : E. Vandervliet : P. M. Parizel Department of Radiology, University Hospital of Antwerp, Wilrijkstraat 10, 2650 Edegem, Belgium e-mail: [email protected] J. Michiels Siemens Medical Solutions, Brussels, Belgium

Keywords BOLD imaging . Spinal fMRI . Fingertapping . Cervical spinal cord

Until recently, functional magnetic resonance imaging (fMRI), introduced in 1990, was mainly used to study brain physiology. It has become a powerful instrument to study local neuronal activity involved in motor, sensory and/or cognitive tasks. Extension of this technique to the spinal cord shows great promise for research and clinical applications. The current knowledge of neuroanatomy of the spinal cord is mainly based upon invasive anatomical and pathological research. fMRI is a promising technique to demonstrate neuronal activity of the spinal cord, as it is

74

performed in a noninvasive way. Thus the technique may have important clinical uses, for example in the rehabilitation of spinal cord injuries. fMR images are mostly based on blood oxygen leveldependent contrast (BOLD). This technique depends on local changes in blood flow, which occur during neuronal activity and which are detected as inhomogeneities in the local magnetic field, due to blood oxygenation metabolites. Oxyhaemoglobin has a negligible effect on the magnetic field. Deoxyhaemoglobin, in contrast, causes inhomogeneities in the magnetic field that decrease the T2* relaxation rate and thus signal response on T2*-weighted images. Neuronal activity initiates an increase in local blood flow (and thus of oxyhaemoglobin), which actually exceeds the increased local metabolic demands. This gives rise to a reduction in the relative concentration of paramagnetic deoxyhaemoglobin [1, 2]. The latter effect leads to diminished magnetic field inhomogeneities and increased T2*-weighted signal intensity [3]. To correlate changes in blood flow with specific neuronal activity, the technique has to be sufficiently fast. To achieve a sufficient time resolution, the signal-to-noise ratio and image quality is less. Nevertheless, the fMR imaging technique gives reproducible results, as demonstrated by Yetkin et al. [4]. The activity of the spinal cord and spinal nerves gives rise to haemodynamic effects [5]. Motor tasks activate motor neurons in the anterior horns of the spinal grey matter, whereas sensory tasks produce a more widespread response, involving multiple sensory tracts synapsing at different levels. The number of published investigations on fMRI of the spinal cord is limited, suggesting major technical challenges in measuring the BOLD effect at the level of the spinal cord. The most substantial problems are due to limited spatial resolution, periodic movements (pulsations of the cerebrospinal fluid and local blood flow) and haemodynamic washout (rapid diffusion of local blood flow caused by the many anastomoses of the venous network of the cervical spine). As early as the 1990s, previous researchers had demonstrated fMRI signal changes in the spinal cord, associated with motor neuronal activity [6, 7]. In more recent years, task dependency has been demonstrated in studies involving [5, 8–10] gross motion (elbow flexion, wrist extension, finger abduction and fist clenching). As well as studies involving motor tasks, a few studies using sensory stimuli (a thermal stimulus, a pain stimulus, blowing puffs on the palm of the hand or a small tactile stimulus) have demonstrated sensory activity in the spinal cord [10–14]. The purpose of our study was to determine whether the current fMRI BOLD technique, with a conventional field strength of 1.5 T and a standard cervical coil, has sufficient

Neuroradiology (2007) 49:73–81

sensitivity and resolution to demonstrate functional activity of spinal nerves and spinal cord segments involved in a complex fine motor task, namely fingertapping.

Methods Population A group of 12 healthy, right-handed volunteers (5 men, 7 women) without a history of neurological or psychiatric disease participated. We selected a young population in order to rule out age-dependent signal decreases in BOLD contrast [15, 16] and the ages of our subjects ranged from 18 to 25 years. The study was approved by the local Committee for Medical Ethics and informed consent was obtained from every volunteer. An experienced neuroradiologist studied the anatomical MR images to rule out subjects with anomalies of the spinal cord or spinal canal. All selected volunteers were included in the study.

Paradigm The 12 volunteers were instructed to perform a fingertapping task. Fingertapping is the sequential apposition of the different fingers towards the thumb. This task was executed at an individually fixed rapid rhythm [17], resting and activation were alternated in 30-s epochs, repeated three times [18]. All subjects executed the task with their dominant right hand. The volunteers were asked not to make any other movements and to close their eyes during the experiment.

Positioning The volunteers were positioned supine as comfortably as possible, head and neck fixed and the right arm fixed with straps and weights, to minimise or eliminate motion artefacts.

MRI protocol Imaging was done on a 1.5-T superconducting magnet with a 40 mT/m gradient (Siemens Sonata, Erlangen, Germany). A standard receiver neck coil was used for all images. On the basis of spin-echo T1-weighted anatomical survey images all cervical segments could be identified and the sagittal image through the midline of the cervical cord was

Neuroradiology (2007) 49:73–81

selected. This sagittal section was used to position the fMRI sequence. The BOLD technique was only used in a transverse plane, from C5 to T2. For the fMRI experiment we used the same basic sequence as the one we use in brain fMRI. This is a single shot interleaved EPI-sequence [19, 20] with a TR of 3000 ms, a TE of 50 ms and a flip angle of 90°; K-space was not segmented; 30 adjacent slices, thickness 3 mm, matrix 64×64, 192× 192 mm. This sequence uses a single 180° refocusing pulse to fill the centre of the K-space (spin echo), while the remaining refocusing is done with gradients resulting in a T2*weighted image. We used paravertebral saturation slabs but no gating. During each cycle, ten images of resting and ten images of activation were acquired. For data analysis the first two images of each cycle were omitted, reflecting the attainment of a steady state of oxy- and deoxyhaemoglobin blood concentrations (Fig. 1). The total time needed to perform the paradigm was 3 min. The entire fMRI experiment (including instructions to the volunteer, positioning and acquiring of the images) took about 20 min per subject. Transverse T1-weighted spin-echo images were acquired for anatomical reference using the same slice

75

thickness and field of view as the fMRI images but with a matrix of 128×128 and a TR/TE 700/15.

Data analysis For data processing, specific fMRI software from the manufacturer of the MRI system (Siemens Medical Solutions, Erlangen, Germany) and commercially available software (BrainVoyager, Brain Innovation, Maastricht, The Netherlands) was used. A region of activation was defined by a significant augmentation of the fMRI signal. For each subject the images were statistically analysed pixel by pixel during resting and activation to determine taskdependent activity [21]. The calculated z-scores were colour-coded and superposed on an anatomical image of the same region. The 30 transverse sections of the final fMR image were observed per subject for activity and signal intensity in and around the spinal cord. All images were processed identically with a z-value cut-off of 4. The regions with the highest z-score were coded “3”, those with the lowest z-score “1” and those with an intermediate score “2”. At each level the highest z-score in or around the spinal cord was taken into account. Regions without activity were coded “0”. Disturbances by susceptibility artefacts were interpolated averaging the values of the previous and the next sections. The anatomical location of activity in the spinal cord was referred to the nearest adjacent vertebra by means of the anatomical survey image, since the inferior vertebral notch (the inferior border of the pedicle) is a reliable marker for the nerve root at the cervical level [22]. A distinction between the ipsi- and contralateral sides was also made. For each subject a graph was made of the values of the response score in function of the section, separately for the left and right side. At each level the area under the curve was calculated and divided by the number of sections to calculate the average response score per vertebral level [21]. Using an analysis of variance test (ANOVA repeated measures) the statistical significance of the differences in signal response score between different levels (1), between the left and right side (2), and between levels of the same side (3) were assessed [21].

Results Fig. 1 BOLD response curve. This curve is an example of the percent signal change of the BOLD sequence for each of the ten measurements during activity (numbered 1 to 10) and averaged over all cycles of activity for the significant voxels. As a reference two measurements before and after activity are also included (−1 and 0, and 11 and 12, respectively). Units are percentage signal changes compared to the signal averaged over all rest states (zero level). The signal change after reaching the steady state is about 2% to 3%

All 12 subjects showed significant fMRI responses in the cervical spinal cord, most of them both left and right (Figs. 1 and 2). A significant difference between activation at (or adjacent to) the dorsal or ventral horns was not found. Table 1 gives a summary of the results. For each subject the average response score per vertebral level on the left and the right side is shown.

76

Neuroradiology (2007) 49:73–81

Fig. 2 a fMR image superimposed on the anatomical spin-echo T1weighted image. This image shows a typical response in the voxels adjacent to the spinal cord, probably in a major draining vein. Although the task was right handed, the response was on the left side. b fMR image superimposed on the original EPI image. The same patient and level as in a. This image shows that the fMRI response is in the same anatomical location as shown on the anatomical overlay

image, effectively confirming that there is no major mismatch in the region of interest. c fMR image superimposed on the anatomical spinecho T1-weighted image in a different subject. This image shows a less typical response inside the spinal cord. Most of the activation was seen in the immediate perimedullary region (a), but both responses were found

The sums of the average response scores for each subject for each vertebral level are shown in Fig. 3. Statistically, using the ANOVA repeated measures test, it was not possible to demonstrate a significantly different pattern of activation between different levels (1), between left and right side (2), or between levels of the same side (3); P values were all >0.05.

Discussion Sagittal images from our own previous (unpublished) research demonstrate that fMRI responses in fingertapping are localized at spinal cord levels C5 to T1, consistent with the findings of Backes et al. [5]. This research also shows the superiority of transverse images in localizing regions of

Table 1 Average response score per vertebral level. In all subjects uniquely at C4 on the right side. Subject 7 showed activation at C7 and Table 1 Average response scorethere per vertebral level.variability In all subjects werewas found. general there wassides a marked in the responses were found. In general was a marked in the responses T1, which very In prominent on both of C7 variability and only slightly side of results in this subject were unreliable of responses between the the subjects. subjects. Subject Subject 11showed showedactivation activationonly onlyononthe left present onC5 thebut lefttheside of T1. Subject 8 showed bilateral because activation artefacts on most of the Subject showed little activation, equallyparticularly distributedatfrom C5 to C7. less Subject 3 showed all levels. the left side of C5 butimages. the results in 2this subject were unreliable C4 and much at T1. Subject 9activation showed aatpeak value There was peak value at C4ofonthe theimages. left side,Subject and on2both the left right side there a maximum at C7 and because of aartefacts on most showed littleand the at C4, especially on was the left side. Activation was T1. alsoSubject present 4at showed C5 and activation atequally all levelsdistributed except T1 from and a peak C7, more on right side. Subjecta5striking showedmaximum is activation at C7 on of theC7 right activation, C5 tovalue C7.atSubject 3 pronounced showed C6.the Subject 10 showed ononly the left side as side. Subject 6 showed activation at C4atonC4theonright Subject well 7 showed activation at C7 at andallT1, which was verySubject prominent both activation at all levels. There was uniquely a peak value the side. left side, as diffuse activation vertebral levels. 11 on showed sidesonofboth C7 and slightly present the was left side of T1. Subject 8 showed bilateral activation particularly at C5. C4 and much12less at T1. aSubject and the only left and the right sideon there a maximum at C7 and only minimal activation at C4 and Subject showed marked9 showed a peak value at activation C4, especially the left side. T1 Activation was alsopeak present at on C5 the andright C6. side Subject 10 and showed a striking T1. Subject 4 showed at allonlevels except and a peak value of C7 T1 and diffuse maximum activation on the left side as well as diffuseonactivation all vertebral Subject value at of C7,C7more pronounced the rightatside. Subject 5levels. showed is 11 leftshowed side only minimal activation at C4 and C5. Subject 12 showed a marked peak value on on the the right sideside. of C7 and T16 and diffuse activation on the left side activation only at C7 right Subject showed activation Subject

1 2 3 4 5 6 7 8 9 10 11 12 Total

C4

C5

C6

C7

T1

Left

Right

Left

Right

Left

Right

Left

Right

Left

Right

0 0 2.2 0.8 0 0 0 1 2.33 0.5 0.6 1.5 8.93

0 0 0.6 0.8 0 0.2 0 1 0.5 0 0.2 0 3.3

0.5 0.2 0.6 1.6 0 0 0 0 0.66 0.83 0.2 1 5.59

0 0.6 0.8 0.6 0 0 0 0 0.6 1 0 0 3.6

0 0 0.4 0.33 0 0 0 0 0 0.5 0 0.33 1.56

0 0.6 0.4 0.83 0 0 0 0 1.33 1.16 0 0.66 4.98

0 0.16 1.4 2.16 0.33 0 1.83 0 0 0.5 0 1.66 8.04

0 0.16 1.4 2.16 0.33 0 1.83 0 0 0.5 0 1.66 8.04

0 0 2.2 0 0 0 0.2 0.16 0 0.5 0 0.83 3.89

0 0 2.2 0 0 0 0 0.33 0 0 0 1.66 4.19

Neuroradiology (2007) 49:73–81

77

Fig. 3 Averaged responses for each vertebral level. a The highest fMRI signal is seen at C7 (nerve root C8) and the lowest at C6. b Separate values for left and right also showing the highest responses at C7, both left and right. In general activity on the left side is more pronounced except at C6. When considering only the right side, an increase in activation is seen from C4 to C7, which then decreases again to T1. When considering only the left side, the pattern is less consistent. There is a high value at C4, followed by a decrease to C6 and a maximum at C7, and again a lower value at T1

activity within or around the spinal cord. We made use of only transverse images for localizing the fMRI signal in this study. Functional anatomy Fingertapping requires a fine motor movement of the hand in which multiple muscle groups are involved. The most important are the finger flexors, innervated by root C7-C8T1, and the muscles of the thumb, innervated by root C8T1 [23, 24]. The activation in executing the paradigm is therefore centred at nerve root C8, leaving the spinal cord at the level of the inferior vertebral notch of corpus C7, implying that the fMRI signals would be particularly expected at the level of the spinal cord segment C8 corresponding to the vertebral corpus C7.

Fingertapping activates both motor and sensory neurons in the cervical spinal cord, since sensory feedback is inevitably elicited. This sensory feedback consists of proprioceptive activation rather than tactile stimulation [25]. Multiple spinal tracts are involved in executing this complex paradigm. From the first neuron in the motor cortex signals are transmitted to the motor neurons in the spinal cord by different descending tracts, constituting the corticospinal tract. The soma of the second neuron is situated in the ventral horn of the spinal cord at the level where the spinal nerve leaves the cord. Different extrapyramidal tracts (the rubrospinal tract, the vestibulospinal tract and the reticulospinal tract) are involved in upper limb motor function as well, and run down anteriorly and laterally to the grey matter of the ipsilateral cervical spinal cord and also make their synapses at the level of the exiting

78

nerve root. Therefore, we would expect that fMRI signals relate to motor function ipsilaterally in the grey matter. Multiple other tracts transmit the sensory feedback to the brain. The fasciculus cuneatus passes on afferent input from the hand (unconscious sensation of muscle position and tonus) to the nucleus cuneatus in the brain stem, as previously demonstrated by fMRI [26]. This tract ascends immediately to the nucleus cuneatus without synapsing, and thus does not contribute to the fMRI signal in the cervical spinal cord. The lateral spinothalamic tract also transmits proprioceptive impulses to the brain. Its fibres ascend one level before they synapse in the substantia gelatinosa, and then cross and ascend further in the contralateral funiculus lateralis towards the thalamus. Thus we would expect possible fMRI signals in the ipsilateral substantia gelatinosa, one level above the nerve root entry in the spinal cord. The ventral spinothalamic tract transmits inputs of pressure and touch, which probably do not significantly contribute to the fMRI signal as previously mentioned [25]. The dorsal and ventral spinocerebellar tracts particularly pass on proprioceptive information from joints, tendons and muscle spindles. These fibres synapse immediately in the substantia intermedia at the ipsilateral side and then run towards the cerebellum, most on the ipsilateral side but some crossed. Therefore fMRI signals from this pathway would be expected on the ipsilateral side of the substantia intermedia. So we would expect sensory functional fMRI signal ipsilaterally in the grey matter, more specifically in the substantia intermedia and the substantia gelatinosa at the level where the nerve enters the cord and one higher. In conclusion, we would expect fMRI signals in the grey matter only at the ipsilateral side of activation and at the level where the nerve attaches to the cord and one level higher [27]. Comparing results to functional anatomy The distribution of the activation response showed important variations between subjects. Because of this large variation and the small number of subjects it was not possible to demonstrate a significantly different pattern of activation between different levels (1), between left and right side (2) or between levels of the same side (3). Nevertheless, a number of findings are clear. In six subjects the peak of activation was found at vertebral corpus C7. As expected, there was activation at the adjacent levels also, but in five out of the six subjects the maximal activation was found at C7. There were five nonresponders, that is those who showed no activation at C7. Most subjects show activation at different levels of the spinal cord. Remarkable was the high signal response on the left side of C4 in four subjects.

Neuroradiology (2007) 49:73–81

Figure 3 gives a summary of the results. Grouping the results of the 12 subjects, it is very clear that the peak response was situated at C7, the level of nerve root C8. Taking into account only the right side, the highest activation was at C7, followed by the adjacent levels C6 and T1, and the least activation was at C5 and finally C4. This correlates completely with the functional anatomy: most nerve fibres involved in the paradigm of fingertapping leave the spinal cord at nerve root C8 (at the level of the vertebral corpus of C7). Some of the nerve fibres come from the adjacent nerve roots, particularly C7 and T1 (at the levels of corpora C6 and T1). On the left side the pattern of activation is more difficult to explain. On the left side as well there was a high peak of response at C7 with less activation at the adjacent levels, and at C4 there was a second lower peak. This peak was found in four subjects, which means that it was not due to one extremely different result. Signal was found on both sides in most individuals. There were signals on the contralateral side of the task as well, not correlating with the functional anatomy. In most subjects there were also paradigm-correlated responses outside the spinal cord. Both of these findings can probably be attributed to the signal changes in large draining veins and the anatomy of these veins around the spinal cord. Technical challenges As already mentioned, the number of publications on fMRI of the spinal cord is small, suggesting significant technical challenges in applying the BOLD technique in the spinal cord. The most substantial problems are caused by limited spatial resolution, periodic movements and haemodynamic washout. The cross-sectional diameter of the human spinal cord (<10 mm) represents a major technical challenge with regard to spatial resolution using an 1.5-T imaging system. The fMRI signal has an effective spatial resolution of only 3 mm. Moreover, MRI of the cervical spine itself is subject to considerable local susceptibility artefacts. Multiple transitions between bone, air and soft tissue in the neck render the magnetic field inhomogeneous. Despite motion correction algorithms in BOLD MRI, regions of false activation can arise near these interfaces [28]. They are caused by the pulmonary apices, the blood flow in the carotid artery, air in the trachea and motion of the larynx and spinal bone. In the cervical spinal cord this source of artefacts is more important than in the brain. Therefore the fMRI signal in the cervical spinal cord is limited to providing an approximate localization of neuronal activity. The flow and pulsations of cerebrospinal fluid around the spinal cord are an unlikely cause of signal changes [5]. On the other hand, the cervical spinal cord itself oscillates in a craniocaudal direction after each cardiac systole [29].

Neuroradiology (2007) 49:73–81

This means that capillary networks make this oscillatory motion as well, which causes fluctuations of blood flow and oxygenation. Yet these fluctuations occur at a higher rate than the slower activity-related fMRI signals, and they are averaged out, therefore decreasing but not disturbing the final fMRI signal. Haemodynamic washout is another disturbing variable. Demonstrated changes in blood flow do not exactly correlate with the underlying oxygen demand. Blood flow changes appear in a slightly larger area than only the area of neuronal activity [17, 30]. The organization of the venous system in the cervical spinal cord—a network with many anastomoses—explains the greater variability of venous return compared to the brain. The BOLD effect is based on an oxygen level increase mainly in small venous structures, i.e. postcapillary venules. The draining venules of the spinal cord are arranged in a radial pattern in an axial plane until they reach the surface of the spinal cord. At the surface there is an extensive venous network with multiple collaterals extending both up and down. The major part of the venous blood is drained into large anterior and posterior midline veins, whereas the sulcal veins have smaller diameters [31]. This may produce significant artefactual fMRI signals. Discussion of the results Variation between subjects The results varied greatly between different subjects. This may have been due to individual variations in anatomical and/or functional organization of the spinal cord and differences in vascular anatomy. The presence of this individual variation has already been demonstrated previously [6, 30]. Furthermore, additional movements, unconsciously made by the volunteers, may have contributed the to variation in the results.

79

request to make no other movements. The fingertapping paradigm could for example elicit an additional slight flexion of the wrist or even an activation of certain muscles without movement. Limited movements of 1 mm or less or muscle activation without movement, unconsciously made by the subject and only moderately correlated with the paradigm, can cause artefacts [32]. Activation ipsi- and contralaterally, dorsal and ventral horns We did not succeed in demonstrating the expected difference in activation between the ipsi- and contralateral sides. This has also been a problem in previous studies [5, 6], although Stroman and Ryner were able to show lateralization [10]. Furthermore, it was not possible to differentiate between the localizations of the nerve tracts mentioned above or between dorsal and ventral horns within one level. All these limitations were due to the limited spatial resolution of fMRI and the organization of the venous system with its multiple anastomoses. Activation outside the spinal cord fMRI signals were observed outside the spinal cord as well. Despite motion correction there were remaining motionrelated active voxels, which possible arose from the large veins on the spinal cord’s surface. Motion correction consisted of time-based rotation and translation of all images. This corrected gross spinal cord (or patient) movement but could not correct for fast CSF flow or pulse-related spinal cord movement. Madi et al. demonstrated that fMRI signals outside the spinal cord are present in all tasks, although much less in an isometric task [8]. This finding suggests that gross motion contributes to this activation as well. Absence of activation

Activation at multiple levels The observation that regions of activation during the fingertapping task are situated at multiple levels in the cervical spinal cord, some of them different from the expected innervation pattern, merits explanation. Firstly, as described above, all muscle groups are innervated by at least two adjacent spinal nerve roots. Secondly, fingertapping is a very complex paradigm involving agonists as well as antagonists and accessory muscles. Thirdly, complex afferent sensory input is transmitted through the spinal cord during activation, which makes synapses at different levels. Fourthly, muscle innervation shows individual variation. Finally, there is the possibility of variation in executing the task, despite fixation of the right arm and the

Why activation was absent at the level of C7 in a few subjects, despite correct execution of the task, remains unclear. This finding currently limits the clinical use of the technique because lack of activation in these subjects cannot be considered as neurological dysfunction. This problem has also been described by Komisaruk et al. for the lower brain stem [26]. Activation on the left side of C4 The high signal response on the left side of C4 in different subjects is remarkable and cannot easily be explained on functional anatomical grounds. The motor nerve root C5 innervates the ipsilateral elbow flexors. The probability that

80

four subjects unknowingly flexed their left elbow during right-handed execution of the task is low. Moreover, we were able to observe the subjects executing the task and we did not notice any additional movements. The sensory nerve root C5 conducts afferent signals from the dermatomes of the elbow fold and thumb. It is possible that the activation was caused by sensory activation, although this seems unlikely. Stracke et al. reported a similar finding in their fMRI investigation with somatosensory stimuli. They almost invariably observed additional activation at a certain level in the spinal cord, independent of which dermatome was stimulated. This activation was situated at segments C3-C4 and they explained it by the existence of an interneuronal system at this spinal level [14]. It is not clear whether or not this mechanism played a role in the activation we observed. Recent developments In our study we evaluated the potential of a ‘classic’ BOLD EPI technique in the visualization of spinal cord function. We did so because we believe this technique has shown its merit in fMRI of the brain. The BOLD EPI sequences are readily available on imagers of most manufacturers. We did not focus on alternative methods since not all of them are readily available. Stroman et al. have been able to demonstrate a nonBOLD contribution to the observed signal changes in fMRI of the spinal cord. They assume that this contribution is caused by an augmentation of local proton density, due to increased exudation of water out of the capillaries when blood flow is increased during neuronal activation. They call this effect “signal enhancement by extravascular protons” (SEEP) [33, 34]. SEEP has been successfully advocated as a viable alternative for spinal fMRI. However, we had problems with this technique on our equipment. Clinical applications Validation of a new technique is important before using it as a tool in clinical practice. Spinal fMRI is still relatively new. We tried to establish whether BOLD EPI fMRI in the spinal cord is possible. At the time of this report the results of clinical studies with spinal fMRI were being presented (e.g. “Functional MR imaging studies of pain in the spinal cord and brain stem” by P.W. Stroman, Queen’s University Kingston, Ontario, Canada at the ASNR in San Diego, May 2006). Conclusion Functional MRI allows us to study the haemodynamic changes of the spinal cord which result from motor

Neuroradiology (2007) 49:73–81

paradigms and sensory triggers and thus to map the neuronal activity of the spinal cord and spinal nerves. We demonstrated that the effect of right-handed fingertapping can be imaged by the fMRI BOLD technique at 1.5 T. Transverse images showed that the regions of activation were localized in spinal cord segments C5 to T1 with a peak at C8, consistent with the known functional neuroanatomy. Acknowledgements We are most grateful to Prof. Gert Verpooten for reviewing the statistical analysis. Conflict of interest statement of interest.

We declare that we have no conflict

References 1. Forster BB, MacKay AL, Whittall KP, Kiehl KA, Smith AM, Hare RD, Liddle PF (1998) Functional magnetic resonance imaging: the basics of blood-oxygen-level dependent (BOLD) imaging. Can Assoc Radiol J 49:320–329 2. Howseman AM, Bowtell RW (1999) Functional magnetic resonance imaging: imaging techniques and contrast mechanisms. Philos Trans R Soc Lond B Biol Sci 354:1179–1194 3. Ogawa S, Tank DW, Menon R, Ellermann JM, Kim SG, Merkle H, Ugurbil K (1992) Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci USA 89:5951–5955 4. Yetkin FZ, McAuliffe TL, Cox R, Haughton VM (1996) Testretest precision of functional MR in sensory and motor task activation. AJNR Am J Neuroradiol 17:95–98 5. Backes WH, Mess WH, Wilmink JT (2001) Functional MR imaging of the cervical spinal cord by use of median nerve stimulation and fist clenching. AJNR Am J Neuroradiol 22:1854– 1859 6. Yoshizawa T, Nose T, Moore GJ, Sillerud LO (1996) Functional magnetic resonance imaging of motor activation in the human cervical spinal cord. Neuroimage 4(3 Pt 1):174–182 7. Stroman PW, Nance PW, Ryner LN (1999) BOLD MRI of the human cervical spinal cord at 3 tesla. Magn Reson Med 42:571– 576 8. Madi S, Flanders AE, Vinitski S, Herbison GJ, Nissanov J (2001) Functional MR imaging of the human cervical spinal cord. AJNR Am J Neuroradiol 22:1768–1774 9. Wilmink JT, Backes WH, Mess WH (2003) Functional MRI of the spinal cord: will it solve the puzzle of pain? JBR-BTR 86:293–294 10. Stroman PW, Ryner LN (2001) Functional MRI of motor and sensory activation in the human spinal cord. Magn Reson Imaging 19:27–32 11. Stroman PW, Tomanek B, Krause V, Frankenstein UN, Malisza KL (2002) Mapping of neuronal function in healthy and injured human spinal cord with spinal fMRI. Neuroimage 17(4):1854– 1860 12. Stroman PW, Krause V, Malisza KL, Frankenstein UN, Tomanek B (2002) Functional magnetic resonance imaging of the human cervical spinal cord with stimulation of different sensory dermatomes. Magn Reson Imaging 20:1–6 13. Porszasz R, Beckmann N, Bruttel K, Urban L, Rudin M (1997) Signal changes in the spinal cord of the rat after injection of formalin into the hindpaw: characterization using functional magnetic resonance imaging. Proc Natl Acad Sci USA 94:5034– 5039

Neuroradiology (2007) 49:73–81 14. Stracke CP, Pettersson LG, Schoth F, Moller-Hartmann W, Krings T (2005) Interneuronal systems of the cervical spinal cord assessed with BOLD imaging at 1.5 T. Neuroradiology 47:127–133 15. Hesselmann V, Zaro Weber O, Wedekind C, Krings T, Schulte O, Kugel H, Krug B, Klug N, Lackner KJ (2001) Age related signal decrease in functional magnetic resonance imaging during motor stimulation in humans. Neurosci Lett 308:141–144 16. Carusone LM, Srinivasan J, Gitelman DR, Mesulam MM, Parrish TB (2002) Hemodynamic response changes in cerebrovascular disease: implications for functional MR imaging. AJNR Am J Neuroradiol 23:1222–1228 17. Yousry I, Naidich TP, Yousry TA (2001) Functional magnetic resonance imaging. Factors modulating the cortical activation pattern of the motor system. Neuroimaging Clin North Am 11:195–202 18. Mohamed FB, Tracy JI, Faro SH, Emperado J, Koenigsberg R, Pinus A, Tsai FY (2000) Investigation of alternating and continuous experimental task designs during single finger opposition fMRI: a comparative study. J Comput Assist Tomogr 24:935–941 19. Song AW, Mao H, Muthupillai R, Haist F, Dixon WT (1999) Segmented spin-echo pulses to increase fMRI signal: repeated intrinsic diffusional enhancement. Magn Reson Med 42:631–635 20. Norris DG, Zysset S, Mildner T, Wiggins CJ (2002) An investigation of the value of spin-echo-based fMRI using a Stroop color-word matching task and EPI at 3 T. Neuroimage 15:719–726 21. Bosch V (2000) Statistical analysis of multi-subject fMRI data: assessment of focal activations. J Magn Reson Imaging 11:61–64 22. Hung LK, Zhao X (2003) Relationship of cervical spinal rootlets and the inferior vertebral notch. Clin Orthop 409:131–137 23. Palastanga N, Field D, Soames R (1998) Anatomy and human movement. Structure and function. Butterworth-Heinemann, Oxford, pp 110–132 24. Monkhouse S (2001) Clinical anatomy. A core text with selfassessment. Churchill Livingstone, London, pp 172–179

81 25. Jansma JM, Ramsey NF, Kahn RS (1998) Tactile stimulation during finger opposition does not contribute to 3D fMRI brain activity pattern. Neuroreport 9:501–505 26. Komisaruk BR, Mosier KM, Liu W-C, Criminale C, Zaborszky L, Whipple B, Kalnin A (2002) Functional localization of brainstem and cervical spinal cord nuclei in humans with fMRI. AJNR Am J Neuroradiol 23:609–617 27. Kahle W, Leonhardt H, Platzer W (1999) Sesam Atlas van de anatomie. Deel 3 zenuwstelsel en zintuigen. Bosch & Keuning, Baarn Nederland, pp 44–53 28. Wu DH, Lewin JS, Duerk JL (1997) Inadequacy of motion correction algorithms in functional MRI: role of susceptibilityinduced artifacts. J Magn Reson Imaging 7:365–370 29. Mikulis DJ, Wood ML, Zerdoner OA, Poncelet BP (1994) Oscillatory motion of the normal cervical spinal cord. Radiology 192:117–121 30. Muto N, Shinomiya K, Komori H, Mochida K, Furuya K (1995) Spinal cord monitoring of the ventral funiculus function. Analysis of spinal field potentials after galvanic vestibular stimulation. Spine 20:2429–2434 31. Krings T (2006) Introduction to fMR Imaging of the Spinal Cord. ASNR abstract book, pp 66–67 32. Field AS, Yen YF, Burdette JH, Elster AD (2000) False cerebral activation on BOLD functional MR images: study of lowamplitude motion weakly correlated to stimulus. AJNR Am J Neuroradiol 21:1388–1396 33. Stroman PW, Krause V, Malisza KL, Frankenstein UN, Tomanek B (2002) Extravascular proton-density changes as a non-BOLD component of contrast in fMRI of the human spinal cord. Magn Reson Med 48:122–127 34. Stroman PW, Krause V, Malisza KL, Frankenstein UN, Tomanek B (2001) Characterization of contrast changes in functional MRI of the human spinal cord at 1.5 T. Magn Reson Imaging 19:833–838