1,489
Views
29
CrossRef citations to date
0
Altmetric
Biomedical Paper

Surgical targeting accuracy analysis of six methods for subthalamic nucleus deep brain stimulation

, &
Pages 325-334 | Received 10 Apr 2007, Accepted 19 Sep 2007, Published online: 06 Jan 2010

Abstract

A commonly adopted surgical target in deep brain stimulation (DBS) procedures, the subthalamic nucleus (STN) is located deep within the brain and is surrounded by delicate deep-brain structures. Symptoms of Parkinson's disease can be reduced by precisely implanting a multi-electrode stimulator at a specific location within the STN and delivering the appropriate signal to the target. A number of techniques have recently been proposed to facilitate STN DBS surgical targeting and thereby improve the surgical outcome. This paper presents a retrospective study evaluating the target localization accuracy and precision of six approaches in 55 STN DBS procedures. The targeting procedures were performed using a neurosurgical visualization and navigation system, which integrates normalized and standardized anatomical and functional information into the planning environment. In this study, we employed as the “gold standard” the actual surgical target locations determined by an experienced neurosurgeon using both pre-operative image-guided surgical target/trajectory planning and intra-operative electrophysiological exploration and confirmation. The surgical target locations determined using each of the six targeting methods were compared with the “gold standards”. The average displacement between the actual surgical targets and those planned with targeting approaches was 3.0 ± 1.3 mm, 3.0 ± 1.3 mm, 3.0 ± 1.0 mm, 2.6 ± 1.1 mm, 2.5 ± 0.9 mm, and 1.7 ± 0.7 mm for approaches based on T2-weighted MRI, a brain atlas, T1 and T2 maps, an electrophysiological database, a collection of final surgical targets from previous patients, and the combination of these functional and anatomical data, respectively. The technique incorporating both anatomical and functional data provides the most reliable and accurate target position for STN DBS.

Introduction

Effective application of deep brain stimulation (DBS) to the subthalamic nucleus (STN) is capable of dramatically relieving the symptoms of Parkinson's disease and essential tremor in medically refractory patients Citation[1–5]. Although this obliquely oriented, biconvex, lens-shaped mass of grey matter is very small (typically 5.9 mm, 3.7 mm, and 5.0 mm in the anteroposterior, mediolateral, and dorsoventral dimensions, respectively Citation[6]), the STN is considered to be the main “driving” factor in the basal ganglia and plays an important role in the basal ganglia-thalamocortical motor circuit. Located within the caudal diencephalon in the ventral part of the subthalamus, the STN, which is dorsal to the internal capsule and rostral to the substantia nigra, uses the excitatory neurotransmitter glutamate and innervates both basal ganglia output nuclei, the internal segment of the globus pallidus (GPi) and the pars reticulata of the substantia nigra (SNr) Citation[1]. Deep brain stimulation eliminates excessive or abnormally patterned activity from the STN by implanting a multi-electrode stimulator at a specific location within the nucleus, where, continuous high-frequency electrical simulation generated by a neuro pacemaker is delivered Citation[1–5]. Hence, motor abnormalities such as tremor, rigidity, bradykinesia, postural instability, and akinesia induced by the hyperactivity of the STN can be alleviated by STN DBS surgery.

As small deviations in the electrode positioning may cause severe side effects, such as speech and language disorders, muscle contraction, ocular deviation, and visual defects Citation[1–5], it is critical to perform accurate and precise surgical target localization and DBS electrode implantation to achieve the optimal surgical outcome. To minimize the uncertainty and variability of direct targeting on the pre-operative volumetric images, multiple invasive exploratory electrodes are currently required intra-operatively to acquire electrophysiological information for mapping functional borders and refining the final surgical target. Accurate initialization of the surgical target and trajectory planning may diminish procedure-induced intracranial hemorrhage, brain tissue damage, and other related complications by reducing the number of exploratory electrodes and the duration of the surgical process.

Due to the lack of contrast between the STN and surrounding structures on regular CT and T1-weighted MR images, information from these modalities is often complemented with T2-weighted MR images Citation[7], Citation[8], printed Citation[9–11] and digitized anatomical brain atlases Citation[12], Citation[13], Citation[14], Citation[15], histological brain atlases Citation[16], Citation[17], high resolution T1 and T2 maps Citation[18], Citation[19], and functional atlases Citation[20], Citation[21], Citation[22] and databases Citation[23]. In addition, registered surgical targets from previous patients Citation[24], as well as integration of multiple functional and anatomical references Citation[25], Citation[26], may also be employed to facilitate STN DBS surgical targeting. Although the STN is often distinguishable on patient-specific T2-weighted MR images, which allow better visualization of this nucleus, the discrepancy between the target positions localized with T2-weighted MR images and those finalized using electrophysiological measurements should be taken into consideration, and the use of T2-weighted MR images alone should be employed cautiously Citation[27], Citation[28]. Anatomical atlases, once linearly Citation[29] or nonlinearly Citation[14] mapped to individual pre-operative brain images, may provide information about subdivisions of deep-brain nuclei that are not directly visible on standard MR images. Unfortunately, atlases have limitations, such as brain structure discontinuity, insufficient statistical representation of anatomical variability, and lack of functional information, that significantly restrict their applicability and generalization in STN DBS targeting.

To enhance the planning of surgical procedures, several research groups have contributed a number of histological atlases. Chakravarty et al. Citation[16] presented a contiguous 3D histological volume containing a set of 3D objects of the basal ganglia and the thalamic nuclei. The 3D deformable histological atlas described by Yelnik et al. Citation[17] includes not only detailed histological information on the human basal ganglia, but also functional information derived from immuno-histochemical studies. Aligning and fusing high resolution and high signal-to-noise ratio (SNR) MR “standard brain” T1 and T2 maps to each individual patient brain can be helpful, even if subtle anatomical variations not captured by non-linear image registration remain. Patient-specific T1 and T2 maps, acquired within a clinically acceptable timeframe Citation[18], may overcome this problem and enable more accurate surgical target localization.

In addition, functional atlases Citation[20], Citation[21], Citation[22] and databases Citation[23] containing standardized electrophysiological information from multiple patients, normalized into a reference brain template, have been developed to complement existing anatomical and histological atlases. Such electrophysiological information is important for characterizing the function of each deep-brain nucleus and adjacent structures, and for estimating the surgical targets. We believe that stereotactic targeting of the STN using registered surgical targets from previous patients and electrophysiological databases is more promising, since such an approach provides valuable probabilistic maps of actual population-based target information and deep-brain electrophysiological activity. Furthermore, the integration of multiple functional and anatomical datasets can establish the relationship between brain functional organization and anatomic structures for refining and finalizing the surgical targets.

This paper extends earlier work Citation[26] through a more comprehensive retrospective study to evaluate and compare the target localization accuracy and precision of six targeting methods in STN DBS surgery. We addressed three questions in this study: (1) What is the spatial correlation between the actual surgical target and the predicted surgical target based on each targeting method? (2) Can any of these target identification approaches be used to accurately delineate the surgical target? (3) What level of confidence does each technique provide for surgical targeting?

Materials and methods

Targeting techniques

Herzog et al. Citation[30] demonstrated that the effective DBS contacts located in the dorsolateral STN border produced the best clinical results. Neurosurgeons at our institution also regard the dorsolateral portion of the STN as the most favorable stimulation site in STN DBS. Therefore, all the target positions estimated using the following six targeting techniques were selected based on this principle.

T2-weighted MRI-based targeting

Located lateral to the red nucleus and dorsolateral to the substantia nigra, the STN has a higher concentration of iron compared to surrounding tissues. This characteristic causes the anterior part of the STN to have hypointense signal intensity on T2-weighted MR images. Furthermore, the red nucleus can be best visualized on an axial T2-weighted MRI, and its anterior border can be easily identified. Therefore, although it is true that the hypointense signal is not always present in the posterior part of the STN, the visibility of this signal in the anterior portion and the correlation between the STN and the red nucleus and substantia nigra can assist the identification of the surgical target location within the STN. In our study, we employed an inter-modality affine registration algorithm (the FMRIB Linear Image Registration Tool–FLIRT) Citation[31] to align the T2-weighted coronal and axial acquisitions of each patient with his/her pre-operative T1-weighted image. Once the T2-weighted images were co-registered and fused with the T1-weighted data, we chose a point dorsolateral to the center of the STN on the three intersecting slices of these overlapped images.

Anatomical brain atlas-based targeting

Because of its established role in neuroscience research, we used the digitized version of the Schaltenbrand and Wahren stereotactic brain atlas Citation[9]. The STN in this atlas was manually segmented and the binary data resulting from the segmentation were passed through the Marching Cubes algorithm to create the surface model of the nucleus. These results were then non-rigidly transformed to each patient brain image space using the AtamaiWarp algorithm Citation[32]. We finally used the combination of a color-coded brain atlas and a 3D mesh representation of the segmented nucleus to intuitively identify the surgical target inside the STN. As the centroid of the STN surface model can be calculated automatically and displayed along with the atlas and the mesh object, it was also considered as a reference when defining the ideal target location.

T1 and T2 map-based targeting

Noninvasive quantitative imaging methods, DESPOT1 and DESPOT2 (see ), permit the acquisition of high resolution (0.34 mm3 isotropic) and high SNR T1 and T2 maps, respectively Citation[18]. Here, T1 and T2 maps of the deep brain region of a young healthy male volunteer were generated by mutually co-registering and averaging 55 T1 maps and 25 T2 maps. Although some deep-brain nuclei appear indistinguishable from adjacent structures on regular MR images, differences between them are actually visible on these T1 and T2 maps. A generic algorithm (GA) incorporating characteristics of the k-means clustering algorithm Citation[19] was employed to segment the STN on the T1 and T2 maps. Our surgical target identification methodology using the T1 and T2 maps is similar to that employed when using the anatomical brain atlas.

Table I.  Imaging parameters of DESPOT1 and DESPOT2.

Targeting based on the electrophysiological database

Coded in a standard form, the intra-operative electrophysiological microelectrode recording (MER) and electrical stimulation data acquired during each STN DBS procedure were saved in the patient MR image space, then non-rigidly mapped to a standard brain template (Colin27) Citation[33] using AtamaiWarp Citation[32]. Our electrophysiological database for STN DBS currently contains data from a total of 55 procedures. The approximate delineation of the STN for each patient was performed by retrieving the data coded with respect to STN-specific firing patterns from the database, and non-rigidly transforming them to each specific patient brain space. The location of the ideal surgical target was determined from locations in the database that indicated optimal control of the Parkinson's disease symptoms with the fewest side effects. Each patient's own data were excluded when retrieved for target localization for that patient.

Targeting based on actual surgical target collection

For each STN DBS procedure, the actual surgical target position where the effective contact of the final DBS lead was placed within the patient brain space was also warped to the standard brain coordinate system, along with the electrophysiological data. The clustering of surgical targets after non-rigid registration was demonstrated around and within the STN. This targeting method made use of the collection of actual surgical targets from 55 STN DBS procedures by non-rigidly mapping them from the standard brain space to the native brain space of each patient. The automatically generated center of mass (COM) of this cluster of actual targets was directly employed as the target estimation in this targeting protocol. To avoid the influence of the patient's own data on targeting, we removed the actual surgical target of the patient from the cluster when we calculated the COM for that patient.

Comprehensive functional and anatomical information-based targeting

With our visualization and navigation system, we localized the surgical targets of these 55 procedures using all the customized functional and anatomical data. The pre-operative T2-weighted images of each patient co-registered with the T1-weighted image of the same patient were used to initialize the target planning, with the assistance of the warped anatomical brain atlas and T1 and T2 maps. The final surgical target was then further refined and delineated with the non-rigidly mapped electrophysiological data and the COM of the target cluster. The combination of the T2-weighted images, brain atlas, and T1 and T2 maps allows visualization and identification of more detailed deep-brain structures. When the above anatomical information is incorporated with the neurophysiological corroboration provided by the electrophysiological database and real surgical data that are non-rigidly accommodated within the same brain space, the correlation between the deep-brain anatomical structures and the electrophysiological organization can be established.

Visualization and navigation system

Surgical target localization with the six targeting procedures evaluated in this paper was conducted using a neurosurgical visualization and navigation system Citation[25], which is integrated with normalized functional and anatomical information within the standard brain space defined by the Colin27 volume. This comprehensive neurosurgical system readily accommodates inter-subject anatomical variability by the use of a fast 3D non-rigid image registration approach, AtamaiWarp Citation[32]. In addition to a variety of anatomical references, such as the digitized and segmented Schaltenbrand and Wahren brain atlas Citation[9] and high resolution and high SNR quantitative T1 and T2 maps Citation[18], Citation[19], our visualization and navigation system also incorporates functional references, including 3D electrophysiological databases Citation[25] and collections of actual surgical targets from previous patients. This system is capable of not only interactively displaying the linked patient image and standard brain template, reformatting them along arbitrary axes, but also registering and fusing images from different imaging modalities, digitized brain atlases, and functional data with the pre-operative patient images. Manipulation of virtual surgical probes, simulated according to their actual geometry within this surgical environment, is also possible. Data-sampling points may therefore be readily visualized with respect to pre-defined targets and electrophysiological data previously imported from the database.

Patient selection

Our study included data from 28 patients (16 male: age 46–79 years, mean age 60.1 years; 12 female: age 21–76 years, mean age 55.5 years). From May 2000 to November 2006, these 28 patients underwent a total of 55 STN DBS procedures to alleviate their Parkinson's disease symptoms at the London Health Sciences Centre (LHSC). The condition of each patient was diagnosed with a complete neurological exam by a team of specialists consisting of a neurologist, neurosurgeon, neurophysiologist and neuropsychologist. Patients were assessed in both on and off medication states. Only those who had significant improvement with levodopa treatment of their motor symptoms were considered appropriate surgical candidates for STN DBS surgery.

Acquisition of image data

The pre-operative 3D T1-weighted MR brain image volumes of the 28 patients were acquired using a 1.5T GE Signa scanner with a 3D (256 × 256 × 248) SPGR sequence (TR: 8.9 ms, TE: 1.9 ms, flip angle 20°, NEX 2, voxel size 1.17 mm × 1.17 mm × 1 mm). To enhance the definition of the STN, we also obtained 2 sets of coronally (256 × 256 × 21) and axially (256 × 256 × 20) oriented 2D T2-weighted MR image slices using a 2D fast spin-echo sequence (TR: 2800 ms, TE: 110 ms, flip angle 90°, NEX 4, pixel size 1.0 mm × 1.0 mm, slice thickness 1.5 mm, interslice spacing 0 mm) on the same scanner immediately after the T1-weighted image acquisition. The placement of the Leksell G stereotactic frame (Elekta Instruments AB, Stockholm, Sweden) with an attached MR-compatible localizer was performed prior to imaging for each patient.

Surgical procedure

The surgical strategies employed at our institute include pre-operative image-guided surgical target/trajectory planning, intra-operative electrophysiological confirmation, and final stimulator implantation. The first phase is carried out based on pre-operative T1- and T2-weighted MR images, whereas the second step involves intra-operative electrophysiological measurements to refine the optimal surgical targets. The Leksell G frame was carefully mounted on the patient's head with local anesthesia immediately before imaging, and removed immediately after surgery. Following acquisition, the images were transferred to the MR console where the neurosurgeon identified benchmark anatomical structures, such as the anterior commissure (AC) and posterior commissure (PC), relative to the stereotactic frame system on the MR images. The image volumes were then resampled such that the “axial plane” was parallel to the AC-PC plane. The section of an anatomical brain atlas closest to the images was employed to select the initial surgical target. Stereotactic coordinates were assigned to the target relative to the frame center. Intra-operatively, before the final placement of the deep brain stimulator, surgeons at LHSC use a five-electrode assembly (“Ben gun”) Citation[2] to obtain the microelectrode recording (MER) and electrical stimulation measurements on a track extending 10 mm above and below the initially planned target to map the functional characteristics of subcortical nuclei and refine target localization.

“Gold standard”

In this work, we considered the actual surgical target locations determined by an experienced stereotactic neurosurgeon using standard surgical procedures as the “gold standard”, or true locations of the theoretical targets. The target positions determined with each of the six proposed targeting techniques were compared with the corresponding “gold standard” to evaluate the target localization accuracy and precision of each procedure.

Results

displays the absolute difference between the actual surgical target location determined by a stereotactic neurosurgeon using pre-operative MR image-based planning and intra-operative electrophysiological explorations and that estimated by a non-neurosurgeon using each of the six targeting approaches based on the T2-weighted MRI, brain atlas, T1 and T2 maps, electrophysiological database, previous surgical targets, and combination data for each of the 55 surgical procedures. Two-tailed Pearson correlation tests demonstrated that there is no statistically significant dependence between the results generated by any two target localization methods (−0.21 < r < 0.23, p > 0.1). The average, maximum, minimum, and standard deviation for the displacements between the “gold standard” and the predicted surgical targets based on each targeting method were computed and are summarized in .

Figure 1. Scatter plot displaying the absolute difference between the actual surgical target for each of the 55 procedures and that estimated with each of the six targeting approaches. [Color version available online.]

Figure 1. Scatter plot displaying the absolute difference between the actual surgical target for each of the 55 procedures and that estimated with each of the six targeting approaches. [Color version available online.]

Table II.  Absolute differences between actual surgical targets and targets estimated with each targeting method.

We employed a one-way analysis of variance (ANOVA) to analyze the differences in surgical targeting accuracy among the six target localization methods, which revealed significant statistical differences between the groups (F = 12.57, df = 5, p < 0.001). Post-hoc tests were carried out to demonstrate which specific targeting technique could provide more accurate target localization. The results of the Tukey's HSD post-hoc test showed that the localization accuracy of the comprehensive targeting approach was significantly better than that of the methods based on the T1 and T2 maps, brain atlas, T2-weighted MRI, and electrophysiological database, as well as the collection of previous final surgical targets (p < 0.01). The mean error of the comprehensive targeting approach was at least 0.8 mm less than that of the other five methods. shows that the hypointense signal of the STN on the patient T2-weighted image has better correlation with the segmented nucleus derived from the T1 and T2 maps than with that from the digitized Schaltenbrand atlas. The spatial relationship between the actual surgical target of a patient and those identified using the six targeting methods is clearly displayed in .

Figure 2. T2-weighted MR image of a patient demonstrating direct visualization of STN: (a) axial slice; (b) coronal slice. The contour of the segmented STN based on T1 and T2 maps is marked in yellow; that based on the digitized Schaltenbrand atlas is marked in blue. The small white sphere is the actual surgical target. The digitized atlas and T1 and T2 maps were registered to the patient brain image space. [Color version available online.]

Figure 2. T2-weighted MR image of a patient demonstrating direct visualization of STN: (a) axial slice; (b) coronal slice. The contour of the segmented STN based on T1 and T2 maps is marked in yellow; that based on the digitized Schaltenbrand atlas is marked in blue. The small white sphere is the actual surgical target. The digitized atlas and T1 and T2 maps were registered to the patient brain image space. [Color version available online.]

Figure 3. Positions of surgical targets estimated with each targeting approach. Yellow sphere: target based on the T2-weighted MR image. Green sphere: target based on the Schaltenbrand atlas. Magenta sphere: target based on T1 and T2 maps. Blue sphere: target based on the electrophysiological database. Cyan sphere: target based on previous surgical targets. Red sphere: target based on combined functional and anatomical data. White sphere: “gold standard” actual surgical target. [Color version available online.]

Figure 3. Positions of surgical targets estimated with each targeting approach. Yellow sphere: target based on the T2-weighted MR image. Green sphere: target based on the Schaltenbrand atlas. Magenta sphere: target based on T1 and T2 maps. Blue sphere: target based on the electrophysiological database. Cyan sphere: target based on previous surgical targets. Red sphere: target based on combined functional and anatomical data. White sphere: “gold standard” actual surgical target. [Color version available online.]

The results illustrate that the location of the surgical targets estimated with the comprehensive functional and anatomical information-based targeting method fell within 2 mm of the optimal surgical targets, indicating the importance of correlating functional organization with deep-brain anatomy. Since, in practice, the current spread generated at the commonly administered stimulation levels is in the range of 2 mm Citation[34], neurosurgeons may easily approach the optimal surgical target region by slightly adjusting the electrode whose initial location was planned with this technique, incorporating both anatomical and functional references. The standard deviation for the combined method was also smaller than that for the other five methods, demonstrating that the targeting precision of this technique was superior to that of the others. The integration of different forms of functional and anatomical information contained within the neurosurgical visualization and navigation system can compensate for their inherent shortcomings and take advantage of their respective strengths to effectively localize the surgical target for STN DBS procedures. In summary, the electrophysiological database, the actual surgical target collection, and the integration of approaches based on functional and anatomical information provided more accurate initial estimation of the surgical target positions than those techniques dependent solely on anatomical references.

Discussion

In this work, a retrospective study of 28 patients who had been treated with bilateral or unilateral STN DBS (55 procedures in total) was conducted to evaluate and compare the target localization accuracy and precision of six targeting methods in STN DBS surgery. From the analysis, we determined the spatial correlation between real and predicted surgical targets, and the capability of these target identification techniques to delineate the theoretical surgical target, as well as the level of confidence that each method could provide in surgical targeting. The results demonstrated that each method was effective, but that the target location estimated with the comprehensive functional and anatomical information-based targeting method was much closer to the optimal surgical target position than those planned using the other five approaches. In addition, because using this combined targeting method results in reduced standard deviation, precision in localizing the optimal surgical target can be enhanced.

When it comes to actual clinical applications, in addition to accuracy, the feasibility or applicability of each technique must also be considered. We need to answer questions such as which method requires the shortest time or least effort; which can be conducted at other neurosurgical sites; and which can be performed within a reasonable timeframe without compromising its accuracy? In our study, the fastest approach to planning a surgical target is based on T2-weighted MRI, which must be linearly registered to the T1-weighted image. This procedure can be performed in 15 min including the time for the registration step. Nevertheless, although it is commonly used in many clinical institutions, employing T2-weighted MR images alone may result in less accurate target localization as we, as well as others Citation[27], Citation[28] have shown. The other five targeting methods all require the use of non-rigid registration. AtamaiWarp as adopted in this work takes approximately 9 min on a 3-GHz Pentium 4 machine to perform the deformation with an average registration error of 1.0 ± 0.7 mm Citation[35]. The method based on the registered surgical target collection from previous patients is equally time-efficient, because the automatically generated center of mass (COM) of this cluster of actual targets can be directly employed as the target. The accuracy of this method is highly dependent on the non-rigid registration. Similar to the technique based on the brain atlas, the procedure employing T1 and T2 maps allows intuitive identification of the surgical target inside the 3D mesh representation of the segmented STN relative to its centroid. Each approach can be carried out within 20 min (including the non-rigid registration step) with comparable accuracy. Planning the desired surgical target using the electrophysiological database entails additional time and comprehension of the physiology. On the other hand, the functional information is necessary for planning the surgical trajectory and is essential to achieving an optimal surgical outcome. The most accurate method employs all the available anatomical and functional data. Although this combined method requires a longer time for the pre-processing of all the data, once displayed in the common brain space the comprehensive representation of anatomical structures and neurophysiological information facilitates the identification of the optimal target location. Since this method was demonstrated to be significantly more accurate than the other four and the time needed (25–30 min total) is practically acceptable, we consider it to be the most favorable targeting approach. The computational speed of the registration steps scales with processor speed and number of processors. Possible errors in target identification may be an indirect result of the registration algorithm, because non-rigid registration was involved in most of the target localization processes.

We analyzed the accuracy of targeting methods based on combinations of each one of the three anatomical references (i.e., the T2-weighted MRI, brain atlas, and T1 and T2 maps) and one or both of the functional references (i.e., the electrophysiological database and actual surgical target collection). However, although in some cases the surgical targets defined using the combination of available data from a smaller group of references are close to those determined by the combination of data from all five, the results vary from case to case, and an optimal estimation of the surgical target cannot be guaranteed. The target defined using the comprehensive functional and anatomical information-based targeting method is closer (> 0.5 mm) to the “gold standard” than that determined using the above combinations. From the Pearson correlation study, we showed that none of the methods significantly correlates with any other method. Therefore, the six targeting methods cannot replace each other in terms of the identification of the optimal surgical target. Since each of the five methods provides different, unique anatomical or functional information, the combination of these approaches can complement the information provided by any individual method.

From a statistical point of view, adopting the average of the repeated measurements performed with each target localization approach could reduce the uncertainty and improve the precision of targeting. In this case, in order to obtain sufficient degrees of freedom, a number of repetitions must be performed. To ensure that the trials are independent (i.e., that no “learning effect” is carried over from one to the next), a significant delay (∼1 hour) must be inserted between trials, which would result in a long duration for the process. This strategy is impractical for actual clinical application, because the target identification procedure must be performed in an efficient and timely manner immediately after pre-operative image acquisition and prior to the surgery. Therefore, performing single target localization for each patient using each targeting technique is more appropriate for this study. The improved precision and reduced variance can also be achieved by using the comprehensive multiple functional and anatomical references. Furthermore, because the functional activity does not always correspond exactly to the anatomy, it is necessary to establish the correlation between the customized population-based electrophysiological database and the patient-specific detailed anatomical structures to improve the accuracy and precision of targeting.

Nowinski et al. constructed a probabilistic functional STN atlas Citation[36] representing the normalized spatial distribution of the best contacts of DBS electrodes from 184 Parkinson's disease patients. In this functional STN atlas, the value at any given location was defined by the number of the best contacts at this location. It has been shown that the part of this functional STN atlas containing medium or high probability values correlates well with the anatomical STN derived from the Schaltenbrand-Wahren brain atlas. Without further clinical validation, the current surgical target localization techniques may be restricted to providing only an initial surgical target position. Although accurate surgical target initiation may reduce the need for invasive intra-operative exploration, and thereby decrease the surgical duration and procedure-related complications, certain intra-operative electrophysiological measurements may still be required to compensate for possible inadequacy of these targeting methods. During the operation, brain tissue may shift slightly from the pre-operative MR image due to cranial opening, electrode insertion, and cerebrospinal fluid (CSF) leakage. Ideally, one would characterize the degree of brain shift by registering the intra-operative MR image to the pre-operative image. However, Bardinet et al. Citation[37] demonstrated that the deformation caused by CSF leakage around the functional target region is minimal (<1 mm) and does not affect the accuracy of target localization performed in pre-operative image space.

References

  • Starr PA, Vitek JL, Bakay RAE. Ablative surgery and deep brain stimulation for Parkinson's disease. Neurosurgery 1998; 43(5)989–1013
  • Benabid AL. Deep brain stimulation for Parkinson's disease. Curr Opin Neurobiol 2003; 13(6)696–706
  • Halpern C, Hurtig H, Jaggi J, Grossman M, Won M, Baltuch G. Deep brain stimulation in neurologic disorders. Parkinsonism Relat Disord 2007; 13(1)1–16
  • Hamani C, Richter E, Schwalb JM, Lozano AM. Bilateral subthalamic nucleus stimulation for Parkinson's disease: A systematic review of the clinical literature. Neurosurgery 2005; 56(6)1313–1324
  • Machado A, Rezai AR, Kopell BH, Gross RE, Sharan AD, Benabid AL. Deep brain stimulation for Parkinson's disease: Surgical technique and perioperative management. Mov Disord 2006; 21(suppl 14)S247–S258
  • Richter EO, Hoque T, Halliday W, Lozano AM, Saint-Cyr JA. Determining the position and size of the subthalamic nucleus based on magnetic resonance imaging results in patients with advanced Parkinson disease. J Neurosurg 2004; 100(3)541–546
  • Patel NK, Heywood P, O'Sullivan K, Love S, Gill SS. MRI-directed subthalamic nucleus surgery for Parkinson's disease. Stereotact Funct Neurosurg 2002; 78(3–4)132–145
  • Slavin KV, Thulborn KR, Wess C, Nersesyan H. Direct visualization of the human subthalamic nucleus with 3T MR imaging. Am J Neuroradiol 2006; 27(1)80–84
  • Schaltenbrand G, Wahren W. Atlas for Stereotaxy of the Human Brain. Thieme, Stuttgart 1977
  • Talairach J, Tournoux P. Co-Planar Stereotaxic Atlas of the Human Brain. Thieme, Stuttgart 1988
  • Van Buren JM, Borke RC. Variations and Connections of the Human Thalamus. Springer, Berlin 1972
  • Nowinski WL, Yang GL, Yeo TT. Computer-aided stereotactic functional neurosurgery enhanced by the use of the multiple brain atlas database. IEEE Trans Med Imag 2000; 19(1)62–69
  • Ganser KA, Dickhaus H, Metzner R, Wirtz CR. A deformable digital brain atlas system according to Talairach and Tournoux. Med Imag Anal 2004; 8(1)3–22
  • St-Jean P, Sadikot AF, Collins L, Clonda D, Kasrai R, Evans AC, Peters TM. Automated atlas integration and interactive three-dimensional visualization tools for planning and guidance in functional neurosurgery. IEEE Trans Med Imag 1998; 17(5)672–680
  • Schiemann T, Hoehne KH, Koch C, Pommert A, Riemer M, Schubert R, Thiede U. Interpretation of tomographic images using automatic atlas lookup. Proceedings of SPIE Visualization in Biomedical Computing 1994 (VBC’94), RA Robb. Rochester, , MN October, 1994, Proc SPIE 1994;2359:457-465
  • Chakravarty MM, Bertrand G, Hodge CP, Sadikot AF, Collins DL. The creation of a brain atlas for image guided neurosurgery using serial histological data. NeuroImage 2006; 30(2)359–376
  • Yelnik J, Bardinet E, Dormont D, Malandain G, Ourselin S, Tande D, Karachi C, Ayache N, Cornu P, Agid Y. A three-dimensional, histological and deformable atlas of the human basal ganglia. I. Atlas construction based on immunohistochemical and MRI data. NeuroImage 2007; 34(2)618–638
  • Deoni SC, Josseau MJ, Rutt BK, Peters TM. Visualization of thalamic nuclei on high resolution, multi-averaged T1 and T2 maps acquired at 1.5T. Hum Brain Mapp 2005; 25(3)353–359
  • Deoni SC, Rutt BK, Parrent AG, Peters TM. Segmentation of thalamic nuclei using a modified k-means algorithm and high-resolution quantitative magnetic resonance imaging at 1.5T. NeuroImage 2007; 34(1)117–126
  • Chakravarty MM, Sadikot AF, Mongia S, Bertrand G, Collins DL (2006) Towards a multi-modal atlas for neurosurgical planning. Proceedings of the 9th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2006), CopenhagenDenmark, October, 2006, R Larsen, M Nielsen, J Sporring. Springer, Berlin, 389–396, Lecture Notes in Computer Science 4191
  • Nowinski WL, Belov D, Benabid AL. An algorithm for rapid calculation of a probabilistic functional atlas of subcortical structures from electrophysiological data collected during functional neurosurgery procedures. NeuroImage 2003; 18(1)143–155
  • D’Haese PF, Cetinkaya E, Konrad PE, Kao C, Dawant BM. Computer-aided placement of deep brain stimulators: From planning to intraoperative guidance. IEEE Trans Med Imag 2005; 24(11)1469–1478
  • Finnis KW, Starreveld YP, Parrent AG, Sadikot AF, Peters TM. Three-dimensional database of subcortical electrophysiology for image-guided stereotactic functional neurosurgery. IEEE Trans Med Imag 2003; 22(1)93–104
  • D’Haese PF, Pallavaram S, Niermann K, Spooner J, Kao C, Dawant BM (2005) Automatic selection of DBS target points using multiple electrophysiological atlases Part II. Lecture Notes in Computer Science 3750. Proceedings of the 8th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2005), Palm Springs, CA, October, 2005, J Duncan, G Gerig. Berlin Springer, 427–434
  • Guo T, Finnis KW, Parrent AG, Peters TM. Visualization and navigation system development and application for stereotactic deep-brain neurosurgeries. Comput Aided Surg 2006; 11(5)231–239
  • Guo T, Finnis KW, Deoni SCL, Parrent AG, Peters TM (2006) Comparison of different targeting methods for subthalamic nucleus deep brain stimulation Lecture Notes in Computer Science 4190. Proceedings of the 9th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2006), CopenhagenDenmark, October, 2006, R Larsen, M Nielsen, J Sporring. Springer, Berlin, 768–775
  • Cuny E, Guehl D, Burbaud P, Gross C, Dousset V, Rougier A. Lack of agreement between direct magnetic resonance imaging and statistical determination of a subthalamic target: The role of electrophysiological guidance. J Neurosurg 2002; 97(3)591–597
  • Hamani C, Richter EO, Andrade-Souza Y, Hutchison W, Saint-Cyr JA, Lozano AM. Correspondence of microelectrode mapping with magnetic resonance imaging for subthalamic nucleus procedures. Surg Neurol 2005; 63(3)249–253
  • Lehman RM, Zheng J, Hamilton JL, Micheli-Tzanako E. Comparison of 3-D stereoscopic MR imaging with pre and post lesion recording in pallidotomy. Acta Neurochir (Wien) 2000; 142(3)319–328
  • Herzog J, Fietzek U, Hamel W, Morsnowski A, Steigerwald F, Schrader B, Weinert D, Pfister G, Muller D, Mehdorn HM, et al. Most effective stimulation site in subthalamic deep brain stimulation for Parkinson's disease. Mov Disord 2004; 19(9)1050–1054
  • Jenkinson M, Smith S. A global optimization method for robust affine registration of brain images. Med Image Anal 2001; 5(2)143–156
  • Starreveld YP. Fast nonlinear registration applied to stereotactic functional neurosurgery, PhD dissertation. University of Western Ontario, London June, 2002
  • Holmes CJ, Hoge R, Collins L, Woods R, Toga AW, Evans AC. Enhancement of MR images using registration for signal averaging. J Comput Assist Tomogr 1998; 22(2)324–333
  • Wu YR, Levy R, Ashby P, Tasker RR, Dostrovsky JO. Does stimulation of the GPi control dyskinesia by activating inhibitory axons?. Mov Disord 2001; 16(2)208–216
  • Guo T, Starreveld YP, Peters TM. Evaluation and validation methods for intersubject non-rigid 3D image registration of the human brain. Proceedings of SPIE Medical Imaging 2005: Visualization, Image-Guided Procedures, and Display, San Diego, CA, February 2005. Proc SPIE 2005;5744:, RL Galloway, Jr, KR Cleary. 594–603
  • Nowinski WL, Thirunavuukarasuu A, Liu J, Benabid AL. Correlation between the anatomical and functional human subthalamic nucleus. Stereotact Funct Neurosurg 2007; 85(2–3)88–93
  • Bardinet E, Cathier P, Roche A, Ayache N, Dormont D (2002) A posteriori validation of pre-operative planning in functional neurosurgery by quantification of brain pneumocephalus Lecture Notes in Computer Science 2488. Proceedings of the 6th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2002), TokyoJapan, September, 2002, T Dohi, R Kikinis. Springer, Berlin, 323–330

Reprints and Corporate Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

To request a reprint or corporate permissions for this article, please click on the relevant link below:

Academic Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

Obtain permissions instantly via Rightslink by clicking on the button below:

If you are unable to obtain permissions via Rightslink, please complete and submit this Permissions form. For more information, please visit our Permissions help page.