J. Biomedical Science and Engineering, 2013, 6, 192-200 JBiSE
http://dx.doi.org/10.4236/jbise.2013.62023 Published Online February 2013 (http://www.scirp.org/journal/jbise/)
Diffusion tensor tractography of the arcuate fasciculus in
patients with brain tumors: Comparison between
deterministic and probabilistic models
Zhixi Li1, Kyung K. Peck1,2, Nicole P. Brennan1, Mehrnaz Jenabi1, Meier Hsu3, Zhigang Zhang3,
Andrei I. Holodny1,4, Robert J. Young1,4
1Department of Radiology, Memorial Sloan-Kettering Cancer Center, New York, USA
2Department of Medical Physics, Memorial Sloan-Kettering Cancer Center, New York, USA
3Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, USA
4Brain Tumor Center, Memorial Sloan-Kettering Cancer Center, New York, USA
Received 11 November 2012; revised 11 December 2012; accepted 17 December 2012
ABSTRACT
Purpose: The purpose of this study was to compare
the deterministic and probabilistic tracking methods
of diffusion tensor white matter fiber tractography in
patients with brain tumors. Materials and Methods:
We identified 29 patients with left brain tumors <2
cm from the arcuate fasciculus who underwent pre-
operative language fMRI and DTI. The arcuate fas-
ciculus was reconstructed using a deterministic Fi-
ber Assignment by Continuous Tracking (FACT)
algorithm and a probabilistic method based on an
extended Monte Carlo Random Walk algorithm.
Tracking was controlled using two ROIs corre-
sponding to Broca’s and Wernicke’s areas. Tra cts in
tumoraff ected hemispheres were examined for exten-
sion between Broca’s and Wernicke’s areas, ante-
rior-posterior length and volume, and compared with
the normal contralatera l tracts. Results: Probabilistic
tracts displayed more complete anterior extension to
Broca’s area than did FACT tracts on the tumor-af-
fected and normal sides (p < 0.0001). The median
length ratio for tumor: normal sides was greater for
probabilistic tracts than FACT tracts (p < 0.0001).
The median tract volume ratio for tumor: normal
sides was also greater for probabilistic tracts than
FACT tracts (p = 0.01). Conclusion: Probabilistic
tractography reconstructs the arcuate fasciculus more
completely and performs better through areas of tu-
mor and/or edema. The FACT algorithm tends to
underestimate the anterior-most fibers of the arcuate
fasciculus, which are crossed by primary motor fi-
bers.
Keywords: Diffusion Tensor Imaging; DTI;
Tractography; Probabilistic; FACT; Arcuate Fasciculus;
Brain Tumors
1. INTRODUCTION
The arcuate fasciculus is an important white matter tract
that connects the frontal (Broca’s) and temporal (Wer-
nicke’s) language regions of the brain. Lesions to the
arcuate fasciculus have been associated with language
deficits [1-4]. Diffusion tensor imaging (DTI) is an
emerging magnetic resonance imaging (MRI) technique
that allows visualization and characterization of white
matter tracts such as the arcuate fasciculus [5,6]. By
modeling the direction and magnitude of water diffusion,
DTI encodes the orientation of white matter fibers on a
voxel-by-voxel basis [7-9]. Tractography algorithms util-
ize the DTI encoded information to reconstruct white
matter tracts [10-12]. Tractography algorithms can be
broadly divided into two categories: deterministic and
probabilistic. FACT is a streamline-based deterministic
method that traces pathways from a seed region by fol-
lowing the primary eigenvector from one voxel to the
next [12,13]. The probabilistic algorithm defines path-
ways by generating multiple curves from seed points
using a Monte Carlo simulation [14-16]. Probability of
connectivity is then assigned to individual voxels based
on the frequency with which the curves traverse the vox-
els.
In regions where white matter tracts cross, merge or
diverge, deterministic tractography often cannot differ-
entiate individual white matter bundles, occasionally re-
constructing erroneous pathways or prematurely termi-
nating true pathways [10,17,18]. Tractography in patients
with brain tumors may be further compromised by mass
effect and/or decreased fractional anisotropy (FA) from
the tumor or tumor-related vasogenic edema, tract infil-

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 193
tration, and destruction [19]. Probabilistic tractography
appears to improve the identification of crossing fibers
and the tracking of long white matter tracts such as the
arcuate fasciculus [20]. Although several studies have
evaluated the use of deterministic or probabilistic tracto-
graphy in the reconstruction of the arcuate fasciculus in
normal subjects [6,20-23], currently none has applied
these techniques to reconstruct the arcuate fasciculus in
patients with brain tumors. The purpose of this study is
to use both FACT and probabilistic methods to map the
arcuate fasciculus connecting two main language tracks
(Broca’s area and Wernicke’s area) and to compare the
methods when a tumor is located nearby.
2. MATERIALS AND METHODS
Whole-brain DTI was performed to map white matter
tracks of arcuate fasciculus in 29 brain tumor patients.
FACT and probabilistic tractography algorithms were
compared based on the volume of the tracks and connec-
tivity in the arcuate fasciculus for both normal and tumor
sides. These methods are described in detail below.
2.1. Subjects
This retrospective study was granted a Waiver of In-
formed Consent by the hospital Institutional Review
Board and was fully compliant with Health Insurance
Portability and Accountability Act regulations. We iden-
tified 29 patients (14 females and 15 males; mean 58.5
years, range 32 - 83 years) from May 2007 to June 2011.
Each patient had a solitary brain tumor in the left hemi-
sphere less than 2 cm from the expected course of the
arcuate fasciculus. None of the patients had magnetic
resonance (MR) evidence of tumor (enhancement, mass
effect or FLAIR hyperintensity) on the right side in the
expected location of Broca’s area, Wernicke’s area or the
arcuate fasciculus, as determined by a board certified
radiologist holding a certificate of added qualification in
neuroradiology. Each patient had DTI performed as part
of their MRI scan for treatment planning (n = 17), and/or
for language difficulties (n = 26) usually expressive
aphasia (n = 22). Patient characteristics are summarized
in Table 1. Tumors included twelve glioblastomas, seven
astrocytomas, four oligodendrogliomas, two metastases,
two large B-cell lymphomas, one ependymoma and one
meningioma. In 13 patients, prior treatments included
surgery, chemotherapy and radiation therapy (n = 7),
surgery and radiation therapy (n = 2), radiation and che-
motherapy, surgery and chemotherapy, surgery, or che-
motherapy (n = 1 each).
2.2. Data Acquisition
Imaging was performed on a 1.5-Tesla (n = 6) or a
Table 1. Summary of subjects and treatments received.
TOTAL
Patients 29
Mean age (range), years 58 (32 - 79)
Sex, n (%)
Female 14 (48.3)
Male 15 (51.7)
Treatment prior to scan, n (%)
No treatment 16 (55.2)
Surgery, radiation therapy and/or chemotherapy 13 (44.8)
Median months to scan for patients who received
treatment (range)
Surgery 19 (1 - 126)
Radiation 12.5 (5 - 135)
Median radiation dose in Gy (range) 54 (30 - 60)
3.0-Tesla (n = 23) magnet (Signa HDx and Excite, GE
Medical Systems, Milwaukee, WI) using a standard
quadrature head coil. All patients underwent DTI using a
single-shot spin-echo echo-planar imaging sequence with
15 (n = 17) or 25 (n = 12) non-collinear gradient direc-
tions; TR/TE, 11000-13,500/60-100 ms; matrix, 128 ×
128; in-plane resolution, 1.88 × 1.88 mm; slice thickness,
3 mm to cover the whole brain; b-value, 1000 s/mm2;
and NEX, 1. Seventeen patients also underwent fMRI
using a single-shot gradient-echo echo-planar imaging
sequence (TR/TE, 4000/(30 - 40) ms; matrix, 128 × 128;
flip angle, 90˚; slice thickness, 4.5 mm). Block-design
language fMRI including phonemic fluency, semantic
fluency and verb generation paradigms was analyzed in
analysis of functional neuroImages (AFNI) [24]. The
standard MRI protocol also included sagittal and axial
T1-weighted images; axial T2-weighted, FLAIR and diffu-
sion-weighted images; axial gradient-echo or susceptibil-
ity-weighted images; and contrast coronal, sagittal and
axial T1-weighted images as well as axial 3D spoiled
gradient recalled images.
2.3. DTI Data Analysis
Head motion and eddy current issues were initially cor-
rected if necessary. Using DTI & FiberTools software
[24,25] (Medical Physics, Department of Diagnostic Ra-
diology, University Hospital, Freiburg, Germany) im-
plemented in MATLAB (Mathworks, Natick, MA), the
arcuate fasciculus was reconstructed by applying a de-
terministic algorithm based on FACT [12] as well as a
probabilistic algorithm similar to the PICO method [16].
2.4. Definition of Seed Regions-of-Interest
(ROIs)
Probabilistic and FACT fiber tracking were performed
for both the left hemisphere (ipsilateral to the tumor in

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 194
every case) and the right hemisphere (normal side) in all
patients controlled by the following input variables: 1) if
fMRI was performed, two language fMRI-defined seed
ROIs (diameter, 8 mm) corresponding to activated re-
gions in Broca’s and Wernicke’s areas [22]; 2) if fMRI
was not available, seed ROIs (diameter 8 mm) that were
approximated by an expert operator (with 3 years ex-
perience in functional imaging) under the direct supervi-
sion of a board certified radiologist holding a certificate
of added qualification in neuroradiology (with 13 years
experience) using anatomical landmarks (Brodmann’s
area 44 [pars opercularis and pars triangularis of the in-
ferior frontal gyrus] for Broca’s area and posterior Brod-
mann’s area 22 [posterior superior temporal gyrus] for
Wernicke’s area) [26].
2.5. FACT Tractography
For FACT, tracts were traced following the principle
eigenvector within each voxel. Termination criteria were:
FA ≤ 0.15 and turning angle ≥ 45˚ [22,23]. White matter
bundles connecting Broca’s and Wernicke’s areas were
selected by using an “and” operation to isolate tracts
passing through both ROIs. In cases where the fMRI or
anatomical ROIs did not produce a tract in FACT tracto-
graphy, the anterior ROI in Broca’s area was replaced by
an ROI in the midportion of the arcuate fasciculus in the
superior longitudinal fasciculus, on the coronal plane
near the rostral corpus callosum and lateral to the corona
radiata fibers [23].
2.6. Probabilistic Tractography
The probabilistic method was based on a Monte Carlo
Random Walk approach similar to the probabilistic index
of connectivity (PICO) implementation [16,20]. In this
method, the extended Monte Carlo Random Walk propa-
gates a high number of trajectories from each ROI and
extracts the directionality of the trajectories passing
through each voxel [20]. The frequency with which a
voxel is visited by trajectories determines the degree of
connectivity to the origin ROI. The number of random
walks was set at 100,000. Termination criteria were: FA
≤ 0.15 and fiber length > 150 voxels. A mask containing
the corona radiata was generated by the FACT method
and removed from the tracking area [20]. Probabilistic
maps from Broca’s and Wernicke’s area seed ROIs were
multiplied together to obtain a map of the tract connect-
ing the two ROIs. Trajectories that started from the two
ROIs and travelled in opposite directions were defined as
connecting tracts, while trajectories that travelled in the
same direction were defined as merging fibers and sepa-
rated [20]. The resulting map contained a voxel-by-voxel
estimation of the degree of connectivity to both ROIs,
and represented the most probable direct pathway be-
tween Wernicke’s and Broca’s areas. Extraneous tracts
were removed from the tracking area by an experienced
neuroradiologist based on anatomical landmarks using a
mask function. In each patient, the same threshold was
applied for probabilistic tracts on both the tumor-affected
as well as the normal sidesto remove any remaining ex-
traneous tracts while retaining the arcuate fasciculus. The
threshold chosen varied between patients (mean 5.74 ×
10−4, range 9.21 × 10−5 to 5.31 × 10−3). The variation in
threshold was used because of the heterogeneity of the
dataset with respect to tumor grade, degree of tract infil-
tration, and edema and mass effects on the tracts. Thresh-
olding and masking techniques have been described pre-
viously [20,27,28].
2.7. Tractography Analysis
Probabilistic and FACT tract volumes were extracted into
masks and the number of voxels was calculated. Within
patients, probabilistic and FACT tract volumes were
compared on each side as well as between sides. The
tumor side: normal side probabilistic volume ratio and
the tumor side: normal side FACT volume ratio were
recorded. For probabilistic tractography, voxels with the
highest connectivity were considered those most likely to
be part of the connecting bundle between two ROIs. We
measured the mean degree of connectivity of voxels that
were part of the connecting bundle on the tumor and
normal sides.
To evaluate the completeness of the arcuate fasciculus
on tumor-affected and normal sides, the FACT and prob-
abilistic tracts were scored for their anterior termination
in a manner similar to Bernal et al. [23]: 0 = no fibers
reached Broca’s area; 1 = few fibers; 2 = most fibers; and
3 = all fibers. The maximum anterior-posterior length of
each tract was measured in the sagittal plane by counting
the number of voxels in a straight line from the most
anterior point of the tract to the most posterior point of
the tract. Length ratios were produced by dividing the
probabilistic tract length on the tumor side by the prob-
abilistic tract length on the normal side, and repeated for
the FACT tract lengths.
2.8. Statistical Analysis
Comparisons were performed using Wilcoxon signed-
rank tests. Results were also stratified by treatment status
and compared with Wilcoxon rank sum tests. To adjust
for multiple comparisons, the false discovery rate proce-
dure was applied [29]. The p-values were ranked from
smallest to largest, compared to false discovery rate sig-
nificance thresholds and declared significant when less
than the false discovery rate threshold.
The template is used to format your paper and style
the text. All margins, column widths, line spaces, and

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 195
text fonts are prescribed; please do not alter them. You
may note peculiarities. For example, the head margin in
this template measures proportionately more than is custo-
mary. This measurement and others are deliberate, using
specifications that anticipate your paper as one part of
the entire journals, and not as an independent document.
Please do not revise any of the current designations.
3. RESULTS
The results are summarized in Table 2. More probabilis-
tic tracts reached Broca’s area than FACT tracts on the
tumor side (p < 0.0001). Probabilistic tracts had anterior
termination scores = 3 (all reached Broca’s area) in 11/29
(37.9%) cases, and scores = 1 - 2 (few or most tracts) in
14 (48.3%) cases. FACT had scores = 1 - 2 in 3 (10.3%)
cases and scores =0 (no tracts reaching Broca’s area) in
26 (89.7%) cases. On the normal side, probabilistic track-
ing of the arcuate fasciculus homologue again outper-
formed FACT (p < 0.0001). Probabilistic tracts had
scores = 3 in 13/29 (44.8%) cases and scores = 1 - 2 in
16 (55.2%) cases, while FACT had scores = 3 in 2/29
(6.9%) cases, = 1 - 2 in 10 (34.5%) cases, and = 0 in 17
(58.6%) cases. Figure 1 shows an example in which the
FACT tract ended before the Broca’s area while the cor-
responding probabilistic tract displayed complete exten-
sion. Figure 2 demonstrates truncation of the FACT tract
by descending corticobulbar fibers in the region of the
centrum semiovale.
The median length ratio of the probabilistic tracts
(tumor side: normal side) was 1.00, greater than the me-
dian of 0.88 for FACT tracts, (p < 0.0001). The median
length ratio of probabilistic: FACT tracts on the tumor
side was 1.27 (p < 0.0001), due to decreased length of
the FACT tracts, while the median length ratio of prob-
abilistic: FACT on the normal side was 1.09. The median
tract volume ratio (tumor: normal) was greater for prob-
abilistic tracts at 1.09 than for FACT tracts at 0.77 (p =
0.01). In the example in Figure 3, the FACT tract thins
near the tumor while the probabilistic tract remains ro-
bust.
The mean connectivity of the voxels comprising the
probabilistic tracts on the tumor side (median 0.043,
range 0.016 - 0.38) was not different from that of the
probabilistic tracts on the normal side (median 0.046,
range 0.019 - 0.079) (p = 0.88). No differences were
found between untreated and treated patients (p ≥ 0.06)
(Table 3).
4. DISCUSSION
In the current study, we used DTI and probabilistic trac-
tography to successfully reconstruct white matter fibers
of the arcuate fasciculus in patients with brain tumors.
Our results demonstrated that the probabilistic tracto-
graphy outperformed the FACT tractography in estimate-
ing the extent and degree of connectivity of the arcuate
fasciculus affected by brain tumors and/or peritumoral
abnormalities. We also found that FACT tended to un-
derestimate the extent of the anterior-most fibers of the
arcuate fasciculus, where the fibers cross the descending
corticobulbar fibers. Localization of the arcuate fascicu-
lus is relevant for neurosurgical planning since direct
brain stimulation is less reliable for white matter tracts
than for grey matter cortical structures [30-32].
Table 2. Summary of results.
Variable Median value (range)Median difference between groups (range) p-value*
Probabilistic anterior termination scores (tumor side) 2 (0 - 3)
FACT anterior termination scores (tumor side) 0 (0 - 2) 2 (0 - 3) <0.0001
Probabilistic anterior termination scores (normal side) 2 (0 - 3)
FACT anterior termination scores (normal side) 0 (0 - 3) 2 (0 - 3) <0.0001
Probabilistic length ratio (tumor side: normal side) 1.00 (0 - 1.47)
FACT length ratio (tumor side: normal side) 0.88 (0 - 1.63) 0.09 (−0.016 - 1) <0.0001
Tumor side length ratio (probabilistic: FACT) 1.27 (1 - 2.31)
Normal side length ratio (probabilistic: FACT) 1.09 (1 - 1.59) 0.10 (−0.11 - 4) <0.0001
FACT volume (tumor side) 303 (0 - 916)
FACT volume (normal side) 389 (153 - 877)
−79 (−773 - 285) 0.02
Probabilistic volume (tumor side) 600 (0 - 1596)
Probabilistic volume (normal side) 637 (280 - 1080) 51 (−1104 - 742) 0.86
FACT volume ratio (tumor side: normal side) 0.77 (0 - 1.85)
Probabilistic volume ratio (tumor side: normal side) 1.09 (0 - 1.87)
−0.28 (−0.76 - 1.36) 0.01
Probabilistic mean connectivity (tumor side) 0.04 (0.02 - 0.38)
Probabilistic mean connectivity (normal side) 0.05 (0.02 - 0.10) 0.00 (−0.03 - 0.33) 0.88
*Italicized results were significant after adjusting for multiple comparisons by the false discovery rate procedure.

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 196
Figure 1. Images in a 62-year-old man with pathologically
proven glioblastoma. (A) Contrast sagittal T1-weighted image
shows a glioblastoma in the centrum semiovale near the middle
portion of the arcuate fasciculus; (B) Probabilistic tracking
demonstrates the entire course of the arcuate fasciculus; (C)
FACT (magenta) fails to reconstruct the anterior-most fibers.
Broca’s area, as activated on fMRI, is highlighted in red; (D)
Probabilistic and FACT tracts overlaid on the same sagittal
slice confirm the absence of anterior fibers in the FACT tract.
Figure 2. Images in a 51-year-old man with pathologically
proven glioblastoma. (A) FACT of the arcuate fasciculus (ma-
genta) and probabilistic tracking of the corticobulbar fibers
(arrow) overlaid on a sagittal contrast T1-weighted image.
Probabilistic tracking of the corticobulbar fibers appear at the
termination of the anterior FACT fibers; (B) Axial; and (C)
Coronal B0 diffusion tensor images show that instead of reach-
ing Broca’s area the arcuate fibers (magenta) are prematurely
truncated by descending corticospinal and corticobulbar fibers
(blue and yellow; arrows) seen with probabilistic tractography.
The arcuate fasciculus is thought to connect Wer-
nicke’s area and Broca’s area [33], as supported by mod-
ern deterministic tractography studies [22,26]. Bernal et
al. recently suggested that the rostral termination of the
arcuate fasciculus at Broca’s area is less definite than
once thought [23]. Using deterministic tractography, the
arcuate fasciculus terminated before the putative Broca’s
area in 58.3% of normal subjects, and in the remaining
cases the connecting fibers were minimal. Rather than
terminating in premotor areas and connecting to Broca’s
area via a relay station, [23,34] however, the arcuate fas-
ciculus is more likely to have prematurely terminated in
the centrum semiovale by crossing fibers.
Figure 3. Images in a 58-year-old woman with pathologically
proven low grade oligodendroglioma. Axial, coronal and sagit-
tal contrast T1 weighted images (A)-(C) and B0 images with
FACT tract overlay (magenta, (D)-(F)) show a nonenhancing
oligodendroglioma in the left inferior frontal gyrus, with dis-
continuity of fibers in the region near the tumor. B0 images
with probabilistic tract overlay (G)-(I) show more robust tracts
around the tumor.
In the centrum semiovale, fibers from the corticospinal
tract, superior longitudinal fasciculus, and interhemispheric
callosal fibers meet and cross [35]. Previous studies
evaluating the integration of deterministic tractography
in surgical navigation have noted an incomplete depic-
tion of the corticospinal tract in these regions [36]. While
some advanced tractography techniques such as high-
angular resolution diffusion imaging, q-ball imaging and
diffusion spectrum imaging can resolve fiber crossings
[37,38], these methods usually require longer scanning
times that may be difficult to achieve in a busy clinical
setting, particularly in patients with brain tumors [39].
In our study, FACT tracts often deviated from the ex-
pected anterior-posterior orientation to veer cortically in
the region of crossing corticobulbar fibers. The probabil-
istic method, on the other hand, was able to track through
this fiber crossing region and reach the expected termi-
nation at Broca’s area. The probabilistic method likely
succeeds by integrating information about trajectory di-
rectionality [20]. In the region of crossing fibers, trajec-
tories stemming from the seed ROIs in Broca and Wer-
nicke’s areas meet and travel in the same direction to-
wards the cortex, and are thus labeled as a merging bun-
dle rather than a true connecting bundle. In our study, all
29 patients had fibers reaching Broca’s area on the nor-

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 197
Table 3. Comparison of results from treated and untreated patients.
No Treatment Treatment
Variable Median (range) Median (range) p-value*
FACT scores (tumor side) 0 (0 - 2) 0 (0 - 1) 0.68
Probabilistic scores (tumor side) 2.5 (0 - 3) 2 (0 - 3) 0.16
Probabilistic tract length (tumor side) 37 (21 - 42) 34 (0 - 45) 0.49
FACT tract length (tumor side) 27.5 (6 - 42) 29 (0 - 44) 0.56
FACT tract volume (tumor side) 291.5 (60 - 916) 306 (0 - 742) 0.59
Probabilistic tract volume (tumor side) 705.5 (156 - 1596) 478 (0 - 857) 0.06
Probabilistic tract mean connectivity (tumor side) 0.04 (0.02 - 0.08) 0.05 (0.02 - 0.38) 0.70
mal side by probabilistic tractography, whereas only 12
(41.7%) did by FACT. The findings on the tumor side of
the brain were similar and implied that the poor results
from FACT on the tumor side were not due solely to the
tumor. The large numbers of trajectories traced by the
probabilistic method per seed point may have helped
compensate for these areas of decreased FA. Our results
support the findings of previous investigators [40], who,
using high-definition fiber tractography requiring more
extensive data collection and computation, successfully
depicted the terminations of the arcuate fasciculus at
Broca’s and Wernicke’s areas.
Recent studies on the incorporation of DTI tractogra-
phy in neurosurgical navigation have predominantly util-
ized deterministic tractography for fiber reconstruction
[41-45]. Many commercial vendor and third-party soft-
ware programs also use deterministic techniques. Previ-
ous investigators have reported underestimation of tracts
by deterministic tractography in regions near tumors [46].
Despite the known limitations of deterministic tracto-
graphy, and the emerging applications of probabilistic
and high-definition fiber techniques, the optimal approach
to tractography reconstruction remains unclear. Further
investigation, including correlation of tractography re-
sults with those of intraoperative stimulation studies, are
necessary.
We found that the tracts generated by FACT tracto-
graphy were smaller in volume in the tumor-affected left
hemisphere than in the normal right hemisphere. Tracts
generated by probabilistic tractography, however, showed
no significant difference in volume between the tumor-
affected and normal sides. Because most individuals are
left language dominant, the left arcuate fasciculus is
normally larger than the right homologue [47,48]. Our
results imply that the probabilistic method was able to
overcome difficulties in tractography such as decreased
FA in tumor regions to generate similar tract volumes
between sides. While thresholding directly affects the
volume of the probabilistic tracts, the tumor side: normal
side tract volume ratio should minimize the confounding
effects of thresholding.
Our study had a few potential limitations. First, it was
a small retrospective study involving a heterogeneous
cohort of patients with different tumor types, some as yet
untreated and the rest previously treated by varying
methods. Surgery, chemotherapy and radiation therapy
have variable effects on white matter [49-53], which may
have affected our tractography results. We attempted to
minimize the potential effects of these treatments by only
including treated patients whose scans were performed at
least 1 month after surgery (median, 19 months) or at
least 5 months after radiation therapy (median, 12.5
months). Although surgery and radiation therapy may
have local or regional effects (as opposed to chemother-
apy that may have more global effects), we found no
differences in anterior termination scores, tract volumes,
or tract lengths between untreated and treated groups.
Second, malignant gliomas are infiltrative tumors known
to invade along white matter tracts and may cause DTI
and tractography abnormalities distinct from metastases
[54-57]. Metastasis, low-grade glioma and meningioma
related edema also do not typically disrupt white matter
tracts in DTI studies [56,58]. In our study, the focus of
the comparison was on probabilistic versus FACT trac-
tography within each patient, rather than between pa-
tients, minimizing the influence of tumor grade on the
results. The small size of our study, however, precluded
comparisons of probabilistic and FACT tractography
results between tumor types.
Third, whereas both methods require setting certain
threshold values (e.g., turning angle for the FACT method
vs. number of walks for the probabilistic method), the
probabilistic method also requires manually adjusting
threshold values on the connectivity map to exclude ex-
traneous tracts. The threshold chosen varied between
patients given the heterogeneity of the tumor pathology,
meaning that a single common threshold could not be
applied on a group-wise basis. However, thresholds were
held constant within each patient on both sides of the
brain. We also performed within-patient analyses using
ratios between sides and between methods to minimize
the effects of these intrinsic differences in the recon-
struct- tion and display of the arcuate fasciculus.
Fourth, the magnetic field strength and gradient en-

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 198
coding directions used in the acquisition of the dataset
varied between 1.5-T and 3.0-T, as well as 15- and 25-
directions. In our study, the comparisons of tracts were
carried out within patients, rather than between patients.
Thus, the differences in field strength and gradient direc-
tions should not confound the results.
Lastly, due to the retrospective nature of the study, in-
traoperative validation was not available to confirm the
reconstructed fibers. Additional work is necessary to
validate the reconstructed probabilistic tracts to confirm
the location, extent, and function of the arcuate fascicu-
lus.
5. CONCLUSION
To the best of our knowledge, this is the first DTI fiber
tractography study to compare outcomes from both FACT
and probabilistic methods in patients with brain tumors.
Probabilistic tractography outperformed the FACT trac-
tography in estimating the extent and degree of connec-
tivity of the arcuate fasciculus language pathway affected
by brain tumors and/or peritumoral abnormalities. We
also found that FACT tended to underestimate the extent
of the anterior-most fibers of the arcuate fasciculus, where
the fibers cross the descending corticobulbar fibers. Lo-
calization of the arcuate fasciculus is relevant for neuro-
surgical planning since direct brain stimulation is less
reliable for white matter tracts than for grey matter cor-
tical structures.
6. ACKNOWLEDGEMENTS
The authors thank Ms. Ada Muellner for her expert medical editing and
advice.
REFERENCES
[1] Breier, J.I., Hasan, K.M., Zhang, W., Men, D. and Pa-
panicolaou, A.C. (2008) Language dysfunction after stroke
and damage to white matter tracts evaluated using diffu-
sion tensor imaging. American Journal of Neuroradiol-
ogy, 29, 483-487. HUdoi:10.3174/ajnr.A0846U
[2] Marchina, S., Zhu, L.L., Norton, A., Zipse, L., Wan, C.Y.
and Schlaug, G. (2011) Impairment of speech production
predicted by lesion load of the left arcuate fasciculus.
Stroke, 42, 2251-2256.
HUdoi:10.1161/STROKEAHA.110.606103U
[3] Yamada, K., Nagakane, Y., Mizuno, T., Hosomi, A., Na-
kagawa, M. and Nishimura, T. (2007) MR tractography
depicting damage to the arcuate fasciculus in a patient
with conduction aphasia. Neurology, 68, 789.
HUdoi:10.1212/01.wnl.0000256348.65744.b2U
[4] Whittle, I.R. and Fraser, D.E. (1991) Resolution of fluent
dysphasia following excision of metastatic carcinoma from
the arcuate fasciculus. British Journal of Neurosurgery, 5,
647-649. HUdoi:10.3109/02688699109002891U
[5] Catani, M., Jones, D.K. and Ffytche, D.H. (2005) Peri-
sylvian language networks of the human brain. Annals of
Neurology, 57, 8-16. HUdoi:10.1002/ana.20319U
[6] Rilling, J.K., Glasser, M.F., Preuss, T.M., Ma, X., Zhao,
T., Hu, X. and Behrens, T.E.J. (2008) The evolution of
the arcuate fasciculus revealed with comparative DTI.
Nature Neuroscience, 11, 426-428. HUdoi:10.1038/nn2072U
[7] Le Bihan, D., Breton, E., Lallemand, D., Grenier, P., Ca-
banis, E. and Laval-Jeantet, M. (1986) MR imaging of
intravoxel incoherent motions: Application to diffusion
and perfusion in neurologic disorders. Radiol ogy , 161,
401-407.
[8] Basser, P.J., Mattiello, J. and LeBihan, D. (1994) MR dif-
fusion tensor spectroscopy and imaging. Biophysic al Jour -
nal, 66, 259-267. HUdoi:10.1016/S0006-3495(94)80775-1U
[9] Moseley, M.E., Cohen, Y., Kucharczyk, J., Mintorovitch,
J., Asgari, H.S., Wendland, M.F., Tsuruda, J. and Nor-
man, D. (1990) Diffusion-weighted MR imaging of ani-
sotropic water diffusion in cat central nervous system.
Radiology, 176, 439-445.
[10] Basser, P.J., Pajevic, S., Pierpaoli, C., Duda, J. and Al-
droubi, A. (2000) In vivo fiber tractography using DT-MRI
data. Magnetic Resonance in Med icine, 44, 625-632.
HUdoi:10.1002/1522-2594(200010)44:4<625::AID-MRM17
>3.0.CO;2-OU
[11] Shimony, J.S., Snyder, A.Z., Conturo, T.E. and Corbetta,
M. (2004) The study of neural connectivity using diffu-
sion tensor tracking. Cortex, 40, 213-215.
HUdoi:10.1016/S0010-9452(08)70955-9U
[12] Mori, S., Crain, B.J., Chacko, V.P. and Van Zijl, P.C.
(1999) Three-dimensional tracking of axonal projections
in the brain by magnetic resonance imaging. Annals of
Neurology, 45, 265-269.
HUdoi:10.1002/1531-8249(199902)45:2<265::AID-ANA21
>3.0.CO;2-3U
[13] Mori, S. and van Zijl, P.C.M. (2002) Fiber tracking: Prin-
ciples and strategies—A technical review. NMR in Bio-
medicine, 15, 468-480. HUdoi:10.1002/nbm.781U
[14] Behrens, T.E.J., Woolrich, M.W., Jenkinson, M., Johan-
sen-Berg, H., Nunes, R.G., Clare, S., Matthews, P.M.,
Brady, J.M. and Smith, S.M. (2003) Characterization and
propagation of uncertainty in diffusion-weighted MR
imaging. Magnetic Resonance in Medicine, 50, 1077-1088.
HUdoi:10.1002/mrm.10609U
[15] Hagmann, P., Thiran, J.-P., Jonasson, L., Vandergheynst,
P., Clarke, S., Maeder, P. and Meuli, R. (2003) DTI map-
ping of human brain connectivity: Statistical fibre track-
ing and virtual dissection. Neuroimage, 19, 545-554.
HUdoi:10.1016/S1053-8119(03)00142-3U
[16] Parker, G.J., Haroon, H.A. and Wheeler-Kingshott, C.A.
(2003) A framework for a streamline-based probabilistic
index of connectivity (PICo) using a structural interpreta-
tion of MRI diffusion measurements. Journal of Mag-
netic Resonance Imaging, 18, 242-254.
HUdoi:10.1002/jmri.10350U
[17] Alexander, A.L., Hasan, K.M., Lazar, M., Tsuruda, J.S.
and Parker, D.L. (2001) Analysis of partial volume ef-
fects in diffusion-tensor MRI. Magnetic Resonance in
Medicine, 45, 770-780. HUdoi:10.1002/mrm.1105U

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 199
[18] Mori, S., Kaufmann, W.E., Davatzikos, C., et al. (2002)
Imaging cortical association tracts in the human brain us-
ing diffusion-tensor-based axonal tracking. Magnetic Re-
sonance in Medicine, 47, 215-223.
HUdoi:10.1002/mrm.10074U
[19] Jellison, B.J., Field, A.S., Medow, J., Lazar, M., Shariar
Salamat, M. and Alexander, A.L. (2004) Diffusion tensor
imaging of cerebral white matter: A pictorial review of
physics, fiber tract anatomy, and tumor imaging patterns.
American Journal of Neurordiology, 25, 356-369.
[20] Kreher, B.W., Schnell, S., Mader, I., Il’yasov, K.A., Hen-
nig, J., Kiselev, V.G. and Saur, D. (2008) Connecting and
merging fibres: Pathway extraction by combining prob-
ability maps. Neuroimage, 43, 81-89.
HUdoi:10.1016/j.neuroimage.2008.06.023U
[21] Saur, D., Kreher, B.W., Schnell, S., Kummerer, D., Kell-
meyer, P., Vry, M.S., Umarova, R., Musso, M., Glauche,
V., Abel, S., Huber, W., Rijntjes, M., Hennig, J. and
Weiller, C. (2008) Ventral and dorsal pathways for lan-
guage. Proceedings of the National Academy of Sciences
of the USA, 105, 18035-18040.
HUdoi:10.1073/pnas.0805234105U
[22] Catani, M., Jones, D.K. and Ffytche, D.H. (2005) Peri-
sylvian language networks of the human brain. Annals of
Neurology, 57, 8-16. HUdoi:10.1002/ana.20319U
[23] Bernal, B. and Altman, N. (2010) The connectivity of the
superior longitudinal fasciculus: A tractography DTI study.
Magnetic Resonance Imaging, 28, 217-225.
HUdoi:10.1016/j.mri.2009.07.008U
[24] Kreher, B., Schnell, S., Hennig, J. and Il’yasov, K. (2006)
DTI calculation, fiber-tracking, and combined evaluation
consolidated in a complete toolbox. Proceedings of the
12th Annual Meeting of the Organization for Human
Brain Mapping, Florence, 11-15 June 2006.
[25] Kreher, B., Henning, J. and Il’yasov, K. (2006) DTI&Fi-
ber-Tools: A complete toolbox for DTI calculation, fiber
tracking, and combined evaluation. Proceeding of ISMRM
14th International Scientific Meeting, Seattle, 6-12 May
2006.
[26] Glasser, M.F. and Rilling, J. (2008) DTI tractography of
the human brain’s language pathways. Cerebral Cortex,
18, 2471-2482. HUdoi:10.1093/cercor/bhn011U
[27] Ciccarelli, O., Behrens, T., Altmann, D.R., Orrell, R.W.,
Howard, R.S., Johansen-Berg, H., Miller, D.H., Matthews,
P.M. and Thompson, A.J. (2006) Probabilistic diffusion
tractography: A potential tool to assess the rate of disease
progression in amyotrophic lateral sclerosis. Brain, 129,
1859-1871. HUdoi:10.1093/brain/awl100U
[28] Guye, M., Parker, G.J.M., Symms, M., Boulby, P., Wheeler-
Kingshott, C.A.M., Salek-Haddadi, A., Barker, G.J. and
Duncan, J.S. (2003) Combined functional MRI and trac-
tography to demonstrate the connectivity of the human
primary motor cortex in vivo. Neuroimage, 19, 1349-
1360. HUdoi:10.1016/S1053-8119(03)00165-4U
[29] Benjamini, Y. and Hochberg, Y. (1995) Controlling the
false discovery rate: A practical and powerful approach to
multiple testing. Journal of the Royal Statistical Society,
Series B (Methodological), 57, 289-300.
[30] Roux, F.E., Boulanouar, K., Ranjeva, J.P., Manelfe, C.,
Tremoulet, M., Sabatier, J. and Berry, I. (1999) Cortical
intraoperative stimulation in brain tumors as a tool to
evaluate spatial data from motor functional MRI. Inves-
tigative Radiology, 34, 225-229.
HUdoi:10.1097/00004424-199903000-00012U
[31] Cosgrove, G.R., Buchbinder, B.R. and Jiang, H. (1996)
Functional magnetic resonance imaging for intracranial
navigation. Neu rosurgery Cli nics of North Ame rica, 7, 313-
322.
[32] Mueller, W.M., Yetkin, F., Hammeke, T.A., Morris, G.L.,
Swanson, S.J., Reichert, K., Cox, R. and Haughton, V.M.
(1996) Functional magnetic resonance imaging mapping
of the motor cortex in patients with cerebral tumors.
Neurosurgery, 39, 515-521.
[33] Geschwind, N. (1970) The organization of language and
the brain. Science, 170, 940-944.
HUdoi:10.1126/science.170.3961.940U
[34] Bernal, B. and Ardila, A. (2009) The role of the arcuate
fasciculus in conduction aphasia. Brain, 132, 2309-2316.
HUdoi:10.1093/brain/awp206U
[35] Mori, S. and van Zijil, P.C.M. (2002) Fiber tracking:
Principles and strategies—A technical review. NMR Bio-
medicine, 15, 468-480. HUdoi:10.1002/nbm.781U
[36] Berman, J.I., Berger, M.S., Mukherjee, P. and Henry, R.G.
(2004) Diffusion-tensor imaging-guided tracking of fibers
of the pyramidal tract combined with intraoperative cor-
tical stimulation mapping in patients with gliomas. Jour-
nal of Neurosurgery, 101, 66-72.
HUdoi:10.3171/jns.2004.101.1.0066U
[37] Tuch, D.S., Reese, T.G., Wiegell, M.R and Wedeen, V.J.
(2003) Diffusion MRI of complex neural architecture.
Neuron, 40, 885-895.
HUdoi:10.1016/S0896-6273(03)00758-XU
[38] Wedeen, V., Reese, T.G., Tuch, D.S., Weigel, M.R., Dou,
J.G., Weiskoff, R.M. and Chessler, D. (2000) Mapping
fiber orientation spectra in cerebral white matter with fou-
rier-transform diffusion MR [abstract]. Proceedings of
the 8th Meeting of the International Society for Magnetic
Resonance in Medicine, Berkeley, 82.
[39] Ciccarelli, O., Catani, M., Johansen-Berg, H., Clark, C.
and Thompson, A. (2008) Diffusion-based tractography
in neurological disorders: Concepts, applications, and fu-
ture developments. The Lancet Neurology, 7, 715-727.
HUdoi:10.1016/S1474-4422(08)70163-7U
[40] Fernandez-Miranda, J.C., Pathak, S. and Schneider, W.
(2010) High-definition fiber tractography and language.
Journal of Neurosurgery, 113, 156-157.
[41] Nimsky, C., Ganslandt, O., Merhof, D., Sorensen, A.G.
and Fahlbusch, R. (2006) Intraoperative visualization of
the pyramidal tract by diffusion tensor-imaging-based fi-
ber tracking. Neuroimage, 30, 1219-1229.
HUdoi:10.1016/j.neuroimage.2005.11.001U
[42] Yu, C.S., Li, K.C., Xuan, Y., Ji, X.M. and Qin, W. (2005)
Diffusion tensor tractography in patients with cerebral
tumors: A helpful technique for neurosurgical planning
and postoperative assessment. Radiology, 56, 197-204.
[43] Okada, T., Mikuni, N., Miki, Y., Kikuta, K., Urayama, S.,
Hanakawa, T., Fushimi, Y., Yamamoto, A., Kanagaki, M.,

Z. X. Li et al. / J. Biomedical Science and Engineering 6 (2013) 192-200 200
Kukuyama, H., Hashimoto, N. and Togashi, K. (2006)
Corticospinal tract localization: Integration of diffusion-
tensor tractography at 3-T MR Imaging with intraopera-
tive white matter stimulation mapping-preliminary results.
Radiology, 240, 849-857. HUdoi:10.1148/radiol.2403050916U
[44] Nimsky, C., Ganslandt, O., Hastreiter, P., Wang, R., Ben-
ner, T., Sorensen, A.G. and Fahlbusch, R. (2005) Preop-
erative and intraoperative diffusion tensor imaging-based
fiber tracking in glioma surgery. Neurosurgery, 56, 130-
137.
[45] Berman, J.I., Berger, M., Chung, S.W., Nagarajan, S.S.
and Henry, R.G. (2007) Accuracy of diffusion tensor
magnetic resonance imaging tractography assessed using
intraoperative subcortical stimulation mapping and mag-
netic source imaging. Journal of Neurosurgery, 107 , 488-
494. HUdoi:10.3171/JNS-07/09/0488U
[46] Kinoshita, M., Yamada, K., Hashimoto, N., Kato, A., Izu-
moto, S., Baba, T., Maruno, M., Nishimura, T. and Yo-
shimine, T. (2005) Fiber-tracking does not accurately es-
timate size of fiber bundle in pathological condition: Ini-
tial neurosurgical experience using neuronavigation and
subcortical white matter stimulation. Neuroimage , 25, 424-
429. HUdoi:10.1016/j.neuroimage.2004.07.076U
[47] Parker, G.J., Luzzi, S., Alexander, D.C., Wheeler-King-
shott, C.A., Ciccarelli, O. and Lambon Ralph, M.A. (2005)
Lateralization of ventral and dorsal auditory-language
pathways in the human brain. Neuroimage, 24, 656-666.
HUdoi:10.1016/j.neuroimage.2004.08.047U
[48] Nucifora, P.G.P., Verma, R. and Melhem, E.R. (2005)
Leftward asymmetry in relative fiber density of the arcu-
ate fasciculus. Neuroreport, 16, 791-794.
HUdoi:10.1097/00001756-200505310-00002U
[49] Chapman, C.H., Nagesh, V., Sundgren, P.C., et al. (2012)
Diffusion tensor imaging of normal-appearing white mat-
ter as biomarker for radiation-induced late delayed cogni-
tive decline. International Journal of Radiation Oncology,
Biology and Physics, 82, 2033-2040.
HUdoi:10.1016/j.ijrobp.2011.01.068U
[50] Welzel, T., Niethammer, A., Mende, U., et al. (2008) Dif-
fusion tensor imaging screening of radiation-induced
changes in the white matter after prophylactic cranial ir-
radiation of patients with small cell lung cancer: First re-
sults of a prospective study. American Journal of Neuro-
radiology, 29, 379-383. HUdoi:10.3174/ajnr.A0797U
[51] Mabbott, D.J., Noseworthy, M.D., Bouffet, E., Rockel, C.
and Laughlin, S. (2006) Diffusion tensor imaging of white
matter after cranial radiation in children for medulloblas-
toma: Correlation with IQ. Neuro-Oncology, 8, 244-252.
HUdoi:10.1215/15228517-2006-002U
[52] Abraham, J., Haut, M.W., Moran, M.T., Filburn, S., Le-
miuex, S. and Kuwabara, H. (2008) Adjuvant chemo-
therapy for breast cancer: effects on cerebral white matter
seen in diffusion tensor imaging. Clinical Breast Cancer,
8, 88-91. HUdoi:10.3816/CBC.2008.n.007U
[53] Sijens, P.E., Heesters, M.A., Enting, R.H., et al. (2007)
Diffusion tensor imaging and chemical shift imaging as-
sessment of heterogeneity in low grade glioma under te-
mozolomide chemotherapy. Cancer Investigation, 25, 706-
710. HUdoi:10.1080/07357900701564028U
[54] Matsukado, Y., MacCarty, C.S. and Kernohan, J.W. (1961)
The growth of glioblastoma multiforme (astrocytomas,
grades 3 and 4) in neurosurgical practice. Journal of
Neurosurgery, 18, 636-644.
HUdoi:10.3171/jns.1961.18.5.0636U
[55] Johnson, P.C., Hunt, S.J. and Drayer, B.P. (1989) Human
cerebral gliomas: Correlation of postmortem MR imaging
and neuropathologic findings. Radiology, 170, 211-217.
[56] Price, S.J., Burnet, N.G., Donovan, T., Green, H.A., Pena,
A., Antoun, N.M., Pickard, J.D., Carpenter, T.A. and Gil-
lard, J.H. (2003) Diffusion tensor imaging of brain tu-
mours at 3T: A potential tool for assessing white matter
tract invasion? C linical Radiology, 58, 455-462.
HUdoi:10.1016/S0009-9260(03)00115-6U
[57] Kleihues, P. and Cavenee, W.K. (2000) Pathology and
genetics of tumours of the nervous system. IARC Press,
Lyon.
[58] Wieshmann, U.C., Symms, M., Parker, G.J.M., Clark, C.,
Lemieux, L., Barker, G. and Shorvon, S. (2000) Diffusion
tensor imaging demonstrates deviation of fibres in normal
appearing white matter adjacent to a brain tumour. Jour-
nal of Neurology, Neurosurgery & Psychiatry, 68, 501-
503. HUdoi:10.1136/jnnp.68.4.501U