Neuronavigation using susceptibility-weighted venography: application to deep brain stimulation and comparison with gadolinium contrast

Technical note

Full access

Careful trajectory planning on preoperative vascular imaging is an essential step in deep brain stimulation (DBS) to minimize risks of hemorrhagic complications and postoperative neurological deficits. This paper compares 2 MRI methods for visualizing cerebral vasculature and planning DBS probe trajectories: a single data set T1-weighted scan with double-dose gadolinium contrast (T1w-Gd) and a multi–data set protocol consisting of a T1-weighted structural, susceptibility-weighted venography, and time-of-flight angiography (T1w-SWI-TOF). Two neurosurgeons who specialize in neuromodulation surgery planned bilateral STN DBS in 18 patients with Parkinson's disease (36 hemispheres) using each protocol separately. Planned trajectories were then evaluated across all vascular data sets (T1w-Gd, SWI, and TOF) to detect possible intersection with blood vessels along the entire path via an objective vesselness measure. The authors' results show that trajectories planned on T1w-SWI-TOF successfully avoided the cerebral vasculature imaged by conventional T1w-Gd and did not suffer from missing vascular information or imprecise data set registration. Furthermore, with appropriate planning and visualization software, trajectory corridors planned on T1w-SWI-TOF intersected significantly less fine vasculature that was not detected on the T1w-Gd (p < 0.01 within 2 mm and p < 0.001 within 4 mm of the track centerline). The proposed T1w-SWI-TOF protocol comes with minimal effects on the imaging and surgical workflow, improves vessel avoidance, and provides a safe cost-effective alternative to injection of gadolinium contrast.

Abbreviations used in this paper:DBS = deep brain stimulation; IBIS = Interactive Brain Imaging System; MER = microelectrode recording; MIP = maximum intensity projection; mIP = minimum intensity projection; PD = Parkinson's disease; STN = subthalamic nucleus; SWI = susceptibility-weighted imaging; TOF = time of flight; T1w = T1-weighted.

Abstract

Careful trajectory planning on preoperative vascular imaging is an essential step in deep brain stimulation (DBS) to minimize risks of hemorrhagic complications and postoperative neurological deficits. This paper compares 2 MRI methods for visualizing cerebral vasculature and planning DBS probe trajectories: a single data set T1-weighted scan with double-dose gadolinium contrast (T1w-Gd) and a multi–data set protocol consisting of a T1-weighted structural, susceptibility-weighted venography, and time-of-flight angiography (T1w-SWI-TOF). Two neurosurgeons who specialize in neuromodulation surgery planned bilateral STN DBS in 18 patients with Parkinson's disease (36 hemispheres) using each protocol separately. Planned trajectories were then evaluated across all vascular data sets (T1w-Gd, SWI, and TOF) to detect possible intersection with blood vessels along the entire path via an objective vesselness measure. The authors' results show that trajectories planned on T1w-SWI-TOF successfully avoided the cerebral vasculature imaged by conventional T1w-Gd and did not suffer from missing vascular information or imprecise data set registration. Furthermore, with appropriate planning and visualization software, trajectory corridors planned on T1w-SWI-TOF intersected significantly less fine vasculature that was not detected on the T1w-Gd (p < 0.01 within 2 mm and p < 0.001 within 4 mm of the track centerline). The proposed T1w-SWI-TOF protocol comes with minimal effects on the imaging and surgical workflow, improves vessel avoidance, and provides a safe cost-effective alternative to injection of gadolinium contrast.

With a reported incidence rate as high as 5% in recent literature,2,8,37 hemorrhagic complications pose a significant risk of neurological deficits in deep brain stimulation (DBS) surgery. While surgical protocols that minimize or omit the use of microelectrode recordings (MERs) may reduce hemorrhage rates,19,37 many centers rely on physiology to determine the location and extent of the targeted nucleus and to maximize treatment efficacy.24,27,29 Hence, careful trajectory selection on preoperative vascular imaging remains an essential step to keep the central and any parallel (lateral) MER tracks at a safe distance from the cerebral vasculature. In this paper we compare 2 MRI approaches to visualizing the cerebral vasculature in the context of neuronavigation for DBS surgery: a T1-weighted MRI with gadolinium contrast (T1w-Gd) and a multi–data set protocol consisting of T1-weighted, susceptibility-weighted imaging (SWI)–based venography22,23 and time-of-flight (TOF) MR angiography (T1w-SWI-TOF).

Magnetic resonance imaging protocols involving the injection of gadolinium contrast (T1w-Gd) are commonly used in DBS planning and offer key advantages over alternative angiographic protocols. These T1w-Gd protocols simultaneously provide vascular and tissue contrast via a single acquisition, thereby reducing scan time and avoiding registration errors, and permit surgical planning using a single data set. However, the injection of gadolinium contrast has some associated medical risks (for example, uncommon nephropathy or systemic fibrosis, local skin reactions, and rare systemic allergy)30,31,33 and cost. Most importantly, gadolinium contrast only achieves visualization of large vessels as the contrast decreases rapidly in smaller vessels due to partial volume averaging with surrounding tissue.

Susceptibility-weighted imaging (SWI)22,23 is a relatively new MRI technique that is primarily sensitive to deoxygenated blood and to deep brain structures rich in iron content. Susceptibility-weighted imaging already provides useful information in a variety of clinical applications including traumatic brain injury, vascular malformations, strokes, and neurodegenerative disorders.26 For DBS specifically, SWI has been recently introduced as a direct method to target the subthalamic nucleus (STN),7,17,28,34,36 a major target for patients with Parkinson's disease (PD) that is otherwise indirectly approximated based on other anatomical landmarks10 or atlases.11,13,21 Finally, SWI can image cerebral veins with contrast superior to that of conventional gadolinium-based MRI, especially for smaller veins in subcortical and deep brain areas.

However, some methodological challenges have prevented the use of SWI-based venography in neurosurgical applications for preoperative planning and intraoperative navigation. Susceptibility-weighted imaging is only sensitive to venous blood and should be accompanied by another angiographic acquisition, such as TOF, to visualize the arteries and standard T1-weighted MRI to visualize anatomy, resulting in increased total scan time and requiring postacquisition registration. The use of multiple data sets (T1w-SWI-TOF) also poses some visualization challenges and may increase the overall planning complexity. Indeed, multiple data sets must be inspected concurrently, and manual identification of a vessel-free path using information-rich SWI data sets can be more challenging. However, emerging computer-assisted planning software4,6,18,32 that automatically evaluates different insertion options and generates graphical maps of most recommended trajectories may facilitate decision making, reduce planning complexity, and improve safety.

In this work we experimentally evaluate the application of combined SWI-based venography and TOF angiography for DBS neuronavigation by comparing it to the current gold-standard T1w-Gd methodology. Specifically, probe trajectories are statistically compared to assess whether all the necessary vessel information imaged on standard, T1w-Gd MRI is also present and accurately registered on T1w-SWI-TOF MRI. Furthermore, we assess whether the additional vascular information provided by the SWI-TOF acquisition is useful to improve the choice of lead trajectory. Finally, we review the main methodological challenges of the technique and investigate practical solutions to facilitate the integration of SWI-TOF within the overall surgical workflow of DBS.

Methods

Patient Population

This study was approved by the Research Ethics Committee of the Montreal Neurological Institute. Twenty-one patients with PD short-listed for STN DBS provided informed written consent and underwent 2 MRI sessions. The patients underwent T1w-Gd imaging, as part of their regular clinical treatment, and our proposed T1w-SWI-TOF protocol. Two patients were excluded due to patient motion, and 1 patient was excluded because clinical T1w-Gd data were not acquired. The remaining 18 patients (11 males and 7 females, age range 44–72 years) were used for this study and the data were analyzed as 36 brain hemispheres.

MRI Acquisition

Protocol 1 (T1w-Gd)

Whole-head T1w-Gd data sets were acquired using a clinical 1.5-T Signa EXCITE scanner (GE). Patients were injected with double-dose gadolinium contrast and were scanned with a 3D fast spoiled gradient echo sequence with transverse orientation and 0.59 × 0.59 × 1.5–mm resolution (TR 23 msec, TE 8 msec, α = 8°, acquisition time 10:00 minutes).

Protocol 2 (T1w-SWI-TOF)

The T1-weighted, SWI, and TOF acquisitions were performed within a single-scanning session on a 3-T TIM Trio scanner (Siemens) with a 32-channel head coil. Standard T1-weighted anatomical contrast of the whole head was obtained from a 3D, magnetization-prepared, rapid gradient-echo (MPRAGE) sequence with sagittal orientation and 1 × 1 × 1–mm resolution (TR 2300 msec, TE 2.98 msec, TI 900 msec, α = 9°, iPAT = 2, acquisition time 5:30 minutes). Venous blood contrast was obtained by SWI based on the 3D gradient echo sequence of Denk et al.15 with transverse orientation, 0.5 × 0.5 × 1–mm resolution, and 5 equally spaced echoes (TR 48 msec; TE 13, 20, 27, 34, and 41 msec; α = 17°, iPAT = 2, acquisition time 10:24 msec). The first echo is fully flow compensated, and the third and fifth echoes are flow compensated in the readout direction. Arterial contrast was obtained using a flow-compensated 3D multislab TOF with 1 × 1 × 1–mm resolution (TR 22 msec, TE 3.85 msec, α = 18°, 4 slabs, 44 slices/slab, acquisition time 9:08 msec).

Data Set Registration

Intrasubject T1-weighted, SWI, and TOF data sets, all acquired within the same scanning session, were rigidly coregistered by mutual information using the MINC toolkit (http://packages.bic.mni.mcgill.ca/). For the purpose of comparison, the T1w-Gd data sets (from Protocol 1) were also linearly registered to their corresponding T1-weighted data sets (from Protocol 2) by mutual information. The registration accuracy between the 2 protocols was carefully examined (see Results) to ensure a fair comparison.

Target Identification

Prior to the experiment, realistic DBS targets were identified within the left and right motor regions of the STN of each subject. The STN targets were directly identified by the neurosurgeons via visual inspection of the SWI data sets (coregistered to all other data sets).

Trajectory Planning Experiments

The case series (36 hemispheres) was divided between 2 neurosurgeons with DBS expertise. On all hemispheres, the neurosurgeons planned STN DBS trajectories that fulfill typical surgical requirements2 and minimize intersection with MR-visible vasculature. All hemispheres were planned twice: once using only the T1w-Gd protocol and once, by the same neurosurgeon, using only the T1w-SWI-TOF protocol. To avoid bias due to learning and order effects, the 2 planning sessions were separated by a minimum of 2 weeks, the subject names were blinded, and the case order and protocol order were shuffled using an automatic program.

All planning experiments were performed using the Interactive Brain Imaging System (IBIS),5,25 a prototype neuronavigation platform developed within our center. The planning experiments were conducted in 2 passes as follows.

First Pass

The neurosurgeons planned all 36 DBS hemispheres by manual inspection of the raw MRI data. Candidate trajectories were identified using a combination of standard 3D and 2D (slice) views available within IBIS. Essentially, the neurosurgeons would interactively select a candidate trajectory on a 3D cortex view and examine the trajectory on various 2D views, including a probe's eye view, sliced perpendicularly to the trajectory, to assess whether the vasculature and other eloquent brain structures are properly avoided along the entire path. This process was repeated until a surgically suitable trajectory was found.

Second Pass

The multimodal T1w-SWI-TOF data sets were passed to computer-assisted path-planning software described in a previous study.6 This software models the planning process as a set of constraints to optimize (for example, trajectories anterior to the primary motor cortex that avoid blood vessels, sulci, ventricles, and caudate) and automatically processes thousands of possible entry points to provide neurosurgeons with intuitive graphical maps of software-recommended insertion options (see Fig. 1 for an example). For each hemisphere, 2 alternate trajectories were automatically computed by the software and were presented to the neurosurgeons. The neurosurgeons could either choose to replace or keep their original manual plan (found during the first pass) with one proposed by the automatic software. This second pass investigates whether the use of computer-assisted software helps to improve the identification of vessel-free trajectories on denser and more complex SWI-TOF angiograms. A second pass was not needed for the T1w-Gd data sets because the manually planned trajectories already avoided the complete vasculature visible on gadolinium contrast.

Fig. 1.
Fig. 1.

Second pass of planning experiments. A: Three-dimensional cortex view. B–D: Two-dimensional probe's eye trajectory visualization of T1-weighted, SWI, and TOF (synchronized to selected trajectory). The blue cylinder is the trajectory planned without computer-assisted software feedback (first pass). The gray cylinders are 2 software-optimized trajectories computed during the second pass obtained from the color-coded automatic trajectory analysis (green patches represent software-recommended insertion areas).

Trajectory Comparison

After the experiment, trajectories planned using T1w-Gd and T1w-SWI-TOF data sets (referred as TGD and TSWI, respectively, in the text) were objectively compared, case-by-case, based on the extent they intersect or approach vessels segmented on MRI. Our hypothesis is that a trajectory corridor that intersects more vasculature visible on preoperative MRI is more likely to cause bleeding than a trajectory corridor that intersects less MR-visible vasculature. This coincides with one important objective of DBS planning stage: the neurosurgeon tries to select a trajectory that best avoids MR-visible vasculature.

The raw vascular data sets (T1w-Gd, SWI, and TOF) were postprocessed to segment the cerebral vasculature by multiscale vesselness enhancement. The data sets were resampled to 0.5-mm isotropic resolution, denoised with a nonlocal means filter12 (using a Rician noise model35), and processed with Frangi et al.'s 3D vesselness filter20 using standard parameters (σ= [0.5–2.5], Δσ= 0.25; α = 0.5, β = 0.5, γ = half the maximal Hessian norm. The vesselness filter is sensitive to tubular structures and therefore returns a voxel likelihood [0.0–1.0] of blood vessel presence.). This image-processing pipeline enhances the cerebral vasculature imaged by T1w-Gd and TOF, and nonsurface veins imaged by SWI (see Fig. 2A and B, upper and center). Hypointense SWI surface veins are not properly captured by standard vesselness filtering because the surrounding CSF and skull is also dark, thereby breaking the tubular assumption posed by the filter. Instead, SWI surface veins were enhanced separately using a generalized vesselness measure1 that relaxes the tubular assumption (see Fig. 2A and B, lower).

Fig. 2.
Fig. 2.

Illustrative examples of the image processing steps. A: Raw T1w-Gd, SWI, and TOF data sets displayed as 15-mm minimum/maximum intensity projections taken at the level of the lateral ventricles (upper), the circle of Willis (center), and brain surface (lower). B: Vesselness-filtered T1w-Gd and SWI-TOF data sets. C: Automatic vessel centerline extraction and matching. The blue curve indicates the centerline detected on SWI-TOF; the red curve, the centerline detected on T1w-Gd.

The TGD and TSWI were evaluated using the postprocessed vesselness data sets. The trajectories were modeled as cylindrical corridors on the MRI with a 2-mm cylinder radius (to include all voxels within 2-mm of the central MER track) and with a 4-mm cylinder radius (that is, to include all voxels within 2-mm of any lateral MER track on a “Ben's gun” array of 5 parallel microelectrodes3). The 2- and 4-mm analyses, respectively, detect intersection and proximity with MR-visible vasculature.

Paired comparisons between TGD and TSWI corridors were performed on vesselness-filtered T1w-Gd and SWI-TOF data sets. Vessel intersections on T1w-Gd were quantified according to the maximal vesselness value encountered within 2 and 4 mm of the trajectory. Similarly, vessel intersections on SWI-TOF data sets were quantified according to the sum of all vesselness values within 2 and 4 mm of the trajectory because the maximal value does not provide sufficient discrimination on denser data.6 Voxels with a vesselness value smaller than 0.05 were considered as vessel-free voxels and were excluded to minimize the effect of varying path lengths. A nonparametric Wilcoxon paired 2-tailed test was used for all statistical comparisons between TGD and TSWI with a significance threshold of 0.05.

Results

Data Set Registration

To ensure a fair comparison between TGD and TSWI, we first estimated, case-by-case, the registration precision across the 2 imaging protocols. To do so, we computed the mean centerline displacement among homologous vessels imaged on the T1w-Gd and on the SWI-TOF data sets. Vessel centerlines were automatically extracted from the vesselness-filtered data sets using the skeletonization algorithm of Bouix et al.9 (see Fig. 2C). Then, each centerline voxel on T1w-Gd was matched to the centerline voxel with the smallest Euclidian distance on the SWI-TOF. Since not all vessels are imaged by both protocols, voxels that were matched with a distance greater than 3.0 mm were discarded with the assumption that the registration error is much less than 3.0 mm. As per Table 1, submillimeter mean centerline displacement was observed for all cases, and the displacement did not increase at the brain surface. Furthermore, the centerline displacement matches the resolution precision of the centerline data (voxel size of 0.5 mm).

TABLE 1:

Mean vessel centerline displacement between coregistered SWI-TOF and T1w-Gd data sets for all patients

Case No.No. of MatchesMean Centerline Displacement (mm)
SubcorticalCortex
xyzxyz
156180.340.420.440.500.460.44
283510.360.390.360.480.460.43
366990.450.410.520.430.400.42
473260.460.490.470.490.420.45
560670.320.250.350.420.410.42
668820.360.330.350.450.430.40
773080.380.340.410.480.490.56
845250.430.480.500.490.480.50
973840.340.360.380.440.410.42
1057570.340.380.430.510.480.45
1135750.300.220.390.360.350.39
1274330.310.390.390.420.420.45
1360070.370.470.420.450.440.42
1474710.400.340.320.440.390.48
1589340.280.250.310.360.290.33
1674240.370.370.410.430.430.40
1765640.300.370.410.470.430.42
1810,3040.380.380.390.430.430.40

Trajectory Comparison Results

TGD and TSWI comparison results are illustrated in the box plots of Fig. 3A and B (first pass) and C and D (second pass). A higher vesselness score indicates more intersection or proximity (2- or 4-mm analysis) between a trajectory corridor and the segmented vasculature.

Fig. 3.
Fig. 3.

Comparison of blood vessel avoidance between TGD and TSWI. A: Manually planned TGD and TSWI evaluated on vesselness filtered T1w-Gd data sets. The arrows indicate 2 exceptional cases (vesselness > 0.1) that are further detailed in Fig. 4. B: Manually planned TGD and TSWI evaluated on vesselness-filtered SWI-TOF. C and D: Same comparison as panels A and B after the computer-assisted pass. A lower score (maximal [max]-vesselness or sum-of-vesselness) indicates less intersection or proximity with cerebral vasculature. *p < 0.01; **p < 0.001. The red lines indicate the median; the boxes, the first and third quartiles; the whiskers, range; and the red dots, outliers.

In Fig. 3A and C, TGD and TSWI corridors are compared against vasculature detected on T1w-Gd. Similar distributions for TGD and TSWI validates that SWI-TOF provides vascular contrast that is at least equivalent to T1w-Gd. In Fig. 3B and D, TGD and TSWI trajectory corridors are compared against denser vasculature detected on SWI-TOF. In Fig. 3D, significantly lower scores are observed for TSWI than TGD. Therefore, computer-assisted planning on T1w-SWI-TOF significantly reduces intersection and proximity to MR-visible vasculature, especially for finer subcortical and deep veins that are only partially or not visualized on T1w-Gd. The next 2 sections provide additional comparison details.

First Pass: Manual Planning

Manual identification of DBS trajectories on the T1w-Gd and T1w-SWI-TOF data sets was executed with an average planning time of 3.5 ± 2.1 minutes (maximum 11.0 minutes) and 4.9 ± 3.2 minutes (maximum 14.5 minutes) per hemisphere, respectively.

Evaluation on T1w-Gd Vesselness Data

As illustrated in Fig. 3A, the maximal vesselness value encountered within 2 and 4 mm from all TSWI and TGD is low and not statistically different (p > 0.05). This finding provides a good indication that the TSWI trajectories avoid the entire cerebral vasculature detectable on gadolinium contrast. However, there were 2 TSWI cases where a vesselness value > 0.10 was found within 4 mm from the path. These specific trajectories are illustrated in Fig. 4, at the level of the brain surface (where the maximal vesselness voxel was detected). In neither case was the choice of trajectory due to missing information on SWI-TOF. Indeed, the superficial veins imaged by T1w-Gd MRI (Fig. 4A and D) are also adequately imaged by SWI (Fig. 4B and E) with no apparent registration errors between the protocols.

Fig. 4.
Fig. 4.

Two special cases (A–C and D–F) of TSWI trajectories planned within 4 mm of a superficial vein visible on T1w-Gd (A and D). In both cases, the same vein is visible on SWI (B and E) with no apparent registration errors. Note that superficial veins are only partially imaged (C) or invisible (F) on TOF because this sequence is mainly sensitive to arterial blood.

Evaluation on SWI-TOF Vesselness Data

As illustrated in Fig. 3B, the score (sum of vesselness) computed for TSWI and TGD trajectories is similar and not statistically different (p > 0.05). This result indicates that manually planned TSWI are only equivalent to TGD, in terms of vessel avoidance, even though SWI-TOF produces superior vascular contrast to T1w-Gd. The next section investigates how a computer-assisted planning methodology may further improve TSWI selections.

Second Pass: Computer-Assisted Planning

The neurosurgeons compared their original TSWI plans with alternate, computer-predicted, TSWI plans generated by our automatic trajectory analysis software. In total, there were 18 cases (50%) where the use of the trajectory planning software guided the neurosurgeons toward selecting a different entry point located more than 5 mm away from the original TSWI plan. There were 13 cases (36%) where the plan selected during the first and second passes coincided within 5 mm and 5 cases (14%) where the neurosurgeons decided to keep the original TSWI.

Evaluation on T1w-Gd Vesselness Data

As illustrated in Fig. 3C, the TSWI trajectories identified during the computer-assisted pass effectively avoids the cerebral vasculature imaged by gadolinium contrast. Furthermore, we note the absence of TSWI plans with a maximal vesselness > 0.1.

Evaluation on SWI-TOF Vesselness Data

As illustrated in Fig. 3D, the sum-of-vesselness score is lower for TSWI than TGD, and the difference is statistically significant within 2 mm (p < 0.01) and 4 mm (p < 0.001) from the paths. Furthermore, the median scores of TSWI versus TGD are reduced by 53% (2-mm radius analysis) and 48% (4-mm radius analysis). A lower score indicates less intersection between TSWI trajectory corridors (2-mm and 4-mm) and fine cerebral vasculature imaged by SWI-TOF. Therefore, computer-assisted surgical planning on T1w-SWITOF improves identification of hypovascular trajectory corridors (see Fig. 5 for some illustrative examples).

Fig. 5.
Fig. 5.

Illustrative comparisons between TGD and TSWI plans for 5 cases (arrows above the columns indicate each case). A–E: Probe's eye visualization of 5 TGD (no apparent intersection with the vasculature). F–J: The same TGD trajectories visualized on the coregistered SWI data set (in proximity or intersection to SWI-visible vasculature). K–O: Corresponding TSWI trajectories that improve avoidance of small vessels seen on SWI subcortically (Columns 1–3 and 5) and at the brain surface (Column 4).

Discussion

Based on our experimental results, there is sufficient evidence suggesting that the SWI and TOF angiographic data sets can be considered as an alternative to gadolinium contrast injection for the identification of vessel-free lead trajectories in the context of DBS neuronavigation using MRI. However, as emphasized in the introduction, the use of a multi–data set SWI-based protocol poses important methodological challenges that must be carefully examined prior to routine clinical use. This section reviews the T1w-Gd and T1w-SWI-TOF protocol comparison with respect to 1) vessel imaging and registration, 2) multi–data set visualization, 3) surgical planning complexity, 4) scan time, and 5) integration with surgical workflow. For clarity,Table 2 presents a concise criterion-by-criterion summary of the protocol comparison.

TABLE 2:

Multicriteria comparison of T1w-Gd and T1w-SWI-TOF protocols*

Criteria of ComparisonProtocol
T1w-GdT1w-SWI-TOF
MRI acquisition− injection of contrast++ no contrast needed
▪ can be acquired at 1.5 & 3 T− 10–20 mins of additional scan time
− best used at higher field strength (≥3T)
vessel imaging▪ combined imaging of anatomy & vessels▪ separate imaging of anatomy, veins, & arteries
▪ sensitive to arteries, surface veins, & major veins of the deep venous systems++ sensitive to arteries, surface veins, & deep veins including small subcortical, septal, & subependymal veins
− less sensitive to small subcortical and deep veins▪ presence of other hypointense structures: iron deposition, pallidum, STN, red nucleus, substantia nigra, skull, background air
–susceptibility artifacts at air-tissue interfaces affecting part of the temporal lobes and the orbitofrontal cortex
intrasubject registration mIP/MIP+ no registration required▪ T1-weighted, SWI, & TOF data set must be rigidly registered
▪ arteries & veins (MIP of T1w-Gd)▪ arteries (MIP of TOF)
+ deep venous & subcortical veins (mIP of SWI)
− surface veins must be visualized at native resolution
planning complexity+ inspection of single data set− inspection of multiple data sets
− denser venograms (SWI)
▪ use of computer-assisted software recommended
stereotactic frame+ can be acquired w/ or w/o a frame− must be acquired w/o a frame

++ = major advantage; + = minor advantage; · = remark; − = minor disadvantage.

Vessel Imaging and Registration

The acquisition of SWI and TOF data sets provides excellent vascular contrast and registration accuracy to the anatomical data, allowing a viable alternative to the injection of gadolinium contrast. The TOF acquisition produces comparable arterial blood contrast to T1w-Gd, but limited venous contrast. The SWI complements the TOF and produces excellent venous contrast (with reversed contrast). In addition, SWI can image several much smaller subcortical, subependymal, and septal veins that are only partially visible or completely absent on T1w-Gd. Experimentally, none of the TSWI trajectories were found to intersect a blood vessel imaged by the standard T1w-Gd data set due to missing vascular contrast on SWI-TOF or imprecise data set registration, neither at the cortical surface nor subcortically. Furthermore, TSWI corridors, planned by neurosurgeons with the help of computer-assisted planning software, were found to intersect significantly less of the fine vasculature imaged by SWI-TOF.

Susceptibility-weighted imaging is also sensitive to nonvessel iron-rich brain structures such as the pallidum, STN, red nucleus, and substantia nigra. These hypointense structures did not interfere with the planning of DBS lead trajectories. As stated earlier, a secondary consequence of SWI in DBS is the possibility of targeting the STN directly rather than indirectly from the relative position of other anatomical landmarks. Other artifacts related to iron deposits can arise in SWI acquisitions, but blood vessels are robustly distinguished from these artifacts due to their extended tubular shape and vascular network connectedness.

One disadvantage of SWI is the sensitivity to susceptibility artifacts at the air-tissue interface, due to the use of relatively long echo times, hence limiting the exploration of brain areas near the skull base, next to air-filled sinuses and ear canals. This limitation has no effect on DBS planning of trajectories that enter the brain through the middle frontal gyrus and terminate within basal ganglia or thalamic nuclei. Susceptibility-weighted imaging–based neuronavigation could also extend to several other image-based interventions. However, some application-specific assessment should be performed, especially for applications necessitating navigation targeted at the inferior temporal lobe or inferior frontal cortices.

Visualization

Volumetric angiographic data are often rendered by maximum intensity projections (MIPs) or minimum intensity projections (mIPs). This visualization technique consists of computing a 2D projection, where each pixel takes either the maximal or the minimal voxel intensity encountered along the projection direction, often chosen along the direction of the trajectory (probe's eye view), with the assumption that the vessels have the highest or lowest intensities relatively to other tissues. Thus, the hyperintense vessels imaged by T1w-Gd and TOF can be efficiently visualized by MIP. Similarly, most hypointense veins imaged by SWI can be visualized by mIP, except at the brain surface because the skull and background air are also hypointense. Instead, surface veins must be rendered as multiple thin slices, which may slow down the planning process. Using the IBIS software,5,25 SWI data sets are rendered with an adaptive slice thickness: deep and subcortical venous structures are shown as mIP slabs with a thickness of up to 10 mm for fast trajectory analysis, and surface veins are shown with a 1-mm slice thickness. The software automatically determines the appropriate slice thickness to use based on depth along the trajectory.

Surgical Planning Complexity

Deep brain stimulation planning using either protocol yields comparable planning times even though the T1w-SWI-TOF protocol involves the inspection of multiple data sets. Since most vessel content imaged by SWI and TOF can be efficiently visualized as thick intensity projections, the use of a multi–data set protocol only represents an approximate 30% increase in the number of slices that must be inspected per trajectory. For example, if we consider a typical DBS path of length 75–80 mm, about 80 Probe's eye view slices with 1-mm thickness would be inspected on the T1-weighted or T1w-Gd data set to determine whether the trajectory intersects a sulcus, the lateral ventricles, and more eloquent brain areas. On T1w-SWI-TOF about 105 slices per trajectory would need to be inspected: 80 slices on the T1-weighted data set, 8 MIP slabs with 10-mm thickness for the TOF, 7 mIP slabs with 10-mm thickness for deep SWI veins, and 10 slices with 1-mm thickness for the surface veins.

The main advantage of SWI-TOF is the enhanced vascular contrast, especially in subcortical and deep venous system areas. This can be observed qualitatively in Fig. 2A and B (upper and center) and analytically in the plots of Fig. 6 that estimate the vessel density (averaged across all subjects) for different vesselness thresholds. Within the vesselness threshold range of 0.05–0.10, which increases sensitivity to finer vasculature, the SWI-TOF is denser than T1w-Gd by factors ranging from 1.5 to 9.

Fig. 6.
Fig. 6.

Vessel density (vessel voxels/[vessel + tissue voxels] × 100%) measured on SWI-TOF and on T1w-Gd across all subjects for the deep venous system (upper) and the subcortical vessels (lower). A higher vesselness threshold (> 0.10) is sensitive to large vasculature only (vessel density is similar for SWI-TOF and T1w-Gd). A vesselness threshold in the range of 0.05–0.10 increases sensitivity to finer vessels and shows vessel density increases by factors of 1.5 to 9 for SWI-TOF versus T1w-Gd. The lines represent the means and the whiskers represent standard deviation.

The additional contrast produced by SWI-TOF enables the identification of DBS trajectories that avoid more vessels. In our study, this advantage was realized via a computer-assisted planning protocol. No significant gain in vessel avoidance was observed by manual planning. Furthermore, in 2 cases, manually attempting to avoid the finer vasculature subcortically resulted in the selection of trajectories closer to a superficial vessel (within 4 mm). Hence, optimal planning on T1w-SWI-TOF is more complex. The use of advanced planning tools effectively helps mitigate the increased planning complexity as they can assist neurosurgeons through automatic computation and rapid exploration of software-optimized trajectories.

MRI Scan Time

The T1w-SWI-TOF protocol used in this experiment eliminates the time needed to inject the gadolinium contrast, but adds 20 minutes of scan time. Furthermore, SWI contrast is best acquired at a higher field strength (3 T or more).22 Indeed, whole-brain SWI acquisition at 1.5 T would require a longer echo time and repetition time to obtain T2*-weighted vascular contrast, thus limiting either the resolution or signal-to-noise ratio to maintain a clinically acceptable scan time.22 On the other hand, T1w-Gd can be acquired on widely available 1.5-T clinical scanners since the main factor responsible for the vascular contrast remains the gadolinium dosage.

To reduce the total scan time, the TOF acquisition could be embedded within a similar multiecho SWI acquisition, rather then a separate TOF acquisition. As shown in Fig. 7, it is possible to generate arterial blood contrast with the inclusion of an additional echo (TE < 5 msec), while approaching the quality of conventional TOF angiography.14,16 Without a separate TOF acquisition, the total scan time overhead is reduced by half.

Fig. 7.
Fig. 7.

Combined visualization of cerebral veins and arteries using a modified 6-echo SWI sequence. A: The fully flow-compensated magnitude data set of the first echo (TE 4.57 msec) provides bright arterial blood contrast. B: Phase and magnitude from the other 5 echoes are reconstructed by SWI and averaged to produce a standard SWI venogram. C: Comparison with a standard 3D TOF angiogram.

Integration With Surgical Workflow

Surgical workflows for DBS may or may not include the use of a stereotactic frame. However, within the context of a frame-based DBS workflow, the SWI data sets may need to be acquired before mounting the frame on the patient's head due to the use of a 32-channel head coil (for parallel imaging and high signal-to-noise ratio). Consequently, 2 scanning sessions are needed: one session to acquire high-quality T1w-SWI-TOF data sets and another session, in the early stages of the intervention, to acquire a separate T1-weighted MRI with the frame landmarks. In our institution, T1w-Gd is already acquired as a diagnostic MRI scan prior to surgery. In this case, the MRI protocol substitution of T1w-Gd for T1w-SWI-TOF comes with minimal impact on the surgical workflow.

Limitations

This study compared 2 MRI protocols for imaging cerebral vasculature. Specifically, DBS trajectories planned by neurosurgeons using T1w-Gd and using T1w-SWI-TOF were compared for vessel avoidance. One limitation of this study is that we cannot report, at this stage, a correlation between prospective use of T1w-SWI-TOF and postoperative reduction of hemorrhages. This must be accomplished separately as a large-scale clinical trial. However, the current study successfully evaluated the completeness of SWI-TOF for cerebrovascular imaging in DBS. Our results showed that presurgical navigation using T1w-SWI-TOF may significantly improve an important goal of DBS planning: to identify trajectories that best avoid the cerebral vasculature visible on preoperative MRI.

Conclusions

The proposed T1w-SWI-TOF protocol is appropriate for presurgical DBS planning, constitutes an effective replacement to the injection of gadolinium contrast, and improves visualization of finer vasculature not detected by T1w-Gd. Indeed, trajectories planned with T1w-SWITOF did not intersect blood vessel imaged by conventional T1w-Gd. With appropriate software support, significantly more vessels (p < 0.01 within 2 mm and p < 0.001 within 4 mm from the paths) can be avoided using the proposed protocol. Finally, the main methodological challenges posed by the inclusion of a dense multi–data set T1w-SWI-TOF protocol were extensively considered. Overall, the proposed protocol provides a safe cost-effective alternative to gadolinium contrast, comes with minimal effect on the surgical workflow, and, most importantly, significantly improves vessel avoidance.

Disclosure

The following support was received: Natural Sciences and Engineering Research Council of Canada (NSERC), Graduate Scholarship to Silvain Bériault; Fond de Recherche sur la Nature et les Technologies (FQRNT), Graduate Scholarship to Silvain Bériault; and NSERC Discovery Grant 170426 to G. Bruce Pike (operating funds). The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

Author contributions to the study and manuscript preparation include the following. Conception and design: Bériault, Sadikot, Drouin, Collins, Pike. Acquisition of data: Bériault, Sadikot, Alsubaie. Analysis and interpretation of data: Bériault, Collins, Pike, Sadikot. Drafting the article: Bériault, Pike, Sadikot. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Bériault. Bériault Statistical analysis: Bériault. Study supervision: Collins, Pike, Sadikot.

This article contains some figures that are displayed in color online but in black-and-white in the print edition.

References

  • 1

    Antiga L: Generalizing vesselness with respect to dimensionality and shape. Insight J 2007. (http://hdl.handle.net/1926/576) [Accessed March 28 2014]

  • 2

    Benabid ALChabardes SMitrofanis JPollak P: Deep brain stimulation of the subthalamic nucleus for the treatment of Parkinson's disease. Lancet Neurol 8:67812009

  • 3

    Benabid ALPollak PLouveau AHenry Sde Rougemont J: Combined (thalamotomy and stimulation) stereotactic surgery of the VIM thalamic nucleus for bilateral Parkinson disease. Appl Neurophysiol 50:3443461987

  • 4

    Bériault SAl Subaie FMok KSadikot AFPike GBAutomatic trajectory planning of DBS neurosurgery from multimodal MRI datasets. Fichtinger GMartel ALPeters TM: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2011. 14th International Conference Toronto Canada September 18–22 2011 Proceedings Part I HeidelbergSpringer2011. 259266

  • 5

    Bériault SDrouin SSadikot AFXiao YCollins DLPike GBA prospective evaluation of computer-assisted deep brain stimulation trajectory planning. Drechsler KErdt MLinguraru MG: Clinical Image-Based Procedures: From Planning to Intervention. International Workshop CLIP 2012 HeidelbergSpringer2013. 4249

  • 6

    Bériault SSubaie FACollins DLSadikot AFPike GB: A multi-modal approach to computer-assisted deep brain stimulation trajectory planning. Int J CARS 7:6877042012

  • 7

    Bériault SXiao YBailey LCollins DLSadikot AFPike GBTowards computer-assisted deep brain stimulation targeting with multiple active contacts. Ayache NDelingette HGolland P: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2012 Part I HeidelbergSpringer2012. 487494

  • 8

    Binder DKRau GStarr PA: Hemorrhagic complications of microelectrode-guided deep brain stimulation. Stereotact Funct Neurosurg 80:28312003

  • 9

    Bouix SSiddiqi KTannenbaum A: Flux driven automatic centerline extraction. Med Image Anal 9:2092212005

  • 10

    Brunenberg EJPlatel BHofman PATer Haar Romeny BMVisser-Vandewalle V: Magnetic resonance imaging techniques for visualization of the subthalamic nucleus. A review. J Neurosurg 115:9719842011

  • 11

    Chakravarty MMBertrand GHodge CPSadikot AFCollins DL: The creation of a brain atlas for image guided neurosurgery using serial histological data. Neuroimage 30:3593762006

  • 12

    Coupe PYger PPrima SHellier PKervrann CBarillot C: An optimized blockwise nonlocal means denoising filter for 3-D magnetic resonance images. IEEE Trans Med Imaging 27:4254412008

  • 13

    D'Haese PFCetinkaya EKonrad PEKao CDawant BM: Computer-aided placement of deep brain stimulators: from planning to intraoperative guidance. IEEE Trans Med Imaging 24:146914782005

  • 14

    Deistung ADittrich ESedlacik JRauscher AReichenbach JR: ToF-SWI: simultaneous time of flight and fully flow compensated susceptibility weighted imaging. J Magn Reson Imaging 29:147814842009

  • 15

    Denk CRauscher A: Susceptibility weighted imaging with multiple echoes. J Magn Reson Imaging 31:1851912010

  • 16

    Du YPJin Z: Simultaneous acquisition of MR angiography and venography (MRAV). Magn Reson Med 59:9549582008

  • 17

    Elolf EBockermann VGringel TKnauth MDechent PHelms G: Improved visibility of the subthalamic nucleus on high-resolution stereotactic MR imaging by added susceptibility (T2*) contrast using multiple gradient echoes. AJNR Am J Neuroradiol 28:109310942007

  • 18

    Essert CHaegelen CLalys FAbadie AJannin P: Automatic computation of electrode trajectories for Deep Brain Stimulation: a hybrid symbolic and numerical approach. Int J Comput Assist Radiol Surg 7:5175322012

  • 19

    Foltynie TZrinzo LMartinez-Torres ITripoliti EPetersen EHoll E: MRI-guided STN DBS in Parkinson's disease without microelectrode recording: efficacy and safety. J Neurol Neurosurg Psychiatry 82:3583632011

  • 20

    Frangi AFNiessen WJVincken KLViergever MAMultiscale vessel enhancement filtering. Wells WMColchester ACFDelp SL: Medical Image Computing and Computer-Assisted Intervention–MICCAI 1998 HeidelbergSpringer1998. 130137

  • 21

    Guo TParrent AGPeters TM: Surgical targeting accuracy analysis of six methods for subthalamic nucleus deep brain stimulation. Comput Aided Surg 12:3253342007

  • 22

    Haacke EMMittal SWu ZNeelavalli JCheng YC: Susceptibility-weighted imaging: technical aspects and clinical applications, part 1. AJNR Am J Neuroradiol 30:19302009

  • 23

    Haacke EMXu YCheng YCReichenbach JR: Susceptibility weighted imaging (SWI). Magn Reson Med 52:6126182004

  • 24

    Machado ARezai ARKopell BHGross RESharan ADBenabid AL: Deep brain stimulation for Parkinson's disease: surgical technique and perioperative management. Mov Disord 21:Suppl 14S247S2582006

  • 25

    Mercier LDel Maestro RFPetrecca KKochanowska ADrouin SYan CX: New prototype neuronavigation system based on preoperative imaging and intraoperative freehand ultrasound: system description and validation. Int J Comput Assist Radiol Surg 6:5075222011

  • 26

    Mittal SWu ZNeelavalli JHaacke EM: Susceptibility-weighted imaging: technical aspects and clinical applications, part 2. AJNR Am J Neuroradiol 30:2322522009

  • 27

    Montgomery EB Jr: Microelectrode targeting of the subthalamic nucleus for deep brain stimulation surgery. Mov Disord 27:138713912012

  • 28

    O'Gorman RLShmueli KAshkan KSamuel MLythgoe DJShahidiani A: Optimal MRI methods for direct stereotactic targeting of the subthalamic nucleus and globus pallidus. Eur Radiol 21:1301362011

  • 29

    Rezai ARKopell BHGross REVitek JLSharan ADLimousin P: Deep brain stimulation for Parkinson's disease: surgical issues. Mov Disord 21:Suppl 14S197S2182006

  • 30

    Sadowski EABennett LKChan MRWentland ALGarrett ALGarrett RW: Nephrogenic systemic fibrosis: risk factors and incidence estimation. Radiology 243:1481572007

  • 31

    Sam AD IIMorasch MDCollins JSong GChen RPereles FS: Safety of gadolinium contrast angiography in patients with chronic renal insufficiency. J Vasc Surg 38:3133182003

  • 32

    Shamir RRJoskowicz LTamir IDabool EPertman LBen-Ami A: Reduced risk trajectory planning in image-guided keyhole neurosurgery. Med Phys 39:288528952012

  • 33

    Thomsen HS: Recent hot topics in contrast media. Eur Radiol 21:4924952011

  • 34

    Vertinsky ATCoenen VALang DJKolind SHoney CRLi D: Localization of the subthalamic nucleus: optimization with susceptibility-weighted phase MR imaging. AJNR Am J Neuroradiol 30:171717242009

  • 35

    Wiest-Daesslé NPrima SCoupé PMorrissey SBarillot CRician Noise removal by non-local means filtering for low signal-to-noise ratio MRI: applications to DT-MRI. Metaxas DAxel LFichtinger G: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2008 Part II HeidelbergSpringer2008. 171179

  • 36

    Xiao YBeriault SPike GBCollins DL: Multicontrast multiecho FLASH MRI for targeting the subthalamic nucleus. Magn Reson Imaging 30:6276402012

  • 37

    Zrinzo LFoltynie TLimousin PHariz MI: Reducing hemorrhagic complications in functional neurosurgery: a large case series and systematic literature review. Clinical article. J Neurosurg 116:84942012

If the inline PDF is not rendering correctly, you can download the PDF file here.

Article Information

Address correspondence to: Silvain Bériault, M.Sc., McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, 3801 University St., Rm. WB-325, Montreal, QC H3A 2B4, Canada. email:silvain.beriault@mail.mcgill.ca.

Please include this information when citing this paper: published online May 16, 2014; DOI: 10.3171/2014.3.JNS131860.

© AANS, except where prohibited by US copyright law.

Headings

Figures

  • View in gallery

    Second pass of planning experiments. A: Three-dimensional cortex view. B–D: Two-dimensional probe's eye trajectory visualization of T1-weighted, SWI, and TOF (synchronized to selected trajectory). The blue cylinder is the trajectory planned without computer-assisted software feedback (first pass). The gray cylinders are 2 software-optimized trajectories computed during the second pass obtained from the color-coded automatic trajectory analysis (green patches represent software-recommended insertion areas).

  • View in gallery

    Illustrative examples of the image processing steps. A: Raw T1w-Gd, SWI, and TOF data sets displayed as 15-mm minimum/maximum intensity projections taken at the level of the lateral ventricles (upper), the circle of Willis (center), and brain surface (lower). B: Vesselness-filtered T1w-Gd and SWI-TOF data sets. C: Automatic vessel centerline extraction and matching. The blue curve indicates the centerline detected on SWI-TOF; the red curve, the centerline detected on T1w-Gd.

  • View in gallery

    Comparison of blood vessel avoidance between TGD and TSWI. A: Manually planned TGD and TSWI evaluated on vesselness filtered T1w-Gd data sets. The arrows indicate 2 exceptional cases (vesselness > 0.1) that are further detailed in Fig. 4. B: Manually planned TGD and TSWI evaluated on vesselness-filtered SWI-TOF. C and D: Same comparison as panels A and B after the computer-assisted pass. A lower score (maximal [max]-vesselness or sum-of-vesselness) indicates less intersection or proximity with cerebral vasculature. *p < 0.01; **p < 0.001. The red lines indicate the median; the boxes, the first and third quartiles; the whiskers, range; and the red dots, outliers.

  • View in gallery

    Two special cases (A–C and D–F) of TSWI trajectories planned within 4 mm of a superficial vein visible on T1w-Gd (A and D). In both cases, the same vein is visible on SWI (B and E) with no apparent registration errors. Note that superficial veins are only partially imaged (C) or invisible (F) on TOF because this sequence is mainly sensitive to arterial blood.

  • View in gallery

    Illustrative comparisons between TGD and TSWI plans for 5 cases (arrows above the columns indicate each case). A–E: Probe's eye visualization of 5 TGD (no apparent intersection with the vasculature). F–J: The same TGD trajectories visualized on the coregistered SWI data set (in proximity or intersection to SWI-visible vasculature). K–O: Corresponding TSWI trajectories that improve avoidance of small vessels seen on SWI subcortically (Columns 1–3 and 5) and at the brain surface (Column 4).

  • View in gallery

    Vessel density (vessel voxels/[vessel + tissue voxels] × 100%) measured on SWI-TOF and on T1w-Gd across all subjects for the deep venous system (upper) and the subcortical vessels (lower). A higher vesselness threshold (> 0.10) is sensitive to large vasculature only (vessel density is similar for SWI-TOF and T1w-Gd). A vesselness threshold in the range of 0.05–0.10 increases sensitivity to finer vessels and shows vessel density increases by factors of 1.5 to 9 for SWI-TOF versus T1w-Gd. The lines represent the means and the whiskers represent standard deviation.

  • View in gallery

    Combined visualization of cerebral veins and arteries using a modified 6-echo SWI sequence. A: The fully flow-compensated magnitude data set of the first echo (TE 4.57 msec) provides bright arterial blood contrast. B: Phase and magnitude from the other 5 echoes are reconstructed by SWI and averaged to produce a standard SWI venogram. C: Comparison with a standard 3D TOF angiogram.

References

1

Antiga L: Generalizing vesselness with respect to dimensionality and shape. Insight J 2007. (http://hdl.handle.net/1926/576) [Accessed March 28 2014]

2

Benabid ALChabardes SMitrofanis JPollak P: Deep brain stimulation of the subthalamic nucleus for the treatment of Parkinson's disease. Lancet Neurol 8:67812009

3

Benabid ALPollak PLouveau AHenry Sde Rougemont J: Combined (thalamotomy and stimulation) stereotactic surgery of the VIM thalamic nucleus for bilateral Parkinson disease. Appl Neurophysiol 50:3443461987

4

Bériault SAl Subaie FMok KSadikot AFPike GBAutomatic trajectory planning of DBS neurosurgery from multimodal MRI datasets. Fichtinger GMartel ALPeters TM: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2011. 14th International Conference Toronto Canada September 18–22 2011 Proceedings Part I HeidelbergSpringer2011. 259266

5

Bériault SDrouin SSadikot AFXiao YCollins DLPike GBA prospective evaluation of computer-assisted deep brain stimulation trajectory planning. Drechsler KErdt MLinguraru MG: Clinical Image-Based Procedures: From Planning to Intervention. International Workshop CLIP 2012 HeidelbergSpringer2013. 4249

6

Bériault SSubaie FACollins DLSadikot AFPike GB: A multi-modal approach to computer-assisted deep brain stimulation trajectory planning. Int J CARS 7:6877042012

7

Bériault SXiao YBailey LCollins DLSadikot AFPike GBTowards computer-assisted deep brain stimulation targeting with multiple active contacts. Ayache NDelingette HGolland P: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2012 Part I HeidelbergSpringer2012. 487494

8

Binder DKRau GStarr PA: Hemorrhagic complications of microelectrode-guided deep brain stimulation. Stereotact Funct Neurosurg 80:28312003

9

Bouix SSiddiqi KTannenbaum A: Flux driven automatic centerline extraction. Med Image Anal 9:2092212005

10

Brunenberg EJPlatel BHofman PATer Haar Romeny BMVisser-Vandewalle V: Magnetic resonance imaging techniques for visualization of the subthalamic nucleus. A review. J Neurosurg 115:9719842011

11

Chakravarty MMBertrand GHodge CPSadikot AFCollins DL: The creation of a brain atlas for image guided neurosurgery using serial histological data. Neuroimage 30:3593762006

12

Coupe PYger PPrima SHellier PKervrann CBarillot C: An optimized blockwise nonlocal means denoising filter for 3-D magnetic resonance images. IEEE Trans Med Imaging 27:4254412008

13

D'Haese PFCetinkaya EKonrad PEKao CDawant BM: Computer-aided placement of deep brain stimulators: from planning to intraoperative guidance. IEEE Trans Med Imaging 24:146914782005

14

Deistung ADittrich ESedlacik JRauscher AReichenbach JR: ToF-SWI: simultaneous time of flight and fully flow compensated susceptibility weighted imaging. J Magn Reson Imaging 29:147814842009

15

Denk CRauscher A: Susceptibility weighted imaging with multiple echoes. J Magn Reson Imaging 31:1851912010

16

Du YPJin Z: Simultaneous acquisition of MR angiography and venography (MRAV). Magn Reson Med 59:9549582008

17

Elolf EBockermann VGringel TKnauth MDechent PHelms G: Improved visibility of the subthalamic nucleus on high-resolution stereotactic MR imaging by added susceptibility (T2*) contrast using multiple gradient echoes. AJNR Am J Neuroradiol 28:109310942007

18

Essert CHaegelen CLalys FAbadie AJannin P: Automatic computation of electrode trajectories for Deep Brain Stimulation: a hybrid symbolic and numerical approach. Int J Comput Assist Radiol Surg 7:5175322012

19

Foltynie TZrinzo LMartinez-Torres ITripoliti EPetersen EHoll E: MRI-guided STN DBS in Parkinson's disease without microelectrode recording: efficacy and safety. J Neurol Neurosurg Psychiatry 82:3583632011

20

Frangi AFNiessen WJVincken KLViergever MAMultiscale vessel enhancement filtering. Wells WMColchester ACFDelp SL: Medical Image Computing and Computer-Assisted Intervention–MICCAI 1998 HeidelbergSpringer1998. 130137

21

Guo TParrent AGPeters TM: Surgical targeting accuracy analysis of six methods for subthalamic nucleus deep brain stimulation. Comput Aided Surg 12:3253342007

22

Haacke EMMittal SWu ZNeelavalli JCheng YC: Susceptibility-weighted imaging: technical aspects and clinical applications, part 1. AJNR Am J Neuroradiol 30:19302009

23

Haacke EMXu YCheng YCReichenbach JR: Susceptibility weighted imaging (SWI). Magn Reson Med 52:6126182004

24

Machado ARezai ARKopell BHGross RESharan ADBenabid AL: Deep brain stimulation for Parkinson's disease: surgical technique and perioperative management. Mov Disord 21:Suppl 14S247S2582006

25

Mercier LDel Maestro RFPetrecca KKochanowska ADrouin SYan CX: New prototype neuronavigation system based on preoperative imaging and intraoperative freehand ultrasound: system description and validation. Int J Comput Assist Radiol Surg 6:5075222011

26

Mittal SWu ZNeelavalli JHaacke EM: Susceptibility-weighted imaging: technical aspects and clinical applications, part 2. AJNR Am J Neuroradiol 30:2322522009

27

Montgomery EB Jr: Microelectrode targeting of the subthalamic nucleus for deep brain stimulation surgery. Mov Disord 27:138713912012

28

O'Gorman RLShmueli KAshkan KSamuel MLythgoe DJShahidiani A: Optimal MRI methods for direct stereotactic targeting of the subthalamic nucleus and globus pallidus. Eur Radiol 21:1301362011

29

Rezai ARKopell BHGross REVitek JLSharan ADLimousin P: Deep brain stimulation for Parkinson's disease: surgical issues. Mov Disord 21:Suppl 14S197S2182006

30

Sadowski EABennett LKChan MRWentland ALGarrett ALGarrett RW: Nephrogenic systemic fibrosis: risk factors and incidence estimation. Radiology 243:1481572007

31

Sam AD IIMorasch MDCollins JSong GChen RPereles FS: Safety of gadolinium contrast angiography in patients with chronic renal insufficiency. J Vasc Surg 38:3133182003

32

Shamir RRJoskowicz LTamir IDabool EPertman LBen-Ami A: Reduced risk trajectory planning in image-guided keyhole neurosurgery. Med Phys 39:288528952012

33

Thomsen HS: Recent hot topics in contrast media. Eur Radiol 21:4924952011

34

Vertinsky ATCoenen VALang DJKolind SHoney CRLi D: Localization of the subthalamic nucleus: optimization with susceptibility-weighted phase MR imaging. AJNR Am J Neuroradiol 30:171717242009

35

Wiest-Daesslé NPrima SCoupé PMorrissey SBarillot CRician Noise removal by non-local means filtering for low signal-to-noise ratio MRI: applications to DT-MRI. Metaxas DAxel LFichtinger G: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2008 Part II HeidelbergSpringer2008. 171179

36

Xiao YBeriault SPike GBCollins DL: Multicontrast multiecho FLASH MRI for targeting the subthalamic nucleus. Magn Reson Imaging 30:6276402012

37

Zrinzo LFoltynie TLimousin PHariz MI: Reducing hemorrhagic complications in functional neurosurgery: a large case series and systematic literature review. Clinical article. J Neurosurg 116:84942012

TrendMD

Metrics

Metrics

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 142 142 32
PDF Downloads 132 132 21
EPUB Downloads 0 0 0

PubMed

Google Scholar