+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
Journal of Neuroradiology (2013) xxx, xxx—xxx
Available online at
ScienceDirect www.sciencedirect.com
ORIGINAL ARTICLE
Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors Mehrnaz Jenabi a, Kyung K. Peck a,b, Robert J. Young a,c,d, Nicole Brennan a, Andrei I. Holodny a,c,d,∗ a
Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, 10065 NY, USA b Department of Medical Physics, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, 10065 NY, USA c Brain Tumor Center, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, 10065 NY, USA d Department of Radiology, Weill Medical College of Cornell University, New York, 10065 NY, USA
KEYWORDS DTI; FMRI; Probabilistic fiber tracking; SMA; Brain tumor
Summary Background and purpose: Accurate localization of anatomically and functionally separate SMA tracts is important to improve planning prior to neurosurgery. Using fMRI and probabilistic DTI techniques, we assessed the connectivity between the frontal language area (Broca’s area) and the rostral pre-SMA (language SMA) and caudal SMA proper (motor SMA). Materials and methods: Twenty brain tumor patients completed motor and language fMRI paradigms and DTI. Peaks of functional activity in the language SMA, motor SMA and Broca’s area were used to define seed regions for probabilistic tractography. Results: fMRI and probabilistic tractography identified separate and unique pathways connecting the SMA to Broca’s area — the language SMA pathway and the motor SMA pathway. For all subjects, the language SMA pathway had a larger number of voxels (P < 0.0001) and higher connectivity (P < 0.0001) to Broca’s area than did the motor SMA pathway. In each patient, the number of voxels was greater in the language and motor SMA pathways than in background pathways (P < 0.0001). No differences were found between patients with ipsilateral and those with contralateral tumors for either the language SMA pathway (degree of connectivity: P < 0.36; number of voxels: 0.35) or the motor SMA pathway (degree of connectivity, P < 0.28; number of voxels, P < 0.74).
Abbreviations: BOLD, blood oxygen level-dependent; FA, fractional anisotropy; fMRI, functional MRI; MD, mean diffusivity; ROI, region of interest; SMA, supplementary motor area. ∗ Corresponding author. Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, NY 10021, USA. Tel.: +1 212 639 3182. E-mail address: [email protected] (A.I. Holodny). 0150-9861/$ – see front matter © 2013 Elsevier Masson SAS. All rights reserved. http://dx.doi.org/10.1016/j.neurad.2013.12.001
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
2
M. Jenabi et al. Conclusion: Probabilistic tractography can identify unique white matter tracts that connect language SMA and motor SMA to Broca’s area. The language SMA is more significantly connected to Broca’s area than is the motor subdivision of the SMA proper. © 2013 Elsevier Masson SAS. All rights reserved.
Introduction The supplementary motor area (SMA) comprises of rostral pre-SMA involved in language (language SMA) and a caudal SMA proper involved in motor planning (motor SMA) [1—3]. The locations of these subdivisions of the SMA, however, vary in individual cases. The SMA is relevant to neurosurgical planning, since resection of the SMA can lead to postoperative deficits including mutism in extreme cases [4]. Iatrogenic speech dysfunction may occur in up to half of brain tumor resections near the SMA [5,6], implying that the SMA plays an integral part in the motor execution of speech. Studies in normal subjects have qualitatively described structural connectivity between these subdivisions of the SMA and Broca’s area, using anatomically defined seed regions and probabilistic tractography [7,8]. These prior studies, however, did not quantify nor compare the connectivity of the generated pathways [9,10]. Diffusion tensor imaging (DTI) can non-invasively identify white matter tracts in the human brain [11,12]. In addition, DTI can define the relationships of essential white matter tracts to a brain tumor to optimize the planning and resection of the tumor in patients with brain tumors [4,13—15]. Functional magnetic resonance imaging (fMRI) has confirmed the importance of classical language-related areas, including Broca’s area and Wernicke’s area, as well as secondary language area such as the SMA that contribute to language expression and reception [3,16—18]. Combination of fMRI and DTI tractography by identifying region of interest (ROI) as activation peaks in task-based fMRI data can increase the accuracy of fiber tracking. Identifying the correct ROI for tractography is critical to extract true structural connectivity, especially in brain tumor patients with distortion of normal anatomy. The purpose of our study was to apply probabilistic tractography to identify and quantify the fiber connectivity between the 2 SMA subdivisions and Broca’s area. Based on the voxels activated by the fMRI language and motor paradigms, seed ROIs for tracking were placed in Broca’s area (BA seed), language SMA, motor SMA and 2 background areas. We quantitatively assessed the degrees of connectivity and the numbers of voxels in the pathways from the SMA ROIs to Broca’s area. We hypothesize that there is higher connectivity between Broca’s area and the language SMA than between Broca’s area and the motor SMA, as well as between either subdivision of the SMA and Broca’s area than between the background areas and Broca’s area.
Methods Subjects The Institutional Review Board and Privacy Board approved this retrospective study performed according to Health
Insurance Portability and Accountability Act regulations. Twenty patients (10 males and 10 females, mean age = 43.8 years [range, 28—65]) with unilateral brain tumors in the right or left brain who underwent both fMRI and DTI over an 18-month period were studied. fMRI and DTI were performed in each case because the patient had signs or symptoms of speech difficulty and/or the tumor was located near (≤ 2 cm) the expected language centers and/or tracts. The tumors consisted of astrocytomas (n = 8), oligodendrogliomas (n = 6), glioblastomas (n = 2) and metastases (n = 4). All patients were 100% right-handed as determined by the Edinburgh Handedness Inventory [19] and were native English speakers.
MRI The fMRI and DTI data were acquired as part of the routine clinical workup for each patient. Scanning was performed on a 3T magnet (3T GE, Milwaukee, Wisconsin) using a standard quadrature head coil. BOLD fMRI was acquired with a single-shot gradient-echo echo-planar imaging sequence (TR/TE = 4000/35 ms; 128 × 128 matrix; 4.5-mm thickness with no gap, 240-mm FOV). DTI data were acquired using a single-shot spin-echo echo-planar imaging sequence (25 directions, TR/TE = 11,000/64 ms, 128 × 128 matrix; 3mm thickness, 1000 s/mm2 b value). T1-weighted spin-echo images (TR/TE = 600/8 ms, 256 × 256 matrix, 4.5-mm thickness with no gap, 240-mm FOV) and 3D T1-weighted anatomic images with a spoiled gradient-recalled-echo sequence (TR 22 ms, TE 4 ms, 256 × 256 matrix, 30◦ flip angle, 1.5-mm thickness, 240-mm FOV) were also acquired.
Functional tasks All patients underwent 2 paradigms: a motor task (finger tapping) and a language task (phonemic fluency). Each paradigm was presented as a block paradigm consisting of 90 images, alternating between 20 seconds of activation and 40 seconds for both hands in response to an auditory cue. Self-paced finger tapping task was used at approximately 2 Hz by sequential finger tapping of fingers 2-5 against finger 1 to localize the motor SMA and the primary motor cortex. The language task was used to localize the language SMA and Broca’s area. During the task, letters appeared on the screen and patients were asked to silently generate words that began with that letter. Each subject’s real-time brain activity and head motion were monitored using software (Brainwave, GE Healthcare, Milwaukee, Wisconsin).
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
Probabilistic fiber tracking of the language and motor SMA networks
3
Broca’s area were spurious, we defined 2 additional ROIs (defined as background ROIs) adjacent to the SMA. An anterior background ROI was placed 5-7 voxels anterior to the language SMA on an axial image at the level of the SMA; and a posterior background ROI was placed 5-7 voxels posterior to the motor SMA (Fig. 1).
Probabilistic tractography
Figure 1 The DTI seed ROIs. The language supplementary motor area (SMA), motor SMA and Broca’s Area are obtained from the language and motor fMRI tasks. Activated pixels in the fMRI map represent the voxel with the highest activation strength and they are used as seed ROIs for DTI tractography. Background ROIs are placed anterior to the language SMA and posterior to the motor SMA.
fMRI analysis Image analysis was performed using Analysis of Functional Neuroimaging (AFNI) [20]. Head motion correction was performed using 3D rigid-body registration and spatial smoothing (Gaussian filter with 4-mm full width of half maximum). Corrections for linear trend and high frequency noise were applied when necessary. Functional activity was generated using a correlation analysis. A modeled waveform corresponding to the task performance block was crosscorrelated with all pixel time courses on a pixel-by-pixel basis to identify stimulus locked responses. A threshold of P < 0.001 was used. To reduce false positive activity from large venous structures or head motion, voxels in which the standard deviation of the acquired time series exceeded 8% of the mean signal intensity were set to zero.
Identification of ROIs for tractography Activation maxima obtained from the fMRI were used to determine seed ROIs for DTI fiber tracking. Based on the voxels activated by the fMRI language and motor paradigms, SMA was segregated into 2 functional regions: language and motor SMA. We defined language SMA as the rostral part of the SMA activated during the language task and motor SMA as the caudal part of the SMA that was activated during the motor task. Voxels showing peak functional activity in the language SMA and the motor SMA were selected as the centers of the seed ROIs for fiber tracking in the language SMA and in the motor SMA, respectively. The voxel showing peak functional activity in Broca’s area was chosen as the center of the seed ROI in Broca’s area. All ROIs for tractography were manually drawn on the b0 images based on the activation maxima of the same patients (Fig. 1). To determine whether the white matter connections between the areas of fMRI-defined SMA activation and
DTI data were analyzed using DTI & FiberTools (Department of Diagnostic Radiology, University Hospital, Freiburg, Germany), implemented in MATLAB (Mathworks, Inc., Natick, Massachusetts). Initially, head motion and eddy current issues were corrected if necessary. In DTI & FiberTools, an extended Monte Carlo simulation method similar to the Probabilistic Index of Connectivity (PICO) method [11,12] was used to capture the directional information passing a voxel. To extract the probabilistic map of connectivity between 2 seed regions, a probabilistic map was first generated for each seed ROI; then the maps from the 2 ROIs were combined to compute the probability that each voxel was part of the bundle connecting both seed points [12,21]. Probabilistic maps were generated from individual ROIs for each patient. Tracking areas were defined as regions with mean diffusivity (MD) < 0.002 and fractional anisotropy (FA) > 0.1. The number of random walks was set to 100,000 and the maximum fiber length to 150 voxels (default) [12]. Probabilistic maps generated from the language SMA and Broca’s area ROIs were combined to produce their probabilistic map of connectivity, called the language SMA pathway. Probabilistic maps generated from the motor SMA and Broca’s area ROIs were combined to create the motor SMA pathway. In addition, probabilistic maps generated from each of the background ROIs and Broca’s area ROIs were combined to create background pathways.
Statistical analysis The probabilistic map of connectivity represents the voxels with the highest likelihood of connecting 2 selected ROIs. To compare these maps of each pathway across subjects, we counted all possible connected voxels within the language SMA pathway, as well as the motor SMA pathway of each subject. Only voxels with endpoints in the ROIs were retained as part of the pathway. These voxels were extracted by generating histograms of the connectivity between the seed regions and applying a connectivity threshold to define the areas of highest connectivity. In addition, for both the motor SMA and the language SMA pathways in individual subjects, the mean degree of connectivity and the voxels with the highest connectivity around Broca’s area were measured. Using paired t test, the numbers of voxels, their degrees of connectivity and the degrees of connectivity around Broca’s area were compared across subjects, in the pathways, between pathways and the 2 background pathways, and between the group of patients whose tumors were ipsilateral or contralateral to the pathways.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model
ARTICLE IN PRESS
NEURAD-445; No. of Pages 8
4
M. Jenabi et al.
Figure 2 The probabilistic fiber tracking results of 2 patients. a and b show the language supplementary motor area (SMA) and the motor SMA respectively for a single subject with a tumor in the non-language-dominant hemisphere. c and d show the language SMA and the motor SMA respectively for a different subject with a tumor in the language-dominant hemisphere. Tracks are superimposed on b0 images in the axial and coronal planes at the levels of SMA and BA. Colors represent the degree of connectivity: from 0.00001 in blue to 0.1 in red.
Results In all 20 subjects (including 11 with left- and 9 with righthemisphere tumors), the activated voxels with the most significant P values during the fMRI tasks were located in the left hemisphere. In each subject, fMRI showed 2 separate areas of activation in the SMA: a rostral area within the pre-SMA activated during the language task (defined as the language SMA), and a caudal region within the SMA proper activated during the motor task (defined as the motor SMA). For tractography, the seed ROIs of the language SMA, motor SMA and Broca’s area were in the areas of dominant activation, which were on the left side in all patients. The fMRI determined left hemisphere dominance for language was confirmed by intraoperative electrical stimulation in all patients.
Language SMA and motor SMA pathways The language SMA and motor SMA pathways to Broca’s area were observed as separate and unique in all patients. The language SMA pathway consistently traveled anterior to the motor SMA pathway. Fig. 2 illustrates the probabilistic fiber tracking results of 2 patients with dominant left SMAs: one with a tumor in the non-dominant and another with a tumor in the dominant hemisphere. Fig. 3 shows the volume renderings of both pathways for a representative subject. Regardless of the location or laterality of the tumor, the language SMA and motor SMA pathways traveled in separate
Table 1
Figure 3 Volume rendering image of supplementary motor area (SMA) pathways. It is showing the projections of both language and motor SMA pathways of a single subject from ROIs. Red: the language SMA pathway; blue: the motor SMA pathway; green: the language SMA seed; yellow: the motor SMA seed, magenta: the Broca’s area seed.
but parallel routes from the SMA to the frontal language area. In all subjects, the mean number of voxels in the language SMA pathway (510) was greater than that in the motor SMA pathway (278), (P < 0.0001) (Table 1). For all subjects, the degree of connectivity was stronger for the language SMA pathway (0.12) than for the motor SMA pathway (0.07) (P < 0.001). In addition, the degree of connectivity around Broca’s area was stronger for the language SMA pathway (0.4) than for the motor SMA pathway (0.16) (P < 0.001). The number of voxels and the degree of connectivity of the language SMA pathway were greater than those of the motor SMA pathway within every subject, although they varied among subjects (Table 1).
Background ROIs In each patient, the number of voxels was greater in the language and motor SMA pathways than in either the anterior (P < 0.0001) or posterior background pathways (P < 0.0001) (Table 2), (Fig. 4). Mean numbers of voxels in the language and motor SMA pathways were higher than those in both background pathways (P < 0.0001).
Side of tumor Probabilistic connectivity maps of the language SMA pathways did not differ when stratified as ipsilateral or contralateral to the tumor (mean number of voxels: 474 vs. 555, respectively, with P < 0.35; mean degree of connectivity: 0.1 vs. 0.3, respectively, with P < 0.76; Table 3). Similarly, no differences were found between motor SMA pathways that were ipsilateral to the tumor and those that were contralateral to the tumor (mean number of voxels:
Comparison of the language and motor supplementary motor area (SMA) pathways.
Language SMA pathway Motor SMA pathway P value
Mean number of voxels
Mean degree of connectivity
Degree of connectivity at BA
510 278 < 0.0001
0.1 0.0350 < 0.0001
0.388 0.160 < 0.001
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model
ARTICLE IN PRESS
NEURAD-445; No. of Pages 8
Probabilistic fiber tracking of the language and motor SMA networks
5
Table 2 Comparison of mean number of voxels and degree of connectivity of probabilistic maps between language and motor supplementary motor area (SMA) pathways and background SMA pathways. Background SMA
Anterior pathway Posterior pathway
Mean number of voxels
P value Language SMA pathway
Motor SMA pathway
159 105
< 0.0001 < 0.0001
< 0.001 0.003
Mean degree of connectivity
0.05 0.03
P value Language SMA pathway < 0.0001 < 0.0001
Discussion In the present study, we investigated the existence, positions and relationships of fiber networks between the SMA and Broca’s area using fMRI and probabilistic tractography. Our results indicated that the language SMA is more robustly connected to Broca’s area than is the motor SMA as reflected by a higher mean number of activated voxels (P < 0.0001) and a greater degree of connectivity (P < 0.0001). These results have implications for preoperative planning in patients with brain tumors.
Language and motor pathways
Figure 4 Comparison of language supplementary motor area (SMA) and motor SMA pathways with anterior and posterior background pathways: (X-axis: each patient; y-axis: the number of connected voxels). The mean number of voxels is indicated inside the box plot of each group. Outside value is shown by (*).
266 vs. 293, respectively, with P < 0.74; mean degree of connectivity: 0.05 vs. 0.09, respectively, with P < 0.59). Regardless of the side of the tumor, the SMA pathways were always on the left side, and the probabilistic connectivity of the language SMA pathway was always greater than that of the motor SMA pathway. In the 11 patients with ipsilateral left tumors (in respect to the location of ROI), the mean number of activated voxels (P < 0.001) and the mean degree of connectivity (P < 0.01) were higher in the language SMA pathway than in the motor SMA pathway (Table 3). In the 9 patients with contralateral right tumors, the mean number of activated voxels (P < 0.002) and the mean degree of connectivity (P < 0.02) were also higher in the language SMA pathway than in the motor SMA pathway.
Table 3
We demonstrated separate language and motor white matter pathways starting in regionally specific areas of the SMA that coursed toward the ipsilateral frontal language area via distinct, semi-parallel pathways. The structural and functional connectivity between the SMA and Broca’s area have been discussed in a number of papers [7,8,17,22,23] showed potential connectivity between Broca’s area and the SMA, using anatomically based ROIs and deterministic or adaptive tractography for fiber tracking in normal subjects. Another study [8] used probabilistic tractography to show connections between Broca’s area and both the pre-SMA and SMA proper. The investigators used anatomical landmarks to define these regions in healthy adults. Our results agree with their findings and support the existence of anatomical and functional connectivity between two SMA subregions and the frontal language area. Unlike previous anatomy-based approaches [8,24], our results benefit from using fMRI specific activation to define the ROIs for probabilistic tracking. fMRI-driven tractography has been described as more accurate in patients with brain tumors [25—27]. We showed that the different parts of the SMA responsible for specific tasks have differing probabilities of connectivity to Broca’s area. Although the
Statistical results of 2 pathways in patients with an ipsilateral and contralateral tumor. Ipsilateral
Language SMA pathway Motor SMA pathway P value
Contralateral
Mean number of voxels
Mean degree of connectivity
Mean number of voxels
474 266 < 0.0001
0.100 0.050 0.01
555 293 0.0018
SMA: supplementary motor area.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
6
M. Jenabi et al.
connectivity among these regions has been described before [7,8], the degree of connectivities of the pathways have not previously been quantified. We selected 3 measures of connectivity for comparison: the number of voxels with the highest connectivity value near Broca’s area, the mean value of connectivity along each pathway and the number of voxels within each pathway. 3D tractography renderings were also generated to visualize the spatial orientation of the pathways to each other: in each case, the language SMA pathway contained more voxels than the motor SMA pathway, as demonstrated by a larger surface area (radius), especially around Broca’s area; the lengths of the 2 pathways were similar (Fig. 3). These differences occurred despite constant seed ROI sizes, suggesting that the probabilistic results reflect true differences in connectivity between the language SMA and motor SMA pathways. Since the probabilistic maps representing the language and motor SMA pathways showed that the voxels with the highest connectivity value were located near the frontal language area, it is reasonable to conclude that the connectivity to Broca’s area from the language component of SMA was consistently more robust than the connection to Broca’s area from the motor component of SMA. Traditional brain mapping techniques have relied on post-mortem fiber dissections, indirect evaluation of lesion locations and clinical deficits in neurologically impaired patients, and tracer studies in primate models. The advent of non-invasive tools such as DTI has allowed the in vivo evaluation of anatomical and functional pathways. These tools have been used to describe complex networks that involved in specific task performances, and have rapidly expanded our understanding of supportive language networks such as the SMA pathways. For example, the exact location of the cortico-spinal tract was matter of debate based on clinical material and anatomical dissection. This anatomical question was settled convincingly by tractography [13].
Structural and functional SMA Studies suggested important roles of the pre-SMA and SMA proper in different aspects of word production and motor control [18,22,23,28]. Interactions between different regions within the language SMA and the frontal language areas in many aspects of language, including word production, have been reported previously [8,9,18,29]. We showed that the language SMA pathway implies connectivity of part of the language SMA to the Broca’s area. The motor SMA pathway revealed connectivity of the motor SMA to Broca’s area. Given that patients with postoperative SMA syndromes exhibit a paucity of speech and have a marked deficit in initiation of speech in particular, it is not surprising that motoric components of planning are directly connected to Broca’s area [5,29—31]. The background pathways showed lower connectivity with Broca’s area than did the language and motor SMA pathways. Since connectivity was determined from seed ROIs placed based on fMRI-activated voxels during simple language and motor tasks, it is reasonable to assume that additional regions of functional activity might be distinguished within the pre-SMA and SMA proper if other fMRI
paradigms were applied. Other studies have reached similar conclusions [18,22]. The differences we found between the connectivity parameters of the language SMA and motor SMA pathways and those of the background pathways further support the work of Peck, who described the structural and functional relations between the language SMA and motor SMA and identified joint activation in the border between these two areas during a simultaneous motor and language task [3].
Limitations Data described in the study were acquired from MRI examinations performed for standard clinical care. Hence, it is possible that microscopic tumor infiltration and/or edema could have affected diffusion measurements, particularly for gliomas [26,27,32,33]. A second potential limitation relates to the use of the probabilistic method for tractography. Probabilistic diffusion tractography is a non-invasive tool for predicting and extracting the white matter network and it relies on several non-anatomical assumptions to measure the neural connectivity. However, the advantages of the probabilistic over deterministic fiber tracking methods are especially apparent in cases where the target tracts involve crossing fibers [11,34,35]. The probabilistic method we used belongs to a method of pathway extraction [12] that determines the most probable neural pathways between two regions by identifying point-to-point connections without any a priori information. This method has been validated in different tracts that cross other fiber bundles several times, i.e., optic radiation tracts, and the lateral part of the cortico-spinal tract.
Clinical relevance While injury to both gray and white matter structures may cause devastating neurologic deficits, at surgery the white matter is less accessible (deeper), more difficult to identify by anatomic landmarks, and more susceptible to false negative results at stimulation. Iatrogenic injury to the pre-SMA (language) or the SMA proper (motor) areas or the associated SMA pathways can result in an SMA syndrome that can produce paresis, dysphasia and even mutism depending upon the hemisphere and extent of injury [5,6]. Interestingly, the majority of these functions return weeks to months following the inception of the deficit but permanent injury has been described [5,6]. The fact that these functions usually return offers an interesting opportunity to use diffusion tractography to measure dynamic changes in the cognitive/sensory pathways over time. It is estimated that as many as 26% of superior frontal glioma resections result in SMA syndromes to varying degrees. DTI maps are ever increasing in their behavioral specificity allowing the correlation between DTI measured pathways and expected outcomes [36]. Further, Fortin et al. discuss the utility of DTI maps in neurosurgery saying ‘‘Complex neurocognitive functions are not attributed to a single brain area but depend on the dynamic interactions of distributed brain areas operating in large-scale networks [37]’’. This is important in the field of neurosurgery where
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
Probabilistic fiber tracking of the language and motor SMA networks intervention within a spatially localized area may indirectly lead to unwanted effects on distant areas.
Conclusions We used fMRI-driven probabilistic fiber tracking to extract the most probable tracts and to quantify the patterns of connectivity between two functionally distinct regions of the SMA and the Broca’s area in brain tumor patients. The connectivity between Broca’s area and the pre-SMA (language SMA) was stronger than the connectivity between Broca’s area and the SMA proper (motor SMA).
Disclosure of interest The authors declare that they have no conflicts of interest concerning this article.
Acknowledgments The authors thank Ms. Ada Muellner and Katharine Ristich for her expert editorial advice.
References [1] Picard N, Strick PL. Motor areas of the medial wall: a review of their location and functional activation. Cereb Cortex 1996;6:342—53. [2] Nachev P, Kennard C, Husain M. Functional role of the supplementary and pre-supplementary motor areas. Nat Rev Neurosci 2008;9:856—69. [3] Peck KK, Bradbury M, Psaty EL, et al. Joint activation of the supplementary motor area and presupplementary motor area during simultaneous motor and language functional MRI. Neuroreport 2009;20:487—91. [4] Gupta A, Shah A, Young RJ, et al. Imaging of brain tumors: functional magnetic resonance imaging and diffusion tensor imaging. Neuroimaging Clin N Am 2010;20: 379—400. [5] Krainik A, Lehéricy S, Duffau H, et al. Postoperative speech disorder after medial frontal surgery: role of the supplementary motor area. Neurology 2003;60: 587—94. [6] Rosenberg K, Nossek E, Liebling R, et al. Prediction of neurological deficits and recovery after surgery in the supplementary motor area: a prospective study in 26 patients. J Neurosurg 2010;113:1152—63. [7] Morgan VL, Mishra A, Newton AT, et al. Integrating functional and diffusion magnetic resonance imaging for analysis of structure-function relationship in the human language network. PLoS ONE 2009;4:e6660. [8] Ford A, McGregor KM, Case K, et al. Structural connectivity of Broca’s area and medial frontal cortex. Neuroimage 2010;52:1230—7. [9] Klein JC, Behrens TEJ, Robson MD, et al. Connectivitybased parcellation of human cortex using diffusion MRI: established reproducibility, validity and observer independence in BA 44/45 and SMA/pre-SMA. Neuroimage 2007;34: 204—11. [10] Maddah M, Wells WM, Warfield SK, et al. Probabilistic clustering and quantitative analysis of white matter fiber tracts. Inf Process Med Imaging 2007;20:372—83.
7
[11] Parker JG, Haroon HA, Wheeler-Kingshott CA. A framework for a streamline-based probabilistic index of connectivity (PICo) using a structural interpretation of MRI diffusion measurements. J Magn Reson Imaging 2003;18: 242—54. [12] Kreher BW, Schnell S, Mader I, Il’yasov KA, et al. Connecting and merging fibres: pathway extraction by combining probability maps. Neuroimage 2008;43:81—9. [13] Holodny AI, Watts R, Korneinko VN, Pronin IN, Zhukovskiy ME, Gor DM, et al. Diffusion tensor tractography of the motor white matter tracts in man: current controversies and future directions. Ann N Y Acad Sci 2005;1064: 88—97. [14] Sinha S, Bastin ME, Whittle IR, et al. Diffusion tensor MR imaging of high-grade cerebral gliomas. AJNR Am J Neuroradiol 2002;23:520—7. [15] Nimsky C, Ganslandt O, Hastreiter P, et al. Intraoperative diffusion-tensor MR imaging: shifting of white matter tracts during neurosurgical procedures–initial experience. Radiology 2005;234:218—25. [16] Price C. The anatomy of language: contributions from functional neuroimaging. J Anat 2000;197:335—59. [17] Koechlin E, Jubault T. Broca’s area and the hierarchical organization of human behavior. Neuron 2006;50: 963—74. [18] Tremblay P, Gracco VL. Contribution of the pre-SMA to the production of words and non-speech oral motor gestures, as revealed by repetitive transcranial magnetic stimulation (rTMS). Brain Res 2009;1268:112—24. [19] Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 1971;9: 97—113. [20] Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res 1996;29:162—73. [21] Saur D, Kreher BW, Schnell S, et al. Ventral and dorsal pathways for language. Proc Natl Acad Sci U S A 2008;105: 18035—40. [22] Meister B, Boroojerdi H, Foltys R, et al. Motor cortex hand area and speech: implications for the development of language. Neuropsychologia 2003;41:401—6. [23] de Lafuente V, Romo R. Language abilities of motor cortex. Neuron 2004;41:178—80. [24] Johansen-Berg H, Behrens TEJ, Robson MD, et al. Changes in connectivity profiles define functionally distinct regions in human medial frontal cortex. Proc Natl Acad Sci U S A 2004;101:13335—40. [25] Schonberg T, Pianka P, Hendler T, et al. Characterization of displaced white matter by brain tumors using combined DTI and fMRI. Neuroimage 2006;30(4):1100—11. [26] Kamiya K, Sato N, Saito Y, et al. Accelerated myelination along fiber tracts in patients with hemimegalencephaly. J Neuroradiol 2013 [In press]. [27] Zolal A, Hejcl A, Malucelli A, et al. Distant White-matter diffusion change caused by tumor growth. J Neuroradiol 2013;40:71—80. [28] Hoshi E, Tanji J. Differential roles of neuronal activity in the supplementary and presupplementary motor areas: from information retrieval to motor planning and execution. J Neurophysiol 2004;92:3482—99. [29] Alario FX, Chainay H, Lehericy S, et al. The role of the supplementary motor area (SMA) in word production. Brain Res 2006;1076:129—43. [30] Binkofski F, Buccino G. Motor functions of the Broca’s region. Brain Lang 2004;89:362—9. [31] Amunts K, Lenzen M, Friederici AD, et al. Broca’s region: novel organizational principles and multiple receptor mapping. PLoS Biol 2010;8:e1000489.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
8
M. Jenabi et al.
[32] Schreiber A, Hubbe U, Ziyeh S, et al. The influence of gliomas and nonglial space-occupying lesions on blood-oxygen-level -dependent contrast enhancement. AJNR Am J Neuroradiol 2002;21:1055—63. [33] Ulmer JL, Krouwer HG, Mueller WM, et al. Pseudoreorganization of language cortical function at fMR imaging: a consequence of tumor-induced neurovascular uncoupling. AJNR Am J Neuroradiol 2003;24:213—7. [34] Behrens TE, Woolrich MW, Jenkinson M, et al. Characterization and propagation of uncertainty in diffusion-weighted MR imaging. Magn Reson Med 2003;50:1077—88.
[35] Lazar M, Weinstein JS, Tsuruda KM, et al. White matter tractography using diffusion tensor deflection. Hum Brain Mapp 2003;18:306—21. [36] Russell SM, Kelly PJ. Incidence and clinical evolution of postoperative deficits after volumetric stereotactic resection of glial neoplasms involving the supplementary motor area. Neurosurgery 2003:506—16. [37] Fontaine D, Capelle L, Duffau H. Somatotopy of the supplementary motor area: evidence from correlation of the extent of surgical resection with the clinical patterns of deficit. Neurosurgery 2002:297—303.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
ARTICLE IN PRESS
Journal of Neuroradiology (2013) xxx, xxx—xxx
Available online at
ScienceDirect www.sciencedirect.com
ORIGINAL ARTICLE
Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors Mehrnaz Jenabi a, Kyung K. Peck a,b, Robert J. Young a,c,d, Nicole Brennan a, Andrei I. Holodny a,c,d,∗ a
Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, 10065 NY, USA b Department of Medical Physics, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, 10065 NY, USA c Brain Tumor Center, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, 10065 NY, USA d Department of Radiology, Weill Medical College of Cornell University, New York, 10065 NY, USA
KEYWORDS DTI; FMRI; Probabilistic fiber tracking; SMA; Brain tumor
Summary Background and purpose: Accurate localization of anatomically and functionally separate SMA tracts is important to improve planning prior to neurosurgery. Using fMRI and probabilistic DTI techniques, we assessed the connectivity between the frontal language area (Broca’s area) and the rostral pre-SMA (language SMA) and caudal SMA proper (motor SMA). Materials and methods: Twenty brain tumor patients completed motor and language fMRI paradigms and DTI. Peaks of functional activity in the language SMA, motor SMA and Broca’s area were used to define seed regions for probabilistic tractography. Results: fMRI and probabilistic tractography identified separate and unique pathways connecting the SMA to Broca’s area — the language SMA pathway and the motor SMA pathway. For all subjects, the language SMA pathway had a larger number of voxels (P < 0.0001) and higher connectivity (P < 0.0001) to Broca’s area than did the motor SMA pathway. In each patient, the number of voxels was greater in the language and motor SMA pathways than in background pathways (P < 0.0001). No differences were found between patients with ipsilateral and those with contralateral tumors for either the language SMA pathway (degree of connectivity: P < 0.36; number of voxels: 0.35) or the motor SMA pathway (degree of connectivity, P < 0.28; number of voxels, P < 0.74).
Abbreviations: BOLD, blood oxygen level-dependent; FA, fractional anisotropy; fMRI, functional MRI; MD, mean diffusivity; ROI, region of interest; SMA, supplementary motor area. ∗ Corresponding author. Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275, York avenue, New York, NY 10021, USA. Tel.: +1 212 639 3182. E-mail address: [email protected] (A.I. Holodny). 0150-9861/$ – see front matter © 2013 Elsevier Masson SAS. All rights reserved. http://dx.doi.org/10.1016/j.neurad.2013.12.001
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
2
M. Jenabi et al. Conclusion: Probabilistic tractography can identify unique white matter tracts that connect language SMA and motor SMA to Broca’s area. The language SMA is more significantly connected to Broca’s area than is the motor subdivision of the SMA proper. © 2013 Elsevier Masson SAS. All rights reserved.
Introduction The supplementary motor area (SMA) comprises of rostral pre-SMA involved in language (language SMA) and a caudal SMA proper involved in motor planning (motor SMA) [1—3]. The locations of these subdivisions of the SMA, however, vary in individual cases. The SMA is relevant to neurosurgical planning, since resection of the SMA can lead to postoperative deficits including mutism in extreme cases [4]. Iatrogenic speech dysfunction may occur in up to half of brain tumor resections near the SMA [5,6], implying that the SMA plays an integral part in the motor execution of speech. Studies in normal subjects have qualitatively described structural connectivity between these subdivisions of the SMA and Broca’s area, using anatomically defined seed regions and probabilistic tractography [7,8]. These prior studies, however, did not quantify nor compare the connectivity of the generated pathways [9,10]. Diffusion tensor imaging (DTI) can non-invasively identify white matter tracts in the human brain [11,12]. In addition, DTI can define the relationships of essential white matter tracts to a brain tumor to optimize the planning and resection of the tumor in patients with brain tumors [4,13—15]. Functional magnetic resonance imaging (fMRI) has confirmed the importance of classical language-related areas, including Broca’s area and Wernicke’s area, as well as secondary language area such as the SMA that contribute to language expression and reception [3,16—18]. Combination of fMRI and DTI tractography by identifying region of interest (ROI) as activation peaks in task-based fMRI data can increase the accuracy of fiber tracking. Identifying the correct ROI for tractography is critical to extract true structural connectivity, especially in brain tumor patients with distortion of normal anatomy. The purpose of our study was to apply probabilistic tractography to identify and quantify the fiber connectivity between the 2 SMA subdivisions and Broca’s area. Based on the voxels activated by the fMRI language and motor paradigms, seed ROIs for tracking were placed in Broca’s area (BA seed), language SMA, motor SMA and 2 background areas. We quantitatively assessed the degrees of connectivity and the numbers of voxels in the pathways from the SMA ROIs to Broca’s area. We hypothesize that there is higher connectivity between Broca’s area and the language SMA than between Broca’s area and the motor SMA, as well as between either subdivision of the SMA and Broca’s area than between the background areas and Broca’s area.
Methods Subjects The Institutional Review Board and Privacy Board approved this retrospective study performed according to Health
Insurance Portability and Accountability Act regulations. Twenty patients (10 males and 10 females, mean age = 43.8 years [range, 28—65]) with unilateral brain tumors in the right or left brain who underwent both fMRI and DTI over an 18-month period were studied. fMRI and DTI were performed in each case because the patient had signs or symptoms of speech difficulty and/or the tumor was located near (≤ 2 cm) the expected language centers and/or tracts. The tumors consisted of astrocytomas (n = 8), oligodendrogliomas (n = 6), glioblastomas (n = 2) and metastases (n = 4). All patients were 100% right-handed as determined by the Edinburgh Handedness Inventory [19] and were native English speakers.
MRI The fMRI and DTI data were acquired as part of the routine clinical workup for each patient. Scanning was performed on a 3T magnet (3T GE, Milwaukee, Wisconsin) using a standard quadrature head coil. BOLD fMRI was acquired with a single-shot gradient-echo echo-planar imaging sequence (TR/TE = 4000/35 ms; 128 × 128 matrix; 4.5-mm thickness with no gap, 240-mm FOV). DTI data were acquired using a single-shot spin-echo echo-planar imaging sequence (25 directions, TR/TE = 11,000/64 ms, 128 × 128 matrix; 3mm thickness, 1000 s/mm2 b value). T1-weighted spin-echo images (TR/TE = 600/8 ms, 256 × 256 matrix, 4.5-mm thickness with no gap, 240-mm FOV) and 3D T1-weighted anatomic images with a spoiled gradient-recalled-echo sequence (TR 22 ms, TE 4 ms, 256 × 256 matrix, 30◦ flip angle, 1.5-mm thickness, 240-mm FOV) were also acquired.
Functional tasks All patients underwent 2 paradigms: a motor task (finger tapping) and a language task (phonemic fluency). Each paradigm was presented as a block paradigm consisting of 90 images, alternating between 20 seconds of activation and 40 seconds for both hands in response to an auditory cue. Self-paced finger tapping task was used at approximately 2 Hz by sequential finger tapping of fingers 2-5 against finger 1 to localize the motor SMA and the primary motor cortex. The language task was used to localize the language SMA and Broca’s area. During the task, letters appeared on the screen and patients were asked to silently generate words that began with that letter. Each subject’s real-time brain activity and head motion were monitored using software (Brainwave, GE Healthcare, Milwaukee, Wisconsin).
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
Probabilistic fiber tracking of the language and motor SMA networks
3
Broca’s area were spurious, we defined 2 additional ROIs (defined as background ROIs) adjacent to the SMA. An anterior background ROI was placed 5-7 voxels anterior to the language SMA on an axial image at the level of the SMA; and a posterior background ROI was placed 5-7 voxels posterior to the motor SMA (Fig. 1).
Probabilistic tractography
Figure 1 The DTI seed ROIs. The language supplementary motor area (SMA), motor SMA and Broca’s Area are obtained from the language and motor fMRI tasks. Activated pixels in the fMRI map represent the voxel with the highest activation strength and they are used as seed ROIs for DTI tractography. Background ROIs are placed anterior to the language SMA and posterior to the motor SMA.
fMRI analysis Image analysis was performed using Analysis of Functional Neuroimaging (AFNI) [20]. Head motion correction was performed using 3D rigid-body registration and spatial smoothing (Gaussian filter with 4-mm full width of half maximum). Corrections for linear trend and high frequency noise were applied when necessary. Functional activity was generated using a correlation analysis. A modeled waveform corresponding to the task performance block was crosscorrelated with all pixel time courses on a pixel-by-pixel basis to identify stimulus locked responses. A threshold of P < 0.001 was used. To reduce false positive activity from large venous structures or head motion, voxels in which the standard deviation of the acquired time series exceeded 8% of the mean signal intensity were set to zero.
Identification of ROIs for tractography Activation maxima obtained from the fMRI were used to determine seed ROIs for DTI fiber tracking. Based on the voxels activated by the fMRI language and motor paradigms, SMA was segregated into 2 functional regions: language and motor SMA. We defined language SMA as the rostral part of the SMA activated during the language task and motor SMA as the caudal part of the SMA that was activated during the motor task. Voxels showing peak functional activity in the language SMA and the motor SMA were selected as the centers of the seed ROIs for fiber tracking in the language SMA and in the motor SMA, respectively. The voxel showing peak functional activity in Broca’s area was chosen as the center of the seed ROI in Broca’s area. All ROIs for tractography were manually drawn on the b0 images based on the activation maxima of the same patients (Fig. 1). To determine whether the white matter connections between the areas of fMRI-defined SMA activation and
DTI data were analyzed using DTI & FiberTools (Department of Diagnostic Radiology, University Hospital, Freiburg, Germany), implemented in MATLAB (Mathworks, Inc., Natick, Massachusetts). Initially, head motion and eddy current issues were corrected if necessary. In DTI & FiberTools, an extended Monte Carlo simulation method similar to the Probabilistic Index of Connectivity (PICO) method [11,12] was used to capture the directional information passing a voxel. To extract the probabilistic map of connectivity between 2 seed regions, a probabilistic map was first generated for each seed ROI; then the maps from the 2 ROIs were combined to compute the probability that each voxel was part of the bundle connecting both seed points [12,21]. Probabilistic maps were generated from individual ROIs for each patient. Tracking areas were defined as regions with mean diffusivity (MD) < 0.002 and fractional anisotropy (FA) > 0.1. The number of random walks was set to 100,000 and the maximum fiber length to 150 voxels (default) [12]. Probabilistic maps generated from the language SMA and Broca’s area ROIs were combined to produce their probabilistic map of connectivity, called the language SMA pathway. Probabilistic maps generated from the motor SMA and Broca’s area ROIs were combined to create the motor SMA pathway. In addition, probabilistic maps generated from each of the background ROIs and Broca’s area ROIs were combined to create background pathways.
Statistical analysis The probabilistic map of connectivity represents the voxels with the highest likelihood of connecting 2 selected ROIs. To compare these maps of each pathway across subjects, we counted all possible connected voxels within the language SMA pathway, as well as the motor SMA pathway of each subject. Only voxels with endpoints in the ROIs were retained as part of the pathway. These voxels were extracted by generating histograms of the connectivity between the seed regions and applying a connectivity threshold to define the areas of highest connectivity. In addition, for both the motor SMA and the language SMA pathways in individual subjects, the mean degree of connectivity and the voxels with the highest connectivity around Broca’s area were measured. Using paired t test, the numbers of voxels, their degrees of connectivity and the degrees of connectivity around Broca’s area were compared across subjects, in the pathways, between pathways and the 2 background pathways, and between the group of patients whose tumors were ipsilateral or contralateral to the pathways.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model
ARTICLE IN PRESS
NEURAD-445; No. of Pages 8
4
M. Jenabi et al.
Figure 2 The probabilistic fiber tracking results of 2 patients. a and b show the language supplementary motor area (SMA) and the motor SMA respectively for a single subject with a tumor in the non-language-dominant hemisphere. c and d show the language SMA and the motor SMA respectively for a different subject with a tumor in the language-dominant hemisphere. Tracks are superimposed on b0 images in the axial and coronal planes at the levels of SMA and BA. Colors represent the degree of connectivity: from 0.00001 in blue to 0.1 in red.
Results In all 20 subjects (including 11 with left- and 9 with righthemisphere tumors), the activated voxels with the most significant P values during the fMRI tasks were located in the left hemisphere. In each subject, fMRI showed 2 separate areas of activation in the SMA: a rostral area within the pre-SMA activated during the language task (defined as the language SMA), and a caudal region within the SMA proper activated during the motor task (defined as the motor SMA). For tractography, the seed ROIs of the language SMA, motor SMA and Broca’s area were in the areas of dominant activation, which were on the left side in all patients. The fMRI determined left hemisphere dominance for language was confirmed by intraoperative electrical stimulation in all patients.
Language SMA and motor SMA pathways The language SMA and motor SMA pathways to Broca’s area were observed as separate and unique in all patients. The language SMA pathway consistently traveled anterior to the motor SMA pathway. Fig. 2 illustrates the probabilistic fiber tracking results of 2 patients with dominant left SMAs: one with a tumor in the non-dominant and another with a tumor in the dominant hemisphere. Fig. 3 shows the volume renderings of both pathways for a representative subject. Regardless of the location or laterality of the tumor, the language SMA and motor SMA pathways traveled in separate
Table 1
Figure 3 Volume rendering image of supplementary motor area (SMA) pathways. It is showing the projections of both language and motor SMA pathways of a single subject from ROIs. Red: the language SMA pathway; blue: the motor SMA pathway; green: the language SMA seed; yellow: the motor SMA seed, magenta: the Broca’s area seed.
but parallel routes from the SMA to the frontal language area. In all subjects, the mean number of voxels in the language SMA pathway (510) was greater than that in the motor SMA pathway (278), (P < 0.0001) (Table 1). For all subjects, the degree of connectivity was stronger for the language SMA pathway (0.12) than for the motor SMA pathway (0.07) (P < 0.001). In addition, the degree of connectivity around Broca’s area was stronger for the language SMA pathway (0.4) than for the motor SMA pathway (0.16) (P < 0.001). The number of voxels and the degree of connectivity of the language SMA pathway were greater than those of the motor SMA pathway within every subject, although they varied among subjects (Table 1).
Background ROIs In each patient, the number of voxels was greater in the language and motor SMA pathways than in either the anterior (P < 0.0001) or posterior background pathways (P < 0.0001) (Table 2), (Fig. 4). Mean numbers of voxels in the language and motor SMA pathways were higher than those in both background pathways (P < 0.0001).
Side of tumor Probabilistic connectivity maps of the language SMA pathways did not differ when stratified as ipsilateral or contralateral to the tumor (mean number of voxels: 474 vs. 555, respectively, with P < 0.35; mean degree of connectivity: 0.1 vs. 0.3, respectively, with P < 0.76; Table 3). Similarly, no differences were found between motor SMA pathways that were ipsilateral to the tumor and those that were contralateral to the tumor (mean number of voxels:
Comparison of the language and motor supplementary motor area (SMA) pathways.
Language SMA pathway Motor SMA pathway P value
Mean number of voxels
Mean degree of connectivity
Degree of connectivity at BA
510 278 < 0.0001
0.1 0.0350 < 0.0001
0.388 0.160 < 0.001
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model
ARTICLE IN PRESS
NEURAD-445; No. of Pages 8
Probabilistic fiber tracking of the language and motor SMA networks
5
Table 2 Comparison of mean number of voxels and degree of connectivity of probabilistic maps between language and motor supplementary motor area (SMA) pathways and background SMA pathways. Background SMA
Anterior pathway Posterior pathway
Mean number of voxels
P value Language SMA pathway
Motor SMA pathway
159 105
< 0.0001 < 0.0001
< 0.001 0.003
Mean degree of connectivity
0.05 0.03
P value Language SMA pathway < 0.0001 < 0.0001
Discussion In the present study, we investigated the existence, positions and relationships of fiber networks between the SMA and Broca’s area using fMRI and probabilistic tractography. Our results indicated that the language SMA is more robustly connected to Broca’s area than is the motor SMA as reflected by a higher mean number of activated voxels (P < 0.0001) and a greater degree of connectivity (P < 0.0001). These results have implications for preoperative planning in patients with brain tumors.
Language and motor pathways
Figure 4 Comparison of language supplementary motor area (SMA) and motor SMA pathways with anterior and posterior background pathways: (X-axis: each patient; y-axis: the number of connected voxels). The mean number of voxels is indicated inside the box plot of each group. Outside value is shown by (*).
266 vs. 293, respectively, with P < 0.74; mean degree of connectivity: 0.05 vs. 0.09, respectively, with P < 0.59). Regardless of the side of the tumor, the SMA pathways were always on the left side, and the probabilistic connectivity of the language SMA pathway was always greater than that of the motor SMA pathway. In the 11 patients with ipsilateral left tumors (in respect to the location of ROI), the mean number of activated voxels (P < 0.001) and the mean degree of connectivity (P < 0.01) were higher in the language SMA pathway than in the motor SMA pathway (Table 3). In the 9 patients with contralateral right tumors, the mean number of activated voxels (P < 0.002) and the mean degree of connectivity (P < 0.02) were also higher in the language SMA pathway than in the motor SMA pathway.
Table 3
We demonstrated separate language and motor white matter pathways starting in regionally specific areas of the SMA that coursed toward the ipsilateral frontal language area via distinct, semi-parallel pathways. The structural and functional connectivity between the SMA and Broca’s area have been discussed in a number of papers [7,8,17,22,23] showed potential connectivity between Broca’s area and the SMA, using anatomically based ROIs and deterministic or adaptive tractography for fiber tracking in normal subjects. Another study [8] used probabilistic tractography to show connections between Broca’s area and both the pre-SMA and SMA proper. The investigators used anatomical landmarks to define these regions in healthy adults. Our results agree with their findings and support the existence of anatomical and functional connectivity between two SMA subregions and the frontal language area. Unlike previous anatomy-based approaches [8,24], our results benefit from using fMRI specific activation to define the ROIs for probabilistic tracking. fMRI-driven tractography has been described as more accurate in patients with brain tumors [25—27]. We showed that the different parts of the SMA responsible for specific tasks have differing probabilities of connectivity to Broca’s area. Although the
Statistical results of 2 pathways in patients with an ipsilateral and contralateral tumor. Ipsilateral
Language SMA pathway Motor SMA pathway P value
Contralateral
Mean number of voxels
Mean degree of connectivity
Mean number of voxels
474 266 < 0.0001
0.100 0.050 0.01
555 293 0.0018
SMA: supplementary motor area.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
6
M. Jenabi et al.
connectivity among these regions has been described before [7,8], the degree of connectivities of the pathways have not previously been quantified. We selected 3 measures of connectivity for comparison: the number of voxels with the highest connectivity value near Broca’s area, the mean value of connectivity along each pathway and the number of voxels within each pathway. 3D tractography renderings were also generated to visualize the spatial orientation of the pathways to each other: in each case, the language SMA pathway contained more voxels than the motor SMA pathway, as demonstrated by a larger surface area (radius), especially around Broca’s area; the lengths of the 2 pathways were similar (Fig. 3). These differences occurred despite constant seed ROI sizes, suggesting that the probabilistic results reflect true differences in connectivity between the language SMA and motor SMA pathways. Since the probabilistic maps representing the language and motor SMA pathways showed that the voxels with the highest connectivity value were located near the frontal language area, it is reasonable to conclude that the connectivity to Broca’s area from the language component of SMA was consistently more robust than the connection to Broca’s area from the motor component of SMA. Traditional brain mapping techniques have relied on post-mortem fiber dissections, indirect evaluation of lesion locations and clinical deficits in neurologically impaired patients, and tracer studies in primate models. The advent of non-invasive tools such as DTI has allowed the in vivo evaluation of anatomical and functional pathways. These tools have been used to describe complex networks that involved in specific task performances, and have rapidly expanded our understanding of supportive language networks such as the SMA pathways. For example, the exact location of the cortico-spinal tract was matter of debate based on clinical material and anatomical dissection. This anatomical question was settled convincingly by tractography [13].
Structural and functional SMA Studies suggested important roles of the pre-SMA and SMA proper in different aspects of word production and motor control [18,22,23,28]. Interactions between different regions within the language SMA and the frontal language areas in many aspects of language, including word production, have been reported previously [8,9,18,29]. We showed that the language SMA pathway implies connectivity of part of the language SMA to the Broca’s area. The motor SMA pathway revealed connectivity of the motor SMA to Broca’s area. Given that patients with postoperative SMA syndromes exhibit a paucity of speech and have a marked deficit in initiation of speech in particular, it is not surprising that motoric components of planning are directly connected to Broca’s area [5,29—31]. The background pathways showed lower connectivity with Broca’s area than did the language and motor SMA pathways. Since connectivity was determined from seed ROIs placed based on fMRI-activated voxels during simple language and motor tasks, it is reasonable to assume that additional regions of functional activity might be distinguished within the pre-SMA and SMA proper if other fMRI
paradigms were applied. Other studies have reached similar conclusions [18,22]. The differences we found between the connectivity parameters of the language SMA and motor SMA pathways and those of the background pathways further support the work of Peck, who described the structural and functional relations between the language SMA and motor SMA and identified joint activation in the border between these two areas during a simultaneous motor and language task [3].
Limitations Data described in the study were acquired from MRI examinations performed for standard clinical care. Hence, it is possible that microscopic tumor infiltration and/or edema could have affected diffusion measurements, particularly for gliomas [26,27,32,33]. A second potential limitation relates to the use of the probabilistic method for tractography. Probabilistic diffusion tractography is a non-invasive tool for predicting and extracting the white matter network and it relies on several non-anatomical assumptions to measure the neural connectivity. However, the advantages of the probabilistic over deterministic fiber tracking methods are especially apparent in cases where the target tracts involve crossing fibers [11,34,35]. The probabilistic method we used belongs to a method of pathway extraction [12] that determines the most probable neural pathways between two regions by identifying point-to-point connections without any a priori information. This method has been validated in different tracts that cross other fiber bundles several times, i.e., optic radiation tracts, and the lateral part of the cortico-spinal tract.
Clinical relevance While injury to both gray and white matter structures may cause devastating neurologic deficits, at surgery the white matter is less accessible (deeper), more difficult to identify by anatomic landmarks, and more susceptible to false negative results at stimulation. Iatrogenic injury to the pre-SMA (language) or the SMA proper (motor) areas or the associated SMA pathways can result in an SMA syndrome that can produce paresis, dysphasia and even mutism depending upon the hemisphere and extent of injury [5,6]. Interestingly, the majority of these functions return weeks to months following the inception of the deficit but permanent injury has been described [5,6]. The fact that these functions usually return offers an interesting opportunity to use diffusion tractography to measure dynamic changes in the cognitive/sensory pathways over time. It is estimated that as many as 26% of superior frontal glioma resections result in SMA syndromes to varying degrees. DTI maps are ever increasing in their behavioral specificity allowing the correlation between DTI measured pathways and expected outcomes [36]. Further, Fortin et al. discuss the utility of DTI maps in neurosurgery saying ‘‘Complex neurocognitive functions are not attributed to a single brain area but depend on the dynamic interactions of distributed brain areas operating in large-scale networks [37]’’. This is important in the field of neurosurgery where
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
Probabilistic fiber tracking of the language and motor SMA networks intervention within a spatially localized area may indirectly lead to unwanted effects on distant areas.
Conclusions We used fMRI-driven probabilistic fiber tracking to extract the most probable tracts and to quantify the patterns of connectivity between two functionally distinct regions of the SMA and the Broca’s area in brain tumor patients. The connectivity between Broca’s area and the pre-SMA (language SMA) was stronger than the connectivity between Broca’s area and the SMA proper (motor SMA).
Disclosure of interest The authors declare that they have no conflicts of interest concerning this article.
Acknowledgments The authors thank Ms. Ada Muellner and Katharine Ristich for her expert editorial advice.
References [1] Picard N, Strick PL. Motor areas of the medial wall: a review of their location and functional activation. Cereb Cortex 1996;6:342—53. [2] Nachev P, Kennard C, Husain M. Functional role of the supplementary and pre-supplementary motor areas. Nat Rev Neurosci 2008;9:856—69. [3] Peck KK, Bradbury M, Psaty EL, et al. Joint activation of the supplementary motor area and presupplementary motor area during simultaneous motor and language functional MRI. Neuroreport 2009;20:487—91. [4] Gupta A, Shah A, Young RJ, et al. Imaging of brain tumors: functional magnetic resonance imaging and diffusion tensor imaging. Neuroimaging Clin N Am 2010;20: 379—400. [5] Krainik A, Lehéricy S, Duffau H, et al. Postoperative speech disorder after medial frontal surgery: role of the supplementary motor area. Neurology 2003;60: 587—94. [6] Rosenberg K, Nossek E, Liebling R, et al. Prediction of neurological deficits and recovery after surgery in the supplementary motor area: a prospective study in 26 patients. J Neurosurg 2010;113:1152—63. [7] Morgan VL, Mishra A, Newton AT, et al. Integrating functional and diffusion magnetic resonance imaging for analysis of structure-function relationship in the human language network. PLoS ONE 2009;4:e6660. [8] Ford A, McGregor KM, Case K, et al. Structural connectivity of Broca’s area and medial frontal cortex. Neuroimage 2010;52:1230—7. [9] Klein JC, Behrens TEJ, Robson MD, et al. Connectivitybased parcellation of human cortex using diffusion MRI: established reproducibility, validity and observer independence in BA 44/45 and SMA/pre-SMA. Neuroimage 2007;34: 204—11. [10] Maddah M, Wells WM, Warfield SK, et al. Probabilistic clustering and quantitative analysis of white matter fiber tracts. Inf Process Med Imaging 2007;20:372—83.
7
[11] Parker JG, Haroon HA, Wheeler-Kingshott CA. A framework for a streamline-based probabilistic index of connectivity (PICo) using a structural interpretation of MRI diffusion measurements. J Magn Reson Imaging 2003;18: 242—54. [12] Kreher BW, Schnell S, Mader I, Il’yasov KA, et al. Connecting and merging fibres: pathway extraction by combining probability maps. Neuroimage 2008;43:81—9. [13] Holodny AI, Watts R, Korneinko VN, Pronin IN, Zhukovskiy ME, Gor DM, et al. Diffusion tensor tractography of the motor white matter tracts in man: current controversies and future directions. Ann N Y Acad Sci 2005;1064: 88—97. [14] Sinha S, Bastin ME, Whittle IR, et al. Diffusion tensor MR imaging of high-grade cerebral gliomas. AJNR Am J Neuroradiol 2002;23:520—7. [15] Nimsky C, Ganslandt O, Hastreiter P, et al. Intraoperative diffusion-tensor MR imaging: shifting of white matter tracts during neurosurgical procedures–initial experience. Radiology 2005;234:218—25. [16] Price C. The anatomy of language: contributions from functional neuroimaging. J Anat 2000;197:335—59. [17] Koechlin E, Jubault T. Broca’s area and the hierarchical organization of human behavior. Neuron 2006;50: 963—74. [18] Tremblay P, Gracco VL. Contribution of the pre-SMA to the production of words and non-speech oral motor gestures, as revealed by repetitive transcranial magnetic stimulation (rTMS). Brain Res 2009;1268:112—24. [19] Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 1971;9: 97—113. [20] Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res 1996;29:162—73. [21] Saur D, Kreher BW, Schnell S, et al. Ventral and dorsal pathways for language. Proc Natl Acad Sci U S A 2008;105: 18035—40. [22] Meister B, Boroojerdi H, Foltys R, et al. Motor cortex hand area and speech: implications for the development of language. Neuropsychologia 2003;41:401—6. [23] de Lafuente V, Romo R. Language abilities of motor cortex. Neuron 2004;41:178—80. [24] Johansen-Berg H, Behrens TEJ, Robson MD, et al. Changes in connectivity profiles define functionally distinct regions in human medial frontal cortex. Proc Natl Acad Sci U S A 2004;101:13335—40. [25] Schonberg T, Pianka P, Hendler T, et al. Characterization of displaced white matter by brain tumors using combined DTI and fMRI. Neuroimage 2006;30(4):1100—11. [26] Kamiya K, Sato N, Saito Y, et al. Accelerated myelination along fiber tracts in patients with hemimegalencephaly. J Neuroradiol 2013 [In press]. [27] Zolal A, Hejcl A, Malucelli A, et al. Distant White-matter diffusion change caused by tumor growth. J Neuroradiol 2013;40:71—80. [28] Hoshi E, Tanji J. Differential roles of neuronal activity in the supplementary and presupplementary motor areas: from information retrieval to motor planning and execution. J Neurophysiol 2004;92:3482—99. [29] Alario FX, Chainay H, Lehericy S, et al. The role of the supplementary motor area (SMA) in word production. Brain Res 2006;1076:129—43. [30] Binkofski F, Buccino G. Motor functions of the Broca’s region. Brain Lang 2004;89:362—9. [31] Amunts K, Lenzen M, Friederici AD, et al. Broca’s region: novel organizational principles and multiple receptor mapping. PLoS Biol 2010;8:e1000489.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
+Model NEURAD-445; No. of Pages 8
ARTICLE IN PRESS
8
M. Jenabi et al.
[32] Schreiber A, Hubbe U, Ziyeh S, et al. The influence of gliomas and nonglial space-occupying lesions on blood-oxygen-level -dependent contrast enhancement. AJNR Am J Neuroradiol 2002;21:1055—63. [33] Ulmer JL, Krouwer HG, Mueller WM, et al. Pseudoreorganization of language cortical function at fMR imaging: a consequence of tumor-induced neurovascular uncoupling. AJNR Am J Neuroradiol 2003;24:213—7. [34] Behrens TE, Woolrich MW, Jenkinson M, et al. Characterization and propagation of uncertainty in diffusion-weighted MR imaging. Magn Reson Med 2003;50:1077—88.
[35] Lazar M, Weinstein JS, Tsuruda KM, et al. White matter tractography using diffusion tensor deflection. Hum Brain Mapp 2003;18:306—21. [36] Russell SM, Kelly PJ. Incidence and clinical evolution of postoperative deficits after volumetric stereotactic resection of glial neoplasms involving the supplementary motor area. Neurosurgery 2003:506—16. [37] Fontaine D, Capelle L, Duffau H. Somatotopy of the supplementary motor area: evidence from correlation of the extent of surgical resection with the clinical patterns of deficit. Neurosurgery 2002:297—303.
Please cite this article in press as: Jenabi M, et al. Probabilistic fiber tracking of the language and motor white matter pathways of the supplementary motor area (SMA) in patients with brain tumors. J Neuroradiol (2013), http://dx.doi.org/10.1016/j.neurad.2013.12.001
