Content uploaded by Ines Debove
Author content
All content in this area was uploaded by Ines Debove on Sep 25, 2020
Content may be subject to copyright.
Contents lists available at ScienceDirect
NeuroImage: Clinical
journal homepage: www.elsevier.com/locate/ynicl
Accuracy of different three-dimensional subcortical human brain atlases for
DBS –lead localisation
Andreas Nowacki
a,⁎,1
, T.A-K. Nguyen
a,1
, Gerd Tinkhauser
b,c
, Katrin Petermann
b
, Ines Debove
b
,
Roland Wiest
d
, Claudio Pollo
a
a
Department of Neurosurgery, Inselspital, University Hospital Bern, and University of Bern, Bern, Switzerland
b
Department of Neurology, Inselspital, University Hospital Bern, and University of Bern, Bern, Switzerland
c
Medical Research Council Brain Network Dynamics Unit and Nuffield Department of Clinical Neurosciences, University of Oxford, United Kingdom
d
Department of diagnostic and interventional Neuroradiology, Inselspital, University Hospital Bernand University of Bern, Bern, Switzerland
ARTICLE INFO
Keywords:
Deep brain stimulation
Lead localisation
Human brain atlas
MNI space
ABSTRACT
Background: Accurate interindividual comparability of deep brain stimulation (DBS) lead locations in relation to
the surrounding anatomical structures is of eminent importance to define and understand effective stimulation
areas. The objective of the current work is to compare the accuracy of the DBS lead localisation relative to the
STN in native space with four recently developed three-dimensional subcortical brain atlases in the MNI tem-
plate space. Accuracy is reviewed by anatomical and volumetric analysis as well as intraoperative electro-
physiological data.
Methods: Postoperative lead localisations of 10 patients (19 hemispheres) were analysed in each individual
patient based on Brainlab software (native space) and after normalization into the MNI space and application of
4different human brain atlases using Lead-DBS toolbox within Matlab (template space). Each patient's STN was
manually segmented and the relation between the reconstructed lead and the STN was compared to the 4 atlas-
based STN models by applying the Dice coefficient. The length of intraoperative electrophysiological STN ac-
tivity along different microelectrode recording tracks was measured and compared to reconstructions in native
and template space. Descriptive non-parametric statistical tests were used to calculate differences between the 4
different atlases.
Results: The mean STN volume of the study cohort was 153.3 ± 40.3 mm3 (n= 19). This is similar to the STN
volume of the DISTAL atlas (166 mm3; p= .22), but significantly larger compared to the other atlases tested in
this study. The anatomical overlap of the lead-STN-reconstruction was highest for the DISTAL atlas
(0.56 ± 0.18) and lowest for the PD25 atlas (0.34 ± 0.17). A total number of 47 MER trajectories through the
STN were analysed. There was a statistically significant discrepancy of the electrophysiogical STN activity
compared to the reconstructed STN of all four atlases (p< .0001).
Conclusion: Lead reconstruction after normalization into the MNI template space and application of four
different atlases led to different results in terms of the DBS lead position relative to the STN. Based on elec-
trophysiological and imaging data, the DISTAL atlas led to the most accurate display of the reconstructed DBS
lead relative to the DISTAL-based STN.
1. Introduction
Deep Brain Stimulation (DBS) is a surgical treatment option to al-
leviate symptoms of movement disorders such as Parkinson's Disease
(PD) (Deuschl et al., 2006;Schuepbach et al., 2013). Different
subcortical targets within the basal ganglia and thalamus are typically
chosen for DBS lead placement. For instance, the Subthalamic Nucleus
(STN) is a common target for DBS of PD. Evidence suggests accurate
lead placement to be of paramount importance for postoperative out-
come (Welter et al., 2014;Wodarg et al., 2012). Therefore, exact
https://doi.org/10.1016/j.nicl.2018.09.030
Received 6 May 2018; Received in revised form 17 August 2018; Accepted 25 September 2018
Abbreviations: Deep Brain Stimulation, (DBS); microelectrode recording, (MER); Parkinson's Disease, (PD); Subthalamic Nucleus, (STN); MNI152, (Montreal
Neurological Institute 152)
⁎
Corresponding author at: Department of Neurosurgery, Inselspital, University Hospital Bern, and University of Bern, Bern, Switzerland.
1
The first two authors contributed equally to this work.
E-mail address: neuro.nowacki@gmail.com (A. Nowacki).
NeuroImage: Clinical 20 (2018) 868–874
Available online 27 September 2018
2213-1582/ © 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T
localisation of the DBS lead within the target and in relationship to
surrounding anatomical structures may be helpful to analyse the un-
derlying reason for poor stimulation efficacy. Moreover, with advanced
lead design offering directional stimulation and current steering, pro-
gramming can be adjusted according to the DBS lead location and the
intended target structure to minimize stimulation induced side-effects
(Pollo et al., 2014;Schupbach et al., 2017;Tinkhauser et al., 2018).
Apart from DBS lead localisation in an individual patient (native space),
comparison of DBS lead positions across subjects is an important and
powerful tool for scientific approaches, e.g., to define efficacious sti-
mulation areas (“sweet spots”) within well-known target structures
(Accolla et al., 2016;Akram et al., 2017).
The undertaking of comparing anatomical data across subjects is
challenging because of anatomical heterogeneity. Both position and
sizes of anatomical structures in one brain must correspond to positions
and sizes in another to make meaningful comparisons (Brett et al.,
2002). With advances in imaging technologies and deformable regis-
tration algorithms, projection of DBS leads onto three-dimensional
subcortical atlases has been established more recently to compare DBS
lead locations of different subjects (Cheung et al., 2014;Welter et al.,
2014). The basic underlying workflow usually consists of registering
brains of individuals to a common template by applying non-rigid,
deformable registration algorithms (normalization into template space)
(Klein et al., 2009). Currently, the MNI152 (Montreal Neurological
Institute) space, which is based on anatomical high-resolution images of
152 subjects, is broadly used as a template space. Implementation of
brain-atlases into the template space then enables further anatomical
analyses and comparison of interindividual data. Today, many different
atlases that have been defined on the basis of magnetic resonance
imaging (MRI) and/or histology are available (Chakravarty et al., 2006;
Welter et al., 2014;Xiao et al., 2015;Yelnik et al., 2007). A major
drawback of many atlases is that their spatial definition of anatomical
structures does not correspond inherently to the MNI template (sum-
marised in Ewert et al. (Ewert et al., 2017)). However, this is a crucial
step for the analysis of lead locations in the context of surrounding
structures. Inaccurate registrations will ultimately lead to wrong con-
clusions about DBS lead placements and the stimulated structures and
can therefore render group comparisons erroneous. Another drawback
is that most atlases are based on young and healthy individuals,
whereas the study groups of interest in the field of DBS are usually old
and their brains atrophic (Dalaker et al., 2011). To overcome these
limitations, different groups have recently presented atlases that match
the MNI space or which are based on PD patients instead of healthy
subjects (Ewert et al., 2017;Pauli et al., 2017;Wang et al., 2016;Xiao
et al., 2015).
The objective of the current work is to compare the accuracy of the
DBS lead location relative to the STN in native space with four recently
developed three-dimensional subcortical brain atlases in the MNI space.
Accuracy is reviewed by anatomical and volumetric analysis as well as
intraoperative electrophysiological data.
2. Material and Methods
2.1. Patients
We retrospectively analysed 24 patients undergoing DBS lead im-
plantation into the STN for advanced PD between October 2015 and
April 2017 at our department. Inclusion criteria were age older than
18 years, availability of intraoperative electrophysiological testing re-
sults by specialized movement disorder neurologists, microelectrode
recording results available along at least two parallel trajectories of
each patient's hemisphere. All patients provided written prior consent
for the procedure. The study was approved by the institutional review
board.
2.2. Operative procedure
For detailed description of our targeting approach and operative
procedure we refer to previously published reports of our group
(Nowacki et al., 2017). In summary, each patient underwent pre-
operative MRI to target the STN. Imaging was performed with a 12-
channel head coil for signal reception on a 3 T MRI system (MAGNE-
TOM Trio™Tim, Siemens, Germany). Imaging included T2-weighted
sequences with an echo time of 380 ms, a repetition time of 3000 ms
and a slice thickness of 1 mm. Targets and trajectories were identified
using Brainlab Elements software (Brainlab AG, Germany). On the day
of surgery a Leksell G frame (Elekta instruments, Sweden) was placed
and a high-resolution, stereotactic CT scan was performed and co-re-
gistered with the preoperative MRI (Brainlab AG, Germany).
Intraoperative microelectrode recording (MER) was performed as
Fig. 1. Determination of STN lengths in an example patient set. (A) Image-based measurements with Lead DBS and microelectrode tool that places microelectrodes in
specific trajectories and allows for measurement of STN length (blue points representing entry and exit, entry point for central trajectory covered by lead). The panel
shows the DISTAL atlas. GPe –external globus pallidus, GPi –internal globus pallidus, STN –subthalamic nucleus, RN –red nucleus. (B) Intraoperative assessment of
electrophysiological recording at our centre. Number of ‘+’signs indicate STN activity rated by one experienced rater live intraoperatively at different positions of
the microelectrode.
A. Nowacki et al. NeuroImage: Clinical 20 (2018) 868–874
869
described previously. Here, recording was performed on two to three
parallel trajectories. From 10 mm to 5 mm above the target, electro-
physiological activity was recorded in 1 mm-steps and from 5 mm
above the target in 0.5 mm-steps. Electrophysiological activity was
evaluated by a simple grading system, ‘+’for little activity to ‘+++’
for strong activity, to determine entry and exit of the STN (Fig. 1). This
grading was done live intraoperatively by an experienced rater and not
ad-hoc nor in a blinded fashion.
Postoperative high-resolution CT was performed for assessment of
correct electrode placement on the day after surgery.
2.3. Postoperative DBS lead location
2.3.1. Native space
Postoperative DBS lead localisation was performed based on
Brainlab Elements software in patient's native space. AC-PC coordinates
of the intended target, the DBS lead tip as well as the targeting error
were calculated as described previously (Nowacki et al., 2015). The
integrated workflow was used to fuse preoperative MR images and post-
operative CT images. The STN was manually segmented by the junior
(AN) and senior neurosurgeon (CP) according to the hypointense signal
based on T2-weighted images. The DBS lead was reconstructed ac-
cording to the CT artefact in an automated way by the Brainlab soft-
ware and its position manually adjusted, if necessary.
2.3.2. Normalized Template space
Lead-DBS toolbox version 2.0.0.6 was used within Matlab 2016b
(The MathWorks, USA) for DBS lead visualisation (Horn and Kuhn,
2015). Preoperative MRI and postoperative CT were co-registered using
SPM12 and Advanced Normalization Tools (Avants et al., 2008). Nor-
malization to the MNI 152 2009b space (Montreal Neurological In-
stitute) was also performed by applying Advanced Normalization Tools.
The algorithm's accuracy and effectiveness were evaluated elsewhere
(Klein et al., 2009). The lead was then automatically pre-localised on
CT images with PaCER and manually adjusted, if the automatic loca-
lisation did not match the CT artefact (typical adjustments of less than
half of the lead's diameter or 0.65 mm) (Husch et al., 2018). Lead-DBS
provided a three-dimensional display of nuclei according to a selected
atlas. Four atlases were evaluated: the DISTAL atlas combining MRI
from the 2009b MNI152 template (normative young adult population),
histology and structural connectivity data (Ewert et al., 2017); the
CIT168 atlas derived from data of the Human Connectome Project
based on 168 young adult subjects between the age of 22 and 35 (Pauli
et al., 2017); the Wang atlas based on high-resolution 7 T MR scans of
twelve healthy subjects; and the PD25 atlas derived from 3 T MRI scans
of 25 Parkinson's disease patients (Xiao et al., 2015). Our rationale
behind the selection of these four atlases was to cover a broad spectrum
of modalities, e.g., different MRI data, histology and subject groups.
2.4. Anatomical comparison of STN in native and template space
To assess the accuracy of the four different brain atlases, we com-
pared each patient's individual STN anatomy based on the hypointense
T2-artefact on their MRI and the postoperative DBS lead location (na-
tive space) with the reconstructed DBS lead and STN according to the
four brain atlases after normalization to the MNI space (template
space).
Furthermore, we compared the volumes of each patient's in-
dividually segmented STN (left and right) according to the MRI-based
T2 artefact and the atlas-based STN volume (supplementary fig. 1). In
Lead-DBS, the segmented STNs according to the four atlases were ex-
ported and then analysed in Matlab 2016b to compute the volume.
The Dice coefficient was applied as a measure of the degree of si-
milarity between the native space STN and the atlas-based STN. It was
calculated according to Eq. (1), where A and B are the STN volumes in
native and template space, respectively.
=∣∩∣
+∣ ∣
Dice AB
AB
2
|| (1)
2.5. STN MER-MRI trajectory length comparison
One further pertinent way to assess the accuracy of the re-
constructed DBS lead relative to STN anatomy was to compare the
position of the DBS lead with intraoperative MER results. Therefore, we
determined the length of electrophysiological STN activity along the
two to three intraoperatively used parallel trajectories and compared it
to image-based reconstructed trajectories across the STN according to
the four atlases (template space) as well as to the T2-hypointense signal
of each patient's STN visualised on MRI (native space).
Native-space based analysis was performed with Brainlab software.
Reconstructed trajectories of MER adjacent to the DBS lead were
manually placed in inline view with a fixed 2 mm distance to the lead
according to known features of the five-channel Ben-Gun. The entry and
exit points of the STN were then measured manually in the software to
calculate the resulting STN length. Lead-DBS provided a microelectrode
tool to determine STN lengths along the different microelectrode tra-
jectories such as central, lateral or medial (Fig. 1).
We then calculated the difference of the image- or atlas-based STN
trajectory length (S
MRI/atlas
) and the MER-based trajectory length (S
MER
)
in each patient (ΔS=S
MRI/atlas
-S
MER
). If no electrophysiological ac-
tivity was recorded for a given trajectory, S
MER
was set to zero.
Similarly, if the image- or atlas-based trajectory was projected outside
the STN, S
MRI/atlas
was set to zero for that trajectory.
2.6. Clinical outcome assessment
Preoperative motor subscores of the Unified Parkinson's Disease
Rating Scale (UPDRS part III) in “medication-off”state and post-
operative UPDRS III scores in the “stimulation-on, medication-off”state
(overnight withdrawal of PD medication) were systematically de-
termined 12 months after DBS implantation. DBS outcome was assessed
by analysing the differences of respective preoperative and post-
operative UPDRS III scores.
2.7. Statistical analysis
For both anatomical and trajectory length comparisons, measure-
ments were done separately for the left and right hemispheres to ac-
count for asymmetries. Finally, data were then analysed together with
descriptive/nonparametric statistics using Prism software (GraphPad
Prism 6, USA). One-way ANOVA (Friedman test) was applied to test for
statistical differences between the image-based and electro-
physiological-based STN lengths. Post-hoc analysis was performed by
Dunn's multiple comparisons test. A P-value < .05 was considered
statistically significant.
3. Results
According to our inclusion criteria we included a total number of
ten patients (19 hemispheres) in the final analysis. Altogether, a total
number of 47 MER trajectories through the STN were analysed (cf.
Table I, 10 along a medial trajectory, 19 along a central trajectory, 18
along a lateral trajectory). The mean targeting error was
1.01 ± 0.57 mm (range 0.40–1.95 mm) indicating precise targeting
and no cases of clinically relevant lead displacement.
3.1. Clinical outcome data
Preoperative mean UPDRS III score was 36.9 ± 9.75. Postoperative
mean UPDRS III score was significantly reduced to 10.55 ± 6.40
(p< .0001). The mean percentage improvement was
A. Nowacki et al. NeuroImage: Clinical 20 (2018) 868–874
870
71.72 ± 14.96%. One patient was lost to clinical follow-up as he died
(unrelated to DBS surgery).
3.2. DBS lead relative to STN anatomy in native and template space
The mean AC-PC coordinates of the right lead tip were LAT
11.40 ± 1.10 mm, ANT -3.31 ± 0.39 mm and VERT
-4.43 ± 0.55 mm. After transformation of the DBS leads into the MNI
space the mean MNI coordinates were LAT 9.38 ± 1.33 mm, AP
-4.44 ± 0.83 mm and VERT -9.35 ± 1.2 mm. Table 1 summarizes
patients´ individual DBS lead tip position relative to the MCP in native
space and after normalization in the MNI space.
Electrode reconstruction after normalization and application of the
four different atlases led to different results with respect to the lead
position within the STN and compared to the native space. Fig. 2 vi-
sualises the DBS lead and STN delineation of the manually segmented
STN compared to the different atlases in one representative patient of
the study cohort. In general, we could observe that the DBS lead was
projected more laterally within the STN after applying the DISTAL and
CIT atlas in comparison to the native space. No such shift could be
observed in case of the PD25 and Wang atlases, in which cases, how-
ever, the DBS lead was frequently located outside the atlas-based STN
borders.
The mean STN volume of the study cohort was 153.3 ± 40.3 mm
3
(n= 19). This is similar to the STN volume of the DISTAL atlas
(166 mm
3
;p= .22), but significantly larger compared to the CIT atlas
(136 mm
3
;p= .04), the atlas of Wang et al. (85 mm
3
;p< .001) and
the PD25 atlas (102 mm
3
; p < .001).
The Dice coefficient of the native space STN volume compared to
each of the different atlas-based STN was determined in each patient's
hemisphere. The overlap was highest for the DISTAL atlas
(0.56 ± 0.18). Less similar results were found for the CIT atlas
(0.50 ± 0.15). Furthermore, we found significant differences for the
atlas of Wang et al. (0.42 ± 0.12) as well as the PD25 (0.34 ± 0.17).
3.3. STN MER-MRI trajectory comparison in native space and template
space
Based on all three trajectories, there was no significant difference
between ΔS in native space, however there was a statistically significant
difference of ΔS for all four atlases (p< .0001).
Fig. 3 shows the differences of the electrophysiological and image-
based STN trajectory lengths separately for the central, medial and
lateral trajectories.
The mean differences (ΔS) between the reconstructed MRI-based
trajectory length in native space and intraoperatively measured MER-
based STN length was 0.49 ± 0.91 mm for the central trajectory,
0.45 ± 2.10 mm for the medial and −0.97 ± 2.99 mm for the lateral
trajectory.
The mean differences (ΔS) between the reconstructed DISTAL atlas-
based STN length and MER-based STN length was −0.17 ± 1.68 mm
for the central trajectory, 0.33 ± 2.75 mm for the medial
Table 1
Coordinates of right lead tip in native and MNI space with reference to the mid-commissural point (MCP).
ID MCP Coordinates in native space MCP coordinates in MNI space
LAT AP VERT LAT AP VERT
1 9.49 −3.73 −5.73 7.49 −3.68 −10.57
2 11.47 −3.63 −4.59 8.44 −5.82 −10.73
3 12.93 −3.65 −3.98 11.17 −5.2 −7.33
4 11.74 −3.51 −4 9.95 −4.4 −8.44
5 12.02 −3.03 −4.12 10.72 −4.28 −8.14
6 12.38 −3.41 −4.4 10.02 −5.05 −10.72
7 10.43 −2.6 −4.78 9.67 −3.35 −9.83
8 12.04 −3.6 −3.92 9.18 −4.26 −9.17
9 10.03 −2.91 −4.67 7.11 −3.36 −10.02
10 11.51 −3.01 −4.15 10.09 −5.06 −8.55
mean 11.40 −3.31 −4.43 9.38 −4.44 −9.35
SD 1.10 0.39 0.55 1.33 0.83 1.20
Fig. 2. Lead position in relation to STN. (A) Postoperative CT lead artefact (red) superimposed on preoperative MR T2 images. (B)-(E) Right lead marked as asterisk
in atlas segmented STN for DISTAL (B), CIT (C), Wang (D) and PD25 (E), respectively.
A. Nowacki et al. NeuroImage: Clinical 20 (2018) 868–874
871
and −2.58 ± 2.40 mm for the lateral trajectory. Frequently, the lat-
eral microelectrode trajectory was displayed outside the DISTAL atlas-
based STN, while there was marked electrophysiological activity re-
corded intraoperatively. This led to negative ΔS-values on the lateral
trajectory in those cases. On the other hand medial trajectories were
displayed within the DISTAL atlas-based STN, even though no electro-
physiological activity was measured (Fig. 4).
The mean differences (ΔS) between the reconstructed CIT atlas-
based STN length and MER-based STN length was −0.93 ± 1.50 mm
for the central trajectory, −1.13 ± 1.99 mm for the medial
and −2.67 ± 2.21 mm for the lateral trajectory.
The mean differences (ΔS) between the reconstructed Wang atlas-
based STN length and MER-based STN length was −2.42 ± 1.98 mm
for the central trajectory, −0.7 ± 2.56 mm for the medial
and −2.71 ± 2.28 mm for the lateral trajectory.
The mean differences (ΔS) between the reconstructed PD25 atlas-
based STN length and MER-based STN length was −1.62 ± 1.90 mm
for the central trajectory, −1.45 ± 2.38 mm for the medial
and −2.94 ± 2.05 mm for the lateral trajectory.
These difference calculations were repeated in native space by
transforming the four atlases from MNI space to native patient space.
There we also observed the smallest differences for the DISTAL atlas
(supplementary fig. 2).
Fig. 3. Differences in measured STN lengths between manually segmented, atlas based STN and intraoperative electrophysiology data. Thick red lines indicate mean;
red boxes represent the 95% confidence interval of the mean; blue box represents the standard deviation; gray dots represent individual data.
Fig. 4. Overlay of manually segmented STN (orange) and atlas based STN according to Distal (green), CIT (yellow), Wang (rose) and PD25 (gray) of an example
patient. The lead tip was used as anchor to align manually segmented and atlas-based STNs.
A. Nowacki et al. NeuroImage: Clinical 20 (2018) 868–874
872
4. Discussion
Our study results show that lead reconstruction after normalization
into the MNI template space and application of four different atlases led
to different results in terms of the DBS lead position relative to the STN.
The STN volume and delineation of the four different atlases as well as
the relation to the reconstructed DBS lead were compared to the T2-
hypointense MRI signal in each individual patient and to the STN
electrophysiological profile recorded during intraoperative MER. We
demonstrated that the process of normalization into the MNI template
space and application of any of the four tested atlases resulted in sig-
nificant differences between the electrophysiological profile and the
atlas-based profile of the STN. Furthermore, significant differences were
found between atlases regarding volume and similarity index compared
to native space. Based on electrophysiological and imaging data, the
DISTAL atlas led to the most accurate display of the reconstructed DBS
lead relative to the DISTAL-based STN. Notably, after applying the
DISTAL atlas the reconstructed DBS leads were displayed too lateral
within the STN compared to native space and to the electro-
physiological profile.
Accurate display of the DBS lead in relation to the surrounding
anatomical structures is of paramount importance in clinical routine as
well as for scientific purposes such as cohort comparisons.
Understanding the relation of the lead to the surrounding anatomical
structures may help clinicians to adapt stimulation parameters to avoid
side effects and optimize stimulation. Importantly, given the fact that
the target structures for DBS are usually small in size with only a few
millimetres of diameter, accurate DBS lead reconstruction is crucial
when it comes to comparison of DBS lead placements across subjects to
infer optimal stimulation sites (Horn et al., 2017). Three-dimensional
subcortical human brain atlases (Cheung et al., 2014;Welter et al.,
2014) in conjunction with computational models of volume of tissue
activated or fibre tractography are helpful tools to address these
questions (Gunalan et al., 2018). Different subcortical atlases have been
developed to enable microstructural analysis of brain anatomy. Some
atlases are based on histologic data of single or multiple brains, while
others are defined solely through MRI data. Some more recent atlases,
such as the DISTAL atlas studied herein, combine several modalities to
improve accuracy (Ewert et al., 2018).
From each patient's data set, we concluded that the DBS lead arte-
fact on postoperative CT reflected the actual lead position within the
STN reliably as we found no difference between the reconstructed
image-based trajectories through the STN and their corresponding
electrophysiological activity. This is in line with previous work by our
and other groups (Hamani et al., 2005;Kocabicak et al., 2013;Nowacki
et al., 2017;Schlaier et al., 2011).
On the other hand, normalization of individual MRI patient data
into the MNI space and application of the four atlases tested inevitably
led to certain shifts of the reconstructed DBS lead in relation to the STN
anatomy. One factor is the normalization to the MNI space. A com-
parison of patients lead tip position with reference to MCP in native
space and after normalization into the MNI template space shows a
difference in all coordinates especially in the vertical direction. A
thorough discussion and probabilistic conversion of AC/PC coordinates
to MNI coordinates is provided by Horn et al. (Horn et al., 2017). In
addition, MNI space defines the origin close to the AC point, not quite
on the AC-PC line, but a few millimetres more inferior.
Comparison between atlas-based reconstructed MER trajectories
through the STN showed significant differences compared to the elec-
trophysiological activity. However, the degree of discrepancy between
imaging and electrophysiological data was considerably different be-
tween the four atlases tested. Whereas there was a good correspondence
of electrophysiological activity and the reconstructed MER trajectory
through the STN of the DISTAL and CIT atlases along the central tra-
jectory, there was a low correspondence in case of the Wang and PD25
atlas along each of the tested trajectories. It is important to note that,
frequently, lateral trajectories were displayed outside the DISTAL and
CIT atlas STN, while there was marked electrophysiological activity
leading to negative ΔS-values. On the other hand, medial trajectories
were displayed within the STN even though no electrophysiological
activity was measured. It highlights the tendency of a lateral shift of the
reconstructed DBS lead relative to all four atlas-based STN. This shift
might be due to the normalization algorithm, specifically as the algo-
rithm transforms brain images of Parkinson's disease patients –which
typically show ventricular enlargement but smaller basal ganglia
structures than normal controls –to the MNI template brain (Geng
et al., 2006). This warrants further investigation in order to adapt the
normalization algorithm for DBS research.
Our data further showed that the DISTAL atlas-based STN volume
corresponded well to the mean STN volume in our cohort of patients.
Application of the other atlases led to an underestimation of the STN
volume, despite the fact that the volume of the MNI template brain
tends to be larger than the average human brain (Horn et al., 2017).
Similarity analysis revealed the best correspondence between the STN
shape of the DISTAL atlas compared to each patient's individual STN.
Poor anatomical overlap was found between the Wang and PD25 atlas
STN and our cohort of patients.
Our study has some limitations. First, our relatively small sample
size of ten patients that we have included in the final analysis prevents
generalizing our findings to a broader population. Second, we only
analysed four different atlases. Literature now abounds with different
subcortical brain atlases. Analysing all of them was beyond the scope of
this study. We focused on the four presented atlases here to compare
and contrast the multimodality of the DISTAL atlas, the high resolution
and high subject count (168) of the CIT atlas, the 7 T based Wang atlas
and the disease-specific PD25 atlas based on Parkinson's disease pa-
tients. While the DISTAL atlas was specifically defined in the MNI space
used herein, the other three atlases were each defined on custom
templates. The CIT atlas was based on data from the Human
Connectome Project, the 7 T based atlas from a groupwise template
built from the data of the twelve healthy subjects scanned and the PD25
atlas was defined as an averaged atlas across the 25 Parkinson's disease
patients. These three atlases were transformed to the MNI template
space and the transformed registrations were provided within LEAD-
DBS. This additional transformation may have affected the comparison.
However, for better comparison we specifically intended to analyse all
four atlases in the same template space. Third, analysis of the STN size
and shape in native space was based on manual segmentation of the T2-
hypointense STN signal based on MRI ignoring the fact that different
MRI sequences have been proposed to display the STN, such as sus-
ceptibility weighted sequences or fluid attenuated inversion recovery
sequences (Chandran et al., 2016;Heo et al., 2015;Senova et al., 2016).
Furthermore, delineation of the STN borders by imaging is not always
clear, which has been described previously by our and other groups
(Nowacki et al., 2017;Starr et al., 2002). Especially the anterior-pos-
terior axis of the STN was found to be prone to MRI distortions
(Sumanaweera et al., 1994). Similarly, delineation of the STN with
electrophysiological data was rated live intraoperatively by a single
experienced rater and not in a blinded retrospective fashion with
multiple raters or with an automised algorithm (Zaidel et al., 2010).
Fourth, our analysis focused on STN length and did not study entry and
exit points in detail that could be prone to error by different lead lo-
calization algorithms and manual corrections. Finally, we relied on
previous evaluation of registration and normalization algorithms
15
, but
we think that further improvements in this domain may be warranted.
Furthermore, we performed an early post-operative CT scan as part of
the clinical routine to rule out complications. Due to resorption of
possible postoperative pneumencephalus, the lead position might
slightly change over the postoperative period and introduce a potential
bias into DBS lead position analysis. Furthermore, we did not measure
the electrophysiological activity along an anterior or posterior trajec-
tory, which would have added great value to evaluate the anatomical
A. Nowacki et al. NeuroImage: Clinical 20 (2018) 868–874
873
accuracy of the atlases with respect to the border between the anterior
STN and internal capsule.
5. Conclusions
Normalization of individual MRI data into the MNI template space
and application of different three-dimensional subcortical atlases in-
evitably led to distortions of the reconstructed lead in relation to the
STN. Of the four different atlases tested in this work, the DISTAL atlas
leads to the most accurate results. However, results must be cautiously
interpreted as there seems to be a lateral shift of the reconstructed DBS
lead relative to the STN.
Funding
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.nicl.2018.09.030.
References
Accolla, E.A., Herrojo Ruiz, M., Horn, A., Schneider, G.H., Schmitz-Hubsch, T., Draganski,
B., Kuhn, A.A., 2016. Brain networks modulated by subthalamic nucleus deep brain
stimulation. Brain 139, 2503–2515.
Akram, H., Sotiropoulos, S.N., Jbabdi, S., Georgiev, D., Mahlknecht, P., Hyam, J.,
Foltynie, T., Limousin, P., De Vita, E., Jahanshahi, M., Hariz, M., Ashburner, J.,
Behrens, T., Zrinzo, L., 2017. Subthalamic deep brain stimulation sweet spots and
hyperdirect cortical connectivity in Parkinson's disease. Neuroimage 158, 332–345.
Avants, B.B., Epstein, C.L., Grossman, M., Gee, J.C., 2008. Symmetric diffeomorphic
image registration with cross-correlation: evaluating automated labeling of elderly
and neurodegenerative brain. Med Image Anal 12, 26–41.
Brett, M., Johnsrude, I.S., Owen, A.M., 2002. The problem of functional localization in the
human brain. Nat Rev. Neurosci 3, 243–249.
Chakravarty, M.M., Bertrand, G., Hodge, C.P., Sadikot, A.F., Collins, D.L., 2006. The
creation of a brain atlas for image guided neurosurgery using serial histological data.
Neuroimage 30, 359–376.
Chandran, A.S., Bynevelt, M., Lind, C.R., 2016. Magnetic resonance imaging of the sub-
thalamic nucleus for deep brain stimulation. J Neurosurg 124, 96–105.
Cheung, T., Noecker, A.M., Alterman, R.L., McIntyre, C.C., Tagliati, M., 2014. Defining a
therapeutic target for pallidal deep brain stimulation for dystonia. Ann Neurol 76,
22–30.
Dalaker, T.O., Zivadinov, R., Ramasamy, D.P., Beyer, M.K., Alves, G., Bronnick, K.S.,
Tysnes, O.B., Aarsland, D., Larsen, J.P., 2011. Ventricular enlargement and mild
cognitive impairment in early Parkinson's disease. Mov Disord 26, 297–301.
Deuschl, G., Schade-Brittinger, C., Krack, P., Volkmann, J., Schafer, H., Botzel, K.,
Daniels, C., Deutschlander, A., Dillmann, U., Eisner, W., Gruber, D., Hamel, W.,
Herzog, J., Hilker, R., Klebe, S., Kloss, M., Koy, J., Krause, M., Kupsch, A., Lorenz, D.,
Lorenzl, S., Mehdorn, H.M., Moringlane, J.R., Oertel, W., Pinsker, M.O., Reichmann,
H., Reuss, A., Schneider, G.H., Schnitzler, A., Steude, U., Sturm, V., Timmermann, L.,
Tronnier, V., Trottenberg, T., Wojtecki, L., Wolf, E., Poewe, W., Voges, J., German
Parkinson Study Group, N.S, 2006. A randomized trial of deep-brain stimulation for
Parkinson's disease. N Engl J Med 355, 896–908.
Ewert, S., Plettig, P., Li, N., Chakravarty, M.M., Collins, D.L., Herrington, T.M., Kuhn,
A.A., Horn, A., 2017. Toward defining deep brain stimulation targets in MNI space: A
subcortical atlas based on multimodal MRI, histology and structural connectivity.
Neuroimage. 170, 271–282.
Ewert, S., Plettig, P., Li, N., Chakravarty, M.M., Collins, D.L., Herrington, T.M., Kuhn,
A.A., Horn, A., 2018. Toward defining deep brain stimulation targets in MNI space: A
subcortical atlas based on multimodal MRI, histology and structural connectivity.
Neuroimage 170, 271–282.
Geng, D.Y., Li, Y.X., Zee, C.S., 2006. Magnetic resonance imaging-based volumetric
analysis of basal ganglia nuclei and substantia nigra in patients with Parkinson's
disease. Neurosurgery 58, 256–262.
Gunalan, K., Howell, B., McIntyre, C.C., 2018. Quantifying axonal responses in patient-
specific models of subthalamic deep brain stimulation. Neuroimage 172, 263–277.
Hamani, C., Richter, E.O., Andrade-Souza, Y., Hutchison, W., Saint-Cyr, J.A., Lozano,
A.M., 2005. Correspondence of microelectrode mapping with magnetic resonance
imaging for subthalamic nucleus procedures. Surg Neurol 63, 249–253 discussion
253.
Heo, Y.J., Kim, S.J., Kim, H.S., Choi, C.G., Jung, S.C., Lee, J.K., Lee, C.S., Chung, S.J., Cho,
S.H., Lee, G.R., 2015. Three-dimensional fluid-attenuated inversion recovery se-
quence for visualisation of subthalamic nucleus for deep brain stimulation in
Parkinson's disease. Neuroradiology 57, 929–935.
Horn, A., Kuhn, A.A., 2015. Lead-DBS: a toolbox for deep brain stimulation electrode
localizations and visualizations. Neuroimage 107, 127–135.
Horn, A., Kuhn, A.A., Merkl, A., Shih, L., Alterman, R., Fox, M., 2017. Probabilistic
conversion of neurosurgical DBS electrode coordinates into MNI space. Neuroimage
150, 395–404.
Husch, A., M, V.P., Gemmar, P., Goncalves, J., Hertel, F., 2018. PaCER - A fully automated
method for electrode trajectory and contact reconstruction in deep brain stimulation.
Neuroimage Clin 17, 80–89.
Klein, A., Andersson, J., Ardekani, B.A., Ashburner, J., Avants, B., Chiang, M.C.,
Christensen, G.E., Collins, D.L., Gee, J., Hellier, P., Song, J.H., Jenkinson, M., Lepage,
C., Rueckert, D., Thompson, P., Vercauteren, T., Woods, R.P., Mann, J.J., Parsey,
R.V., 2009. Evaluation of 14 nonlinear deformation algorithms applied to human
brain MRI registration. Neuroimage 46, 786–802.
Kocabicak, E., Aygun, D., Ozaydin, I., Jahanshahi, A., Tan, S., Onar, M., Boke, O., Kurt,
M., Guz, H., Terzi, M., Alptekin, O., Temel, Y., 2013. Does probe's eye subthalamic
nucleus length on T2W MRI correspond with microelectrode recording in patients
with deep brain stimulation for advanced Parkinson's disease? Turk Neurosurg 23,
658–665.
Nowacki, A., Fiechter, M., Fichtner, J., Debove, I., Lachenmayer, L., Schupbach, M.,
Oertel, M.F., Wiest, R., Pollo, C., 2015. Using MDEFT MRI Sequences to Target the
GPi in DBS Surgery. PLoS One 10, e0137868.
Nowacki, A., Debove, I., Fiechter, M., Rossi, F., Oertel, M.F., Wiest, R., Schupbach, M.,
Pollo, C., 2017. Targeting Accuracy of the Subthalamic Nucleus in Deep Brain
Stimulation Surgery: Comparison Between 3 T T2-Weighted Magnetic Resonance
Imaging and Microelectrode Recording Results. Oper Neurosurg (Hagerstown) 15 (1),
66–71.
Pauli, W.M., Nili, A.N., Tyszka, J.M., 2017. A High-Resolution Probabilistic In Vivo Atlas
of Human Subcortical Brain Nuclei. bioRxiv 211,201.
Pollo, C., Kaelin-Lang, A., Oertel, M.F., Stieglitz, L., Taub, E., Fuhr, P., Lozano, A.M.,
Raabe, A., Schupbach, M., 2014. Directional deep brain stimulation: an in-
traoperative double-blind pilot study. Brain 137, 2015–2026.
Schlaier, J.R., Habermeyer, C., Warnat, J., Lange, M., Janzen, A., Hochreiter, A.,
Proescholdt, M., Brawanski, A., Fellner, C., 2011. Discrepancies between the MRI-
and the electrophysiologically defined subthalamic nucleus. Acta Neurochir (Wien)
153, 2307–2318.
Schuepbach, W.M., Rau, J., Knudsen, K., Volkmann, J., Krack, P., Timmermann, L.,
Halbig, T.D., Hesekamp, H., Navarro, S.M., Meier, N., Falk, D., Mehdorn, M.,
Paschen, S., Maarouf, M., Barbe, M.T., Fink, G.R., Kupsch, A., Gruber, D., Schneider,
G.H., Seigneuret, E., Kistner, A., Chaynes, P., Ory-Magne, F., Brefel Courbon, C.,
Vesper, J., Schnitzler, A., Wojtecki, L., Houeto, J.L., Bataille, B., Maltete, D., Damier,
P., Raoul, S., Sixel-Doering, F., Hellwig, D., Gharabaghi, A., Kruger, R., Pinsker, M.O.,
Amtage, F., Regis, J.M., Witjas, T., Thobois, S., Mertens, P., Kloss, M., Hartmann, A.,
Oertel, W.H., Post, B., Speelman, H., Agid, Y., Schade-Brittinger, C., Deuschl, G.,
Group, E.S, 2013. Neurostimulation for Parkinson's disease with early motor com-
plications. N Engl J Med 368, 610–622.
Schupbach, W.M.M., Chabardes, S., Matthies, C., Pollo, C., Steigerwald, F., Timmermann,
L., Visser Vandewalle, V., Volkmann, J., Schuurman, P.R., 2017. Directional leads for
deep brain stimulation: Opportunities and challenges. Mov Disord 32, 1371–1375.
Senova, S., Hosomi, K., Gurruchaga, J.M., Gouello, G., Ouerchefani, N., Beaugendre, Y.,
Lepetit, H., Lefaucheur, J.P., Badin, R.A., Dauguet, J., Jan, C., Hantraye, P.,
Brugieres, P., Palfi, S., 2016. Three-dimensional SPACE fluid-attenuated inversion
recovery at 3 T to improve subthalamic nucleus lead placement for deep brain sti-
mulation in Parkinson's disease: from preclinical to clinical studies. J Neurosurg 125,
472–480.
Starr, P.A., Christine, C.W., Theodosopoulos, P.V., Lindsey, N., Byrd, D., Mosley, A.,
Marks Jr., W.J., 2002. Implantation of deep brain stimulators into the subthalamic
nucleus: technical approach and magnetic resonance imaging-verified lead locations.
J Neurosurg 97, 370–387.
Sumanaweera, T.S., Adler, J.R., Jr., Napel, S., Glover, G.H., 1994. Characterization of
spatial distortion in magnetic resonance imaging and its implications for stereotactic
surgery. Neurosurgery 35, 696–703; discussion 703–694.
Tinkhauser, G., Pogosyan, A., Debove, I., Nowacki, A., Shah, S.A., Seidel, K., Tan, H.,
Brittain, J.S., Petermann, K., di Biase, L., Oertel, M., Pollo, C., Brown, P., Schuepbach,
M., 2018. Directional local field potentials: A tool to optimize deep brain stimulation.
Mov Disord 33, 159–164.
Wang, B.T., Poirier, S., Guo, T., Parrent, A.G., Peters, T.M., Khan, A.R., 2016. Generation
and evaluation of an ultra-high-field atlas with applications in DBS planning. SPIE
Medical Imaging. SPIE, pp. 10.
Welter, M.L., Schupbach, M., Czernecki, V., Karachi, C., Fernandez-Vidal, S., Golmard,
J.L., Serra, G., Navarro, S., Welaratne, A., Hartmann, A., Mesnage, V., Pineau, F.,
Cornu, P., Pidoux, B., Worbe, Y., Zikos, P., Grabli, D., Galanaud, D., Bonnet, A.M.,
Belaid, H., Dormont, D., Vidailhet, M., Mallet, L., Houeto, J.L., Bardinet, E., Yelnik, J.,
Agid, Y., 2014. Optimal target localization for subthalamic stimulation in patients
with Parkinson disease. Neurology 82, 1352–1361.
Wodarg, F., Herzog, J., Reese, R., Falk, D., Pinsker, M.O., Steigerwald, F., Jansen, O.,
Deuschl, G., Mehdorn, H.M., Volkmann, J., 2012. Stimulation site within the MRI-
defined STN predicts postoperative motor outcome. Mov Disord 27, 874–879.
Xiao, Y., Fonov, V., Beriault, S., Al Subaie, F., Chakravarty, M.M., Sadikot, A.F., Pike,
G.B., Collins, D.L., 2015. Multi-contrast unbiased MRI atlas of a Parkinson's disease
population. Int J Comput Assist Radiol Surg 10, 329–341.
Yelnik, J., Bardinet, E., Dormont, D., Malandain, G., Ourselin, S., Tande, D., Karachi, C.,
Ayache, N., Cornu, P., Agid, Y., 2007. A three-dimensional, histological and de-
formable atlas of the human basal ganglia. I. Atlas construction based on im-
munohistochemical and MRI data. Neuroimage 34, 618–638.
Zaidel, A., Spivak, A., Grieb, B., Bergman, H., Israel, Z., 2010. Subthalamic span of beta
oscillations predicts deep brain stimulation efficacy for patients with Parkinson's
disease. Brain 133, 2007–2021.
A. Nowacki et al. NeuroImage: Clinical 20 (2018) 868–874
874
