Abstract BACKGROUND Tractography is a popular tool for visualizing the corticospinal tract (CST). However, results may be influenced by numerous variables, eg, the selection of seeding regions of interests (ROIs) or the chosen tracking algorithm. OBJECTIVE To compare different variable sets by correlating tractography results with intraoperative subcortical stimulation of the CST, correcting intraoperative brain shift by the use of intraoperative MRI. METHODS Seeding ROIs were created by means of motor cortex segmentation, functional MRI (fMRI), and navigated transcranial magnetic stimulation (nTMS). Based on these ROIs, tractography was run for each patient using a deterministic and a probabilistic algorithm. Tractographies were processed on pre- and postoperatively acquired data. RESULTS Using a linear mixed effects statistical model, best correlation between subcortical stimulation intensity and the distance between tractography and stimulation sites was achieved by using the segmented motor cortex as seeding ROI and applying the probabilistic algorithm on preoperatively acquired imaging sequences. Tractographies based on fMRI or nTMS results differed very little, but with enlargement of positive nTMS sites the stimulation-distance correlation of nTMS-based tractography improved. CONCLUSION Our results underline that the use of tractography demands for careful interpretation of its virtual results by considering all influencing variables. Accuracy, Algorithm, fMRI, iMRI, nTMS, Tractography ABBREVIATIONS ABBREVIATIONS AH abductor halluces BOLD blood oxygen level dependent CST corticospinal tract DCS direct cortical stimulation DEC direction-encoded color DSS direct subcortical stimulation DTI diffusion tensor imaging DWI diffusion-weighted imaging EOR extent of tumor resection EPI echo-planar imaging EXT extensor digitorum FA fractional anisotropy FDI first dorsal interosseus fMRI functional magnetic resonance imaging FOV field of view FT fiber tracking iMRI intraoperative magnetic resonance imaging MEP motor evoked potential MERIT MeVis Image Registration Toolkit MRI magnetic resonance imaging NBS Navigated Brain Stimulation nTMS navigated transcranial magnetic stimulation RMT resting motor threshold ROI regions of interest TA tibial anterior WHO World Health Organization The significant effect of the extent of tumor resection (EOR) on patients’ overall survival has been extensively demonstrated for both, low-grade as well as high-grade gliomas.1–4 Nevertheless, these progressive tumors are incurable up to now, and high-grade gliomas (World Health Organization [WHO] grade III and IV) still have a high lethality.2 It is therefore essential to maintain or even improve high-grade glioma patients’ quality of life. As a consequence, gliomas located near eloquent functional brain areas are a major challenge for surgical therapy, confronting neurosurgeons with the conflict of maximizing EOR while preserving neurological function. This problem can best be managed by using intraoperative mapping, such as direct cortical or subcortical stimulation (DCS, DSS). In a huge meta-analysis including 8091 patients operated on for supratentorial infiltrative gliomas, late severe deficits were observed significantly less frequently after surgery with intraoperative mapping.5 Since intraoperative mapping provides reassurance to the surgeon about location of cerebral function, greater EOR can be achieved.6 This effect can be supplemented by the use of intraoperative magnetic resonance imaging (iMRI), providing the surgeon with real-time information on residual tumor and on spatial relationship to functional boundaries, leading to a further maximization of the EOR.7,8 Taking all of these findings into account, the application of iMRI combined with intraoperative mapping can be considered a gold standard in the neurosurgical therapy of eloquently located gliomas. Besides intraoperative accuracy, diffusion tensor imaging (DTI) tractography used for presurgical planning of glioma resection near the corticospinal tract (CST) has been shown to further increase patient safety.9 Bello et al10 reported on a reduction in surgery time, intraoperative seizures, and postoperative patient fatigue applying preoperative DTI tractography and intraoperative DSS for glioma resection. The use of preoperative DTI tractography in combination with DCS/DSS additionally increased both, EOR and functional outcome, especially for subcortical tumors.11,12 However, since DTI tractography based on presurgically acquired radiological sequences does not account for intraoperative brain shift that occurs after craniotomy and tumor removal, intraoperative DTI for tractography control and neuronavigation update has been proposed.13 In particular, resections of contrast enhancing tumors lead to a high brain shift, and therefore demand for intraoperative reassessment.14 Within DTI tractography, several parameters and technologies exist all having influence on the final result.15 Among these, the selection of seeding regions of interest (ROIs), the fiber tracking (FT) algorithm, and the time of DTI acquisition are the most decisive. So far, research on DTI tractography for neurosurgical purposes was confined to the assessment of single aspects and parameters, but failed to integratively analyze a set of variables over 1 patient cohort which we consider a relevant shortcoming. Moreover, standardization still is lacking. Clark et al16 formulated this problem already in 2003, calling FT accuracy into question. Therefore, this study aimed at addressing this problem by investigating the influence of different parameters on FT accuracy, including 3 different seeding methods, one of them with regard on limb representation areas, 2 FT algorithms, and 2 points in time of data acquisition. In doing so, complementing iMRI with intraoperative mapping allowed for minimal periprocedural interferences. Results of the present study should help to better interpret tractography results when considering all influencing parameters. METHODS This study was performed in accordance with the ethics standards laid down in the Declaration of Helsinki17 and was approved by the local ethics committee (approval number SNO-06-2015). Written informed consent was obtained from each participant prior to inclusion. Subjects Between 01/2012 and 02/2014, all patients undergoing surgical resection of brain lesions adjacent to or involving the precentral gyrus and/or the CST were prospectively identified. Subsequent selection criteria for study inclusion were diagnosis of WHO grade III or IV glioma or of brain infiltrating metastasis, preserved motor function sufficient to perform motor tasks for fMRI acquisition, absence of cardiac pacemaker and metal implants, capacity to consent as well as obtained written agreement of the patient to participate in the study. Initially, 15 patients were selected. However, 4 of them had to be excluded from the study analysis, since their image data were insufficient or inappropriate for performing the image registration and subsequent calculations. Thus, 11 patients (8 men and 3 women, aged 15-77, mean 55.2 ± 17.9 yr) suffering from subcortically extending high-grade gliomas or metastasis in proximity to or within the precentral gyrus and/or the CST were included in the study (Table 1). In all patients, surgery was performed under continuous neurophysiological monitoring and repetitive cortical and subcortical monopolar mapping. At the end of tumor resection, iMRI was carried out for EOR control. Subcortical stimulation was repeated after iMRI acquisition in order to record stimulation sites for later coregistration with pre- and postoperative imaging. TABLE 1. Characteristics of Patients, Tumors, Subcortical Stimulation Threshold, Extent of Resection, and Motor Deficits Minimum stimulation threshold Motor deficit (MRC) Case No. Sex/Age (yr) Pathology Tumor location Upper limb Lower limb EoR Preop Postop 3 mo 1 M/52 GBM central R 4 mA 10 mA STR HP L 1/5 HP L 3/5 UL L 4/5 2 F/64 Metastasis frontal R 10 mA – GTR 5/5 5/5 5/5 3 M/77 GBM frontal R 6 mA 17 mA GTR peri facial palsy L HP L 4/5 HP L 1/5* 4 M/66 GBM postcentral R 10 mA 18 mA GTR 5/5 5/5 5/5 5 M/72 GBM parietal R 11 mA 13 mA GTR 5/5 5/5 5/5 6 M/69 GBM parietal R 14 mA 13 mA STR HP L 4/5 HP L 4/5 l LL L 0/5, UL 2/5* 7 M/15 Ependymoma III postcentral L 11 mA 17 mA GTR 5/5 5/5 5/5 8 M/55 GBM postcentral R 8 mA 11 mA STR 5/5 5/5 5/5 9 M/39 AA III precentral L 6 mA 9 mA GTR HP R 4/5 HP R 4/5 5/5 10 F/44 OA III central L 6 mA 3 mA STR 5/5 5/5 5/5 11 F/50 GBM postcentral R 3 mA 3 mA GTR 5/5 5/5 5/5 Minimum stimulation threshold Motor deficit (MRC) Case No. Sex/Age (yr) Pathology Tumor location Upper limb Lower limb EoR Preop Postop 3 mo 1 M/52 GBM central R 4 mA 10 mA STR HP L 1/5 HP L 3/5 UL L 4/5 2 F/64 Metastasis frontal R 10 mA – GTR 5/5 5/5 5/5 3 M/77 GBM frontal R 6 mA 17 mA GTR peri facial palsy L HP L 4/5 HP L 1/5* 4 M/66 GBM postcentral R 10 mA 18 mA GTR 5/5 5/5 5/5 5 M/72 GBM parietal R 11 mA 13 mA GTR 5/5 5/5 5/5 6 M/69 GBM parietal R 14 mA 13 mA STR HP L 4/5 HP L 4/5 l LL L 0/5, UL 2/5* 7 M/15 Ependymoma III postcentral L 11 mA 17 mA GTR 5/5 5/5 5/5 8 M/55 GBM postcentral R 8 mA 11 mA STR 5/5 5/5 5/5 9 M/39 AA III precentral L 6 mA 9 mA GTR HP R 4/5 HP R 4/5 5/5 10 F/44 OA III central L 6 mA 3 mA STR 5/5 5/5 5/5 11 F/50 GBM postcentral R 3 mA 3 mA GTR 5/5 5/5 5/5 EoR, extent of resection; MRC, Medical Research Council Scale for Muscle strength, grades 1 to 5; M, male; F, female; R, right; L, left; GBM, glioblastoma; AA, anaplastic astrocytoma; OA, oligoastrocytoma; GTR, gross total resection; STR, subtotal resection; HP, hemiparesis; UL, upper limb; LL, lower limb. View Large TABLE 1. Characteristics of Patients, Tumors, Subcortical Stimulation Threshold, Extent of Resection, and Motor Deficits Minimum stimulation threshold Motor deficit (MRC) Case No. Sex/Age (yr) Pathology Tumor location Upper limb Lower limb EoR Preop Postop 3 mo 1 M/52 GBM central R 4 mA 10 mA STR HP L 1/5 HP L 3/5 UL L 4/5 2 F/64 Metastasis frontal R 10 mA – GTR 5/5 5/5 5/5 3 M/77 GBM frontal R 6 mA 17 mA GTR peri facial palsy L HP L 4/5 HP L 1/5* 4 M/66 GBM postcentral R 10 mA 18 mA GTR 5/5 5/5 5/5 5 M/72 GBM parietal R 11 mA 13 mA GTR 5/5 5/5 5/5 6 M/69 GBM parietal R 14 mA 13 mA STR HP L 4/5 HP L 4/5 l LL L 0/5, UL 2/5* 7 M/15 Ependymoma III postcentral L 11 mA 17 mA GTR 5/5 5/5 5/5 8 M/55 GBM postcentral R 8 mA 11 mA STR 5/5 5/5 5/5 9 M/39 AA III precentral L 6 mA 9 mA GTR HP R 4/5 HP R 4/5 5/5 10 F/44 OA III central L 6 mA 3 mA STR 5/5 5/5 5/5 11 F/50 GBM postcentral R 3 mA 3 mA GTR 5/5 5/5 5/5 Minimum stimulation threshold Motor deficit (MRC) Case No. Sex/Age (yr) Pathology Tumor location Upper limb Lower limb EoR Preop Postop 3 mo 1 M/52 GBM central R 4 mA 10 mA STR HP L 1/5 HP L 3/5 UL L 4/5 2 F/64 Metastasis frontal R 10 mA – GTR 5/5 5/5 5/5 3 M/77 GBM frontal R 6 mA 17 mA GTR peri facial palsy L HP L 4/5 HP L 1/5* 4 M/66 GBM postcentral R 10 mA 18 mA GTR 5/5 5/5 5/5 5 M/72 GBM parietal R 11 mA 13 mA GTR 5/5 5/5 5/5 6 M/69 GBM parietal R 14 mA 13 mA STR HP L 4/5 HP L 4/5 l LL L 0/5, UL 2/5* 7 M/15 Ependymoma III postcentral L 11 mA 17 mA GTR 5/5 5/5 5/5 8 M/55 GBM postcentral R 8 mA 11 mA STR 5/5 5/5 5/5 9 M/39 AA III precentral L 6 mA 9 mA GTR HP R 4/5 HP R 4/5 5/5 10 F/44 OA III central L 6 mA 3 mA STR 5/5 5/5 5/5 11 F/50 GBM postcentral R 3 mA 3 mA GTR 5/5 5/5 5/5 EoR, extent of resection; MRC, Medical Research Council Scale for Muscle strength, grades 1 to 5; M, male; F, female; R, right; L, left; GBM, glioblastoma; AA, anaplastic astrocytoma; OA, oligoastrocytoma; GTR, gross total resection; STR, subtotal resection; HP, hemiparesis; UL, upper limb; LL, lower limb. View Large Acquisition Parameters Protocol Design The aim of this study was to determine which combination of parameters allows for FT most realistically depicting the CST. Therefore, all available parameters consisting of 3 different seeding methods (“SEEDING”; fMRI, navigated transcranial magnetic stimulation [nTMS], segmentation of the motor cortex), 2 of them considering limb representation areas (“ANATOMY”), 2 FT algorithms (“FT ALGORITHM”; deterministic and probabilistic), and MRI data acquired at 2 points in time (“TIME”; pre- and postoperatively) were combined for tracking the CST, resulting in 20 tractographies for each patient (see Figure 1). Each tractography was evaluated by correlating the distance between the distinctive fiber tract and the post-iMRI stimulation site with the minimum stimulation intensity evoking motor evoked potentials (MEPs) of hand or leg muscles at this site. FIGURE 1. View largeDownload slide Structure of acquired data for each patient. FIGURE 1. View largeDownload slide Structure of acquired data for each patient. MRI Data Acquisition All MRI data sets were collected using a 3 T Magnetom Verio System (Siemens Medical Systems, Erlangen, Germany) with an 8-channel phased array head coil. Besides standard MRI sequences (T1-weighted (w), T2-w, FLAIR, T2*-w, diffusion-weighted imaging [DWI], T1-w with contrast agent), the protocol included the following: 1. A 3-dimensional (3-D) T1-magnetization-prepared rapid gradient-echo (MPRAGE) sequence with repetition time (TR) = 2250 ms; echo time (TE) = 2.25 ms; inversion time (TI) = 900 ms; 9° flip angle; 176 sagittal partitions; 1 mm slice thickness; field of view (FOV) = 250 mm²; 512 × 512 matrix, performed after intravenous administration of gadolinium (0.2 ml/kg body weight) 2. A diffusion-weighted image sequence with 20 diffusion directions and a b-value of 700 s/mm2 allowing DTI. Other parameters were TR = 9200 ms; TE = 72 ms; 90° flip angle; 50 axial slices; 1.9 mm slice thickness without gap; FOV = 260 mm2; 136 × 136 matrix; 2 acquisitions; GRAPPA acceleration factor = 3. 3. A blood oxygen level dependent (BOLD) sensitive echo-planar imaging (EPI) sequence to perform functional MRI of motor function with TR = 2000 ms; TE = 30 ms; 90° flip angle; 36 transverse slices; 3 mm slice thickness; 3 × 3 × 3 mm3 voxel size; matrix = 64 × 64. The paradigm was a simple visually guided motor task. Subjects performed 3 blocks of alternating movements of the tongue and flexion movements of the fingers and feet, each contralateral to the affected brain hemisphere (1 block = 21 s, alternation frequency 2 Hz). Data Processing fMRI ROI fMRI data processing was performed with BrainVoyager® (©Rainer Goebel). The steps included interslice scan time correction with sinc interpolation, 3-D motion correction, and high-pass temporal filtering of 3-D data points, and signal linear drift removal. A linear correlation analysis of the measured data was performed based on a reference time course function predicting the expected BOLD signal. For this reference time course, the resting condition for 1 motor task was defined by the other 2 conditions; for example, the time periods of tongue and finger movements were used as a resting period to analyze the foot activation. In a first step, voxels with a correlation coefficient r, indicating the strength of covariation, greater than 0.4 were color-coded. Subsequently, the highest level of r value which encoded only single voxels in the targeted motor area was fixed as rmax to adapt to the individual level of activation. This visually adjusted individual threshold was multiplied with 0.66 (rmax × 0.66) to guarantee a critical size of the fMRI activation areal.18 The threshold values for the single motor areas were adjusted by visual inspection for quality assurance. The coregistration of functional and 3-D anatomic data sets was initiated by automatic header-based alignment, followed by manual adjustment of rotational and translational alignment. Finally, the functional areas were superimposed on the anatomic 3-D T1-w images. nTMS ROI Navigated single-pulse TMS was performed in all 11 patients with the eXimia Navigated Brain Stimulation (NBS) System (software version 4.2; Nexstim Plc, Helsinki, Finland) using a figure-of-8 coil (coil diameter 50 mm; biphasic pulse, pulse length 280 μs). After loading the T1-weighted magnetization-prepared rapid gradient-echo (MPRAGE) sequence into the NBS system for 3-D anatomic reconstruction of the brain, the patient's head was coregistered with the image data set by using anatomic landmarks and surface matching via the integrated optical tracking system. The depth of the displayed 3-D head model then was adjusted, so that both the cortex and the tumor were visualized. For continuous EMG recording surface, electrodes (Neuroline 720, Ambu GmbH; Bad Nauheim, Germany) were attached in a belly-tendon montage over the contralateral muscles; the abductor pollicis brevis, the first dorsal interosseus (FDI), the extensor digitorum (EXT), the tibial anterior (TA), and abductor hallucis (AH) muscles. Subsequently, the resting motor threshold (RMT) of the FDI and TA were defined, each representing upper and lower limb muscles respectively.19 Thereafter, mapping was conducted at 110% of the according RMT. For mapping upper limb muscle representation areas, the coil was turned perpendicular to the central sulcus, for mapping the leg motor cortex it was oriented in a lateromedial direction.20,21 For each limb representation, at least 100 stimuli were applied, paying attention to delineate the border of the representation area. Stimuli eliciting MEPs after preinnervation as well as MEPs ≤ 50 μV were discarded. Finally, all valid nTMS data of both upper and lower limb mapping were exported in 25 mm peeling depth from the skull in DICOM format for further processing. Manual segmentation of the precentral gyrus For manual segmentation of the precentral gyrus, a 3-D brain mask was created individually for every patient through skull stripping on the T1-weighted MPRAGE sequence in MeVisLab (Version 2.8a, MEVIS Medical Solutions AG, Bremen, Germany; of note that the MeVisLab Software is strictly for noncommercial purposes and not a FDA/CE-certified application). Once the cranial bone had been digitally removed, a 3-D brain model was generated using the 3-D interactive watershed transformation approach. This approach allows for efficient segmentation of the brain, even in the presence of serious pathologies such as malignant brain tumors.22,23 Then, the precentral gyrus was manually segmented for each of the patients’ brain models using sphere-of-influence segmentation.24 For this method, digital markers were set in the precentral gyrus, which were automatically enlarged by a predefined radius of 8 mm, thus creating a converging seeding area (see Figure 2). The generated seeding area was finally cropped against the fractional anisotropy (FA) map in order to include CST fibers only. FIGURE 2. View largeDownload slide Sphere of influence segmentation in MeVisLab (MEVIS Medical Solutions AG). Green dots indicate manually set markers, the yellow seeding ROI was achieved by their concentric enlargement. FIGURE 2. View largeDownload slide Sphere of influence segmentation in MeVisLab (MEVIS Medical Solutions AG). Green dots indicate manually set markers, the yellow seeding ROI was achieved by their concentric enlargement. Time MRI data acquisition, fMRI, and nTMS were performed both before and after neurosurgical tumor resection. Whereas MRI was done during the week before and within 72 h after surgery, fMRI was acquired within 7 d pre- as well as 7 d postoperatively. Preoperative nTMS was always performed the day before surgery, taking advantage of maximum preoperative tumor-associated edema reduction by dexamethasone.25 Likewise, patients were examined by nTMS 5 to 8 d after surgery. Surgical Setup and Intraoperative MRI All patients underwent fluorescence- and iMRI-guided tumor resection in a mobile ultra-low-field strength iMRI-device under general anesthesia (induction through Propofol 200 μg/kg and Fentanyl 250 μg, intubation under Cisatracurium 20 mg, and Midazolam 2 mg, maintenance by total intravenous anesthesia with Propofol 4-8 mg/kg/h and Remifentanil 0.2-0.5 mg/kg/min). iMRI was used for both neuronavigation and resection control. Recent evidence exists that the combined use of iMRI and 5-aminolevulinic acid-derived tumor fluorescence maximizes the EOR.26,27 Six patients of this study were operated on before 2013 and underwent iMRI in a PoleStar N20 scanner (Odin Medical Technologies, Yokneam, Israel/Medtronic Navigation; Medtronic Inc, Dublin, Ireland; field strength 0.15 T). For all following patients, a PoleStar N30 scanner (Medtronic Inc; field strength 0.15 T) was employed. Acquired imaging sequences were selected according to preoperative diagnostic imaging. For contrast-enhancing tumors axial T1-weighted (4-mm slices, scanning time 7 min) sequences with contrast agent (gadolinium-diethylenetriaminepentaacetic acid (DTPA), 0.2 ml per kg bodyweight) were employed. In patients with nonenhancing lesions, T2-weighted (4 mm, 11 min) or fluid-attenuated inversion recovery sequences (6 mm, 9 min) were obtained. All images were automatically loaded into the StealthStationTM neuronavigation system (Medtronic Navigation; Medtronic Inc). Before tumor resection, scanning was performed in the anesthetized patient to obtain data for intraoperative navigation. After tumor resection, a second scan was carried out for extent-of-resection control. If residual tumor was identified at functional boundaries either according to neurophysiological mapping and monitoring or according to anatomy, no further resection was undertaken. In case of detected residual tumor safely accessible to further resection, surgery continued, followed by another control scan. After the last scan, subcortical stimulation was repeated inside the resection cavity. Via an optical tracking device (Sure TrakTM, Medtronic Navigation; Medtronic Inc) mounted on the stimulation probe each positive stimulation site could be registered on the iMRI sequence. Additionally, the stimulation intensity eliciting MEPs at the corresponding site was recorded. Intraoperative Monitoring and Mapping Intraoperative neurophysiological monitoring of somatosensory potentials and MEPs evoked by transcranial electric stimulation has been performed in all patients throughout the entire surgical procedure. Its setup in the iMRI has been described previously.28,29 Using a multipulse stimulation technique (5 pulses, 0.5 ms pulse width each, interstimulus interval 0.4 ms),30 MEPs were recorded from the contralateral thenar, EXT, biceps, and TA muscles. Moreover, responses of hypothenar, FDI, biceps, quadriceps, long AH, orbicularis oris, and genioglossus muscles were registered according to the localization of the tumor. In addition, DCS and DSS were applied (probe tip 1.5 mm; 5 pulses, 0.5 ms pulse width each, interstimulus interval 0.4 ms), aiming at identification and localization of the CST.31 Stimulation polarity was anodal for cortical mapping and cathodal for subcortical mapping.32 The reference electrode was located at Fzp. Starting at a stimulation intensity of 5 mA, it was increased in steps of 1 mA up to a maximum stimulation intensity of 20 mA until MEPs with a minimum amplitude of 30 μV within 4 consecutive trains at a 0.5-Hz repetition rate were recorded.31,33 Stimulation conditions were the same in all patients. Thus, the stimulation charge (C = mA*ms) was only influenced by stimulation intensity, but not pulse duration. Stimulation sites were recorded on iMRI after tumor removal together with the applied currents. For all recordings, the ISIS IOM Neuromonitoring system (Inomed Medizintechnik GmbH, Emmendingen, Germany) was used. Image Registration In a first step, pre- and postsurgically acquired MRI data sets were imported into the commercial planning software StealthVizTM (Medtronic Inc), where ADC maps and diffusion tensors were calculated and visualized in direction-encoded color (DEC) maps. Then, DEC maps and DTI data were converted into DICOM format and loaded into the FT software MeVisLab (Version 2.8a, MEVIS Medical Solutions AG). For each patient, all ROIs generated by nTMS, fMRI, and by manual segmentation of the precentral gyrus were equally imported into the software MeVisLab. According to the acquisition time, data were registered separately for pre- and postsurgical images onto the corresponding T1 MPRAGE sequence using a rigid image transformation of MeVis Image Registration Toolkit (MERIT),34 a software tool implemented in MeVisLab. The same rigid image transformation was also used for the transformation of pre- and postsurgically acquired DWI sequences into the associated T1 MPRAGE coordinate space, necessary as a prerequisite for DTI tractography. Finally, iMRI data and the DCS/DSS stimulation sites were registered on the same individual T1 MPRAGE sequence using MERIT. For later calculation of the distances between the distinctive fiber tracts and DSS stimulation sites, a common coordinate space was defined, using the pre- and postsurgical T1 MPRAGE coordinate space accordingly. Enlargement of nTMS Seeding Areas Since nTMS results are displayed as nonconfluent dots when exported, many examiners enlarge nTMS sites by postprocessing of source data.35–38 We therefore additionally aimed at comparing tractography results using ROIs according to both original and enlarged nTMS data. For this purpose, nTMS data were imported in MeVisLab (MEVIS Medical Solutions AG) and enlarged by both 1 and 2 mm employing an Euclidean distance transformation using MeVisLab. Registration of White Matter Atlas The probabilistic FT algorithm used in this study is based on regionally dependent parameter adaption as the fiber passes different areas of the brain. Therefore, areas with different diffusion and tissue properties need to be defined in advance by using a white-matter atlas. In this study, we used the JHU-MNI-ss atlas, also known as “Eve Atlas,” based on T1 sequences of 152 healthy volunteers.39 This atlas parceled out 176 subregions of the brain using the ICBM-DTI-81 atlas.40,41 The Eve Atlas was then registered to the T1 MPRAGE coordinate space by a linear affine image transformation. The image similarity metric used for all image registration procedures was the intensity-based sum of squared differences metric in MERIT.42 FT Algorithm Tractography was performed in MeVisLab (MEVIS Medical Solutions AG) using the software module NeuroQLab.41 The module has been modified for this study so that it can handle the imported seeding ROIs. Deterministic FT Against the background that deterministic FT algorithms use predefined tracking parameters over all brain areas, we set these deterministic FT parameters for all patients as follows, details have been published by Schlüter et al37: voxel size seed grid 1 mm tracking step length 1 mm window length 3 mm minimal FA threshold 0.05 maximum fiber curvature 0.3 voxel size seed grid 1 mm tracking step length 1 mm window length 3 mm minimal FA threshold 0.05 maximum fiber curvature 0.3 voxel size seed grid 1 mm tracking step length 1 mm window length 3 mm minimal FA threshold 0.05 maximum fiber curvature 0.3 voxel size seed grid 1 mm tracking step length 1 mm window length 3 mm minimal FA threshold 0.05 maximum fiber curvature 0.3 Termination criteria (minimal FA, maximum fiber curvature) were calculated on average over a window length of 3 mm.43 FT was terminated when average FA values were <0.05, indicating undirected diffusion. Average fiber curvature over the predefined window length of > 0.3 equally led to termination of the FT. For deterministic FT, a single include ROI was manually defined in MeVisLab to secure CST delineation. The ROI was placed at the level of the posterior limb of the internal capsule.44,45 Nonplausible fibers in the wrong hemisphere or, on postsurgically acquired data, inside the resection cavity were manually excluded. Probabilistic FT In most papers on probabilistic FT, the modeling of the underlying diffusion process is performed in a probabilistic way. However, we utilized an alternative approach41 which uses only diffusion tensors and a deflection-based FT algorithm46 as basis. The underlying idea is that different diffusion processes and geometric properties of fibers require an algorithm whose parameters dynamically adapt to specific brain regions and do not make any assumptions on a global assignment of parameters. This algorithm randomly sets the following parameters within the following predefined valid intervals: voxel size seed grid 1 mm α(streamline (α = 1) vs ‘deflection-based’ (α = 0) FT) ∈ [0;1] tracking step length ∈ [0.5;7] mm window length ∈ [1.5;6] mm minimal FA threshold ∈ [0.1;0.45] maximum fiber curvature ∈ [0.2;0.65] voxel size seed grid 1 mm α(streamline (α = 1) vs ‘deflection-based’ (α = 0) FT) ∈ [0;1] tracking step length ∈ [0.5;7] mm window length ∈ [1.5;6] mm minimal FA threshold ∈ [0.1;0.45] maximum fiber curvature ∈ [0.2;0.65] voxel size seed grid 1 mm α(streamline (α = 1) vs ‘deflection-based’ (α = 0) FT) ∈ [0;1] tracking step length ∈ [0.5;7] mm window length ∈ [1.5;6] mm minimal FA threshold ∈ [0.1;0.45] maximum fiber curvature ∈ [0.2;0.65] voxel size seed grid 1 mm α(streamline (α = 1) vs ‘deflection-based’ (α = 0) FT) ∈ [0;1] tracking step length ∈ [0.5;7] mm window length ∈ [1.5;6] mm minimal FA threshold ∈ [0.1;0.45] maximum fiber curvature ∈ [0.2;0.65] with ∈ meaning element out of this range. Parameter intervals were further restricted for each white matter atlas region securing local parameter adaption. Hereunto parameter intervals for atlas regions were defined as subsets to the valid parameter intervals defined above. Concerning termination criteria, tracking terminated at FA values or a maximum fiber curvature outside the defined parameter interval for the brain region currently calculated. For probabilistic FT calculation 200 iterations were used, with exception of probabilistic FT based on enlarged nTMS ROIs. This was processed using 50 iterations due to limited computer processing capacity. Furthermore, for probabilistic FT one additional ROI was manually placed in the ventrolateral cerebral peduncle in order to exclude false-positive fibers descending from the internal capsule47 (Figure 3). In total, 1 seeding ROI and 2 include ROIs were used for probabilistic FT. FIGURE 3. View largeDownload slide ROI placement at the level of the ventrolateral cerebral peduncle for probabilistic FT, that is shown in the right figure. FIGURE 3. View largeDownload slide ROI placement at the level of the ventrolateral cerebral peduncle for probabilistic FT, that is shown in the right figure. The deterministic and probabilistic 3-D CST fiber tracts were finally stored for later distance calculation. Calculation of Distances Minimal distances between the distinctive tractography and DSS stimulation sites were calculated for all processed fiber tracts. For that purpose, binary masks of both each fiber tract and each stimulation point were generated, and then subsequently the shortest distance in millimeters between these 2 binary masks was calculated. All calculated distances were stored in table format for later statistical analysis. Statistical Analysis A 2-stage statistical analysis was conducted. First, an F-test securing a significant effect of a variable on FT accuracy was run. Calculated P-values indicate if the discovered coherences are significantly differentiable from 0. This was set as selection criteria for variables to be further analyzed. In a second step, a linear mixed effects model was applied to the variables selected by the F-test. The linear mixed effects model differentiates fixed and random effects on correlation between distances (mm) and stimulation intensities (mA). While the distance between DSS stimulation points and CST fiber tracts was defined as independent variable, stimulation intensity of DSS at the respective stimulation sites was set as dependent variable. Fixed-effects represent direct effects of the independent variable on the dependent variable, thus being a measure for direct correlation defining fiber tract accuracy. FT accuracy across the factors SEEDING, ANATOMY, FT ALGORITHM, and TIME was assessed by comparing R2 marginal, the coefficient of determination representing fixed-effects. Random effects subsume individual differences between patients such as gender, age, and tumor size among others. High R² conditional values represent those patient-specific factors on correlation. For this purpose, data were structured as single-level grouped data using the grouping factor “patient.”48 A Shapiro-Wilk test for Gaussian distribution of residual values from linear regression analysis (LME model) confirmed normal distribution of underlying data. Statistics were calculated using the software R (Version 3.2.0 GUI 1.65 Mavericks build, The R Foundation for Statistical Computing, Vienna, Austria) via the user interface RStudio (Version 0.98.1103, RStudio Inc, Boston, Massachusetts). The nlme (linear and nonlinear mixed-effect model) package was applied within R. The level of significance was set to α = 0.05, thus statistical significance was considered to be P < .05. Results of the LME models were sorted by R2 marginal as the determinant figure for FT accuracy in a descending manner, presenting best results at the top. RESULTS Stimulation Findings In total, MEPs could be evoked at 137 DSS sites after tumor resection. Mean stimulation intensity threshold was 13.06 mA (range 3-20 mA; 87 successful stimulations) for upper and 12.22 mA (range 3-20 mA; 50 successful stimulations) for lower limb muscles. Including all sites where MEPs were evoked at a stimulation intensity ≤20 mA, the F-test did not find a significant influence of the parameters TIME and SEEDING ROIs on the correlation between distances and stimulation intensities. Given that higher intensities of monopolar stimulation activate more axons, but over an unspecific area, it was assumed that including only data obtained at lower stimulation intensities would result in higher accuracy. Thus, the filter was set accordingly, including only data received at a stimulation intensity ≤ 12 mA. Consecutive results indicated that all selected variables had a significant influence on FT accuracy (see Table 2). TABLE 2. F-Test With Filtered Data (Stimulation Intensity ≤ 12 mA) Variable P-value TIME .010* SEEDING ROI .018* ANATOMY .001* FT ALGORITHM .004* Variable P-value TIME .010* SEEDING ROI .018* ANATOMY .001* FT ALGORITHM .004* * P < .05, significant correlation. View Large TABLE 2. F-Test With Filtered Data (Stimulation Intensity ≤ 12 mA) Variable P-value TIME .010* SEEDING ROI .018* ANATOMY .001* FT ALGORITHM .004* Variable P-value TIME .010* SEEDING ROI .018* ANATOMY .001* FT ALGORITHM .004* * P < .05, significant correlation. View Large Consequently, all subsequent statistical analyses were conducted using the filtered data at stimulation intensities ≤12 mA. Time The analysis of the factor TIME clearly showed that FT based on preoperative image data delivers superior accuracy compared to postoperatively acquired data sets (see Table 3). This holds true independently of the seeding method applied. TABLE 3. Impact of TIME on FT Accuracy TIME R2 marginal R2 conditional P-value M1 manual pre 0.227 0.574 <.001* M1 manual post 0.164 0.558 .001* fMRI pre 0.131 0.643 .001* nTMS (0 mm) pre 0.107 0.703 .003* fMRI post 0.099 0.689 .020* nTMS (0 mm) post 0.045 0.769 .050* TIME R2 marginal R2 conditional P-value M1 manual pre 0.227 0.574 <.001* M1 manual post 0.164 0.558 .001* fMRI pre 0.131 0.643 .001* nTMS (0 mm) pre 0.107 0.703 .003* fMRI post 0.099 0.689 .020* nTMS (0 mm) post 0.045 0.769 .050* * P < .05, significant correlation. View Large TABLE 3. Impact of TIME on FT Accuracy TIME R2 marginal R2 conditional P-value M1 manual pre 0.227 0.574 <.001* M1 manual post 0.164 0.558 .001* fMRI pre 0.131 0.643 .001* nTMS (0 mm) pre 0.107 0.703 .003* fMRI post 0.099 0.689 .020* nTMS (0 mm) post 0.045 0.769 .050* TIME R2 marginal R2 conditional P-value M1 manual pre 0.227 0.574 <.001* M1 manual post 0.164 0.558 .001* fMRI pre 0.131 0.643 .001* nTMS (0 mm) pre 0.107 0.703 .003* fMRI post 0.099 0.689 .020* nTMS (0 mm) post 0.045 0.769 .050* * P < .05, significant correlation. View Large Seeding Best FT accuracy was achieved by using the segmented precentral gyrus as seeding ROI (marginal R2 = 0.146, P < .001; see Table 4). Since the marginal R2 of fMRI and nTMS differed very little, none of these methods showed distinct superiority over the other. Nevertheless, current-distance correlation improved after enlarging nTMS results by both, 1 and 2 mm (see Figure 4). While the use of original positive nTMS sites led to a marginal R2 = 0.058, enlargement by 1 and 2 mm resulted in marginal R2 = 0.079 and marginal R2 = 0.087, respectively. Thus, with increasing enlargement of positive nTMS sites better correlation was achieved (see Table 5). TABLE 4. Impact of SEEDING ROI on FT Accuracy SEEDING ROI R2 marginal R2 conditional P-value M1 manual 0.146 0.642 <.001* fMRI 0.077 0.611 .006* nTMS (0 mm) 0.058 0.721 .006* SEEDING ROI R2 marginal R2 conditional P-value M1 manual 0.146 0.642 <.001* fMRI 0.077 0.611 .006* nTMS (0 mm) 0.058 0.721 .006* * P < 0.05, significant correlation. View Large TABLE 4. Impact of SEEDING ROI on FT Accuracy SEEDING ROI R2 marginal R2 conditional P-value M1 manual 0.146 0.642 <.001* fMRI 0.077 0.611 .006* nTMS (0 mm) 0.058 0.721 .006* SEEDING ROI R2 marginal R2 conditional P-value M1 manual 0.146 0.642 <.001* fMRI 0.077 0.611 .006* nTMS (0 mm) 0.058 0.721 .006* * P < 0.05, significant correlation. View Large FIGURE 4. View largeDownload slide Fiber tracks based on original nTMS data and after enlargement of nTMS data by 1 and 2 mm (blue, arm fibers; yellow, leg fibers). FIGURE 4. View largeDownload slide Fiber tracks based on original nTMS data and after enlargement of nTMS data by 1 and 2 mm (blue, arm fibers; yellow, leg fibers). TABLE 5. Impact of nTMS Enlargement on FT Accuracy, Without Differentiation of Algorithms as well as Using Either the Deterministic or Probabilistic Approach R2 R2 marginal conditional P-value nTMS enlargement nTMS 2 mm 0.087 0.541 <.001* nTMS 1 mm 0.079 0.575 .001* nTMS 0 mm 0.058 0.721 .006* nTMS enlargement, deterministic FT nTMS 1 mm det 0.095 0.795 .009* nTMS 2 mm det 0.093 0.851 .003* nTMS 0 mm det 0.055 0.757 .037 nTMS enlargement, probabilistic FT nTMS 0 mm prob 0.088 0.750 .001* nTMS 1 mm prob 0.064 0.604 .006* nTMS 2 mm prob 0.056 0.546 .011* R2 R2 marginal conditional P-value nTMS enlargement nTMS 2 mm 0.087 0.541 <.001* nTMS 1 mm 0.079 0.575 .001* nTMS 0 mm 0.058 0.721 .006* nTMS enlargement, deterministic FT nTMS 1 mm det 0.095 0.795 .009* nTMS 2 mm det 0.093 0.851 .003* nTMS 0 mm det 0.055 0.757 .037 nTMS enlargement, probabilistic FT nTMS 0 mm prob 0.088 0.750 .001* nTMS 1 mm prob 0.064 0.604 .006* nTMS 2 mm prob 0.056 0.546 .011* * P < .05, significant correlation. View Large TABLE 5. Impact of nTMS Enlargement on FT Accuracy, Without Differentiation of Algorithms as well as Using Either the Deterministic or Probabilistic Approach R2 R2 marginal conditional P-value nTMS enlargement nTMS 2 mm 0.087 0.541 <.001* nTMS 1 mm 0.079 0.575 .001* nTMS 0 mm 0.058 0.721 .006* nTMS enlargement, deterministic FT nTMS 1 mm det 0.095 0.795 .009* nTMS 2 mm det 0.093 0.851 .003* nTMS 0 mm det 0.055 0.757 .037 nTMS enlargement, probabilistic FT nTMS 0 mm prob 0.088 0.750 .001* nTMS 1 mm prob 0.064 0.604 .006* nTMS 2 mm prob 0.056 0.546 .011* R2 R2 marginal conditional P-value nTMS enlargement nTMS 2 mm 0.087 0.541 <.001* nTMS 1 mm 0.079 0.575 .001* nTMS 0 mm 0.058 0.721 .006* nTMS enlargement, deterministic FT nTMS 1 mm det 0.095 0.795 .009* nTMS 2 mm det 0.093 0.851 .003* nTMS 0 mm det 0.055 0.757 .037 nTMS enlargement, probabilistic FT nTMS 0 mm prob 0.088 0.750 .001* nTMS 1 mm prob 0.064 0.604 .006* nTMS 2 mm prob 0.056 0.546 .011* * P < .05, significant correlation. View Large The same was true when taking the factor TIME into account. In a subgroup analysis, the correlation for both, pre- and postoperatively acquired image data for FT, was improved applying nTMS enlargement. However, no additional positive effect was observed after enlarging nTMS data by 2 mm compared to 1 mm. Combining the enlargement of positive nTMS sites with the factor FT ALGORITHM, better correlation was obtained by the deterministic approach. However, using the deterministic algorithm enlarging nTMS sites by more than 1 mm did not further improve correlation. Applying the probabilistic algorithm correlation worsened with increasing enlargement (see Table 5). Anatomy Studying whether fiber tracts of the upper or lower limb were depicted more accurately was only possible using ROIs based on nTMS or fMRI data. Both approaches revealed a significant correlation between distances and stimulation intensities for fiber tracts representing the CST of the arm, whereas no such correlation could be found for leg fibers (see Table 6). TABLE 6. Impact of ANATOMY on FT Accuracy ANATOMY R2 marginal R2 conditional P-value fMRI arm 0.091 0.625 .010* nTMS (0 mm) arm 0.089 0.598 .007* fMRI leg 0.019 0.667 .098 nTMS (0 mm) leg 0.003 0.826 .636 ANATOMY R2 marginal R2 conditional P-value fMRI arm 0.091 0.625 .010* nTMS (0 mm) arm 0.089 0.598 .007* fMRI leg 0.019 0.667 .098 nTMS (0 mm) leg 0.003 0.826 .636 * P < .05, significant correlation. View Large TABLE 6. Impact of ANATOMY on FT Accuracy ANATOMY R2 marginal R2 conditional P-value fMRI arm 0.091 0.625 .010* nTMS (0 mm) arm 0.089 0.598 .007* fMRI leg 0.019 0.667 .098 nTMS (0 mm) leg 0.003 0.826 .636 ANATOMY R2 marginal R2 conditional P-value fMRI arm 0.091 0.625 .010* nTMS (0 mm) arm 0.089 0.598 .007* fMRI leg 0.019 0.667 .098 nTMS (0 mm) leg 0.003 0.826 .636 * P < .05, significant correlation. View Large FT Algorithm Regardless of the employed seeding ROIs and not considering nTMS enlargement, the use of the probabilistic FT algorithm led to a better correlation than the deterministic FT algorithm (see Table 7 and Figure 5). TABLE 7. Impact of FT ALGORITHM on FT Accuracy FT ALGORITHM R2 marginal R2 conditional P-value M1 manual prob 0.303 0.657 .007* M1 manual det 0.184 0.751 .005* fMRI prob 0.152 0.690 <.001* fMRI det 0.099 0.555 .014* nTMS prob, 0 mm 0.088 0.750 .001* nTMS det, 0 mm 0.055 0.757 .037* FT ALGORITHM R2 marginal R2 conditional P-value M1 manual prob 0.303 0.657 .007* M1 manual det 0.184 0.751 .005* fMRI prob 0.152 0.690 <.001* fMRI det 0.099 0.555 .014* nTMS prob, 0 mm 0.088 0.750 .001* nTMS det, 0 mm 0.055 0.757 .037* * P < .05, significant correlation. View Large TABLE 7. Impact of FT ALGORITHM on FT Accuracy FT ALGORITHM R2 marginal R2 conditional P-value M1 manual prob 0.303 0.657 .007* M1 manual det 0.184 0.751 .005* fMRI prob 0.152 0.690 <.001* fMRI det 0.099 0.555 .014* nTMS prob, 0 mm 0.088 0.750 .001* nTMS det, 0 mm 0.055 0.757 .037* FT ALGORITHM R2 marginal R2 conditional P-value M1 manual prob 0.303 0.657 .007* M1 manual det 0.184 0.751 .005* fMRI prob 0.152 0.690 <.001* fMRI det 0.099 0.555 .014* nTMS prob, 0 mm 0.088 0.750 .001* nTMS det, 0 mm 0.055 0.757 .037* * P < .05, significant correlation. View Large Figure 5. View largeDownload slide Deterministic vs probabilistic FT results visualized across seeding methods (blue, arm fibers; yellow, leg fibers; green, all CST fibers). Figure 5. View largeDownload slide Deterministic vs probabilistic FT results visualized across seeding methods (blue, arm fibers; yellow, leg fibers; green, all CST fibers). DISCUSSION Correlation of FT with Subcortical Mapping Since it is impossible to expose the pyramidal tract for correlation with tractography, the lowest stimulation intensity evoking MEPs was correlated to the shortest distance between the stimulation site and the respective visualized fiber tract, an approach that is widely accepted.10,49–51 Hitherto, a nonlinear current-distance correlation has been described by some authors,32,52,53 whereas in 6 studies a linear current-distance correlation has been found, but differing among studies in both linear and absolute coefficients.51,54–58 Applying different stimulation conditions in a very systematic approach, Shiban et al32 described a nonlinear current-distance correlation for cathodal subcortical stimulation with a pulse-duration of 0.5 or 0.7 ms. They pointed out that this correlation was near the simple rule that per mm distance to the pyramidal tract, 1 mA of subcortical stimulation intensity would be necessary to elicit MEPs, with the exception to this rule that in the direct vicinity of the CST, lower intensities might suffice. Ohue et al56 observed a strong correlation between bipolar subcortical stimulation and the distance to postoperative tractography in 32 patients. As they stated, intraoperative brain shift negatively influenced the use of preoperative FT for distance correlation. During the last years, this problem has increasingly been counteracted by the use of intraoperative imaging. Nossek et al55 performed monopolar subcortical stimulation and applied intraoperative ultra-sonography for updating preoperative MR data, resulting in a relationship of 0.97 mA for every 1 mm distance to the CST. Maesawa et al54 employed bipolar subcortical stimulation and found a linear current-distance correlation for intraoperative, but not preoperative tractography. However, Ostry et al58 found a linear correlation between the threshold of monopolar cathodal subcortical stimulation and the distance to both, pre- and intraoperative tractography, observing a stronger correlation with preoperative imaging. It is noteworthy, that not only study designs, but also stimulation conditions differed among studies, rendering comparison of results impossible. Moreover, considering that numbers of positive subcortical stimulation sites varied among studies, statistical power of their correlation was of varying strength. While Ostry et al58 registered 155 and 125 stimulation sites for pre- and intraoperative correlation calculation, respectively, Nossek et al55 collected 43 and Maesawa et al54 28 stimulation recordings. Only recently, Javadi et al51 reported on a series of 20 patients who underwent iMRI including intraoperative DTI tractography and DSS at the end of tumor resection. Subcortical stimulation was bipolar in 12 patients and monopolar in 8 patients, yielding 18 positive stimulations in total and resulting in a linear current-distance dependence for bipolar, but not for monopolar stimulation. In addition, no MEPs could be evoked by bipolar DSS in 3 patients, despite the close spatial relationship of ≤2 mm between stimulation sites and fiber tracks, according to the neuronavigation system. Summing up all these studies, we have to conclude, that results of different current-distance correlations need individual and cautious interpretation. In the present study, we employed monopolar cathodal subcortical stimulation in 11 patients, evoking 87 MEPs of upper limb muscles and 50 MEPs of lower limb muscles after tumor removal. Against the experience of other authors, we did not encounter any technical difficulties or seizures during monopolar cortical or subcortical stimulation. Consequently, we used this method consistently to all patients without using electrocorticogramm (ECoG) recording. Stimulation sites where MEPs with a minimum amplitude of 30 μV could be elicited were immediately recorded as 3-D-coordinates in the iMRI data set, and these data were coregistered with pre- and postoperative tractography results. In contrast to previous studies, which showed linear correlations between stimulation currents and distances presuming that all effects measured are direct effects of X on Y,54–56 we applied a linear mixed effects model taking direct effects (distance on stimulation intensity) and random effects into account. Thus, the results of our study are not comparable to previous ones. Taking the variables ROI selection, ANATOMY, FT ALGORITHM, and TIME into account, best correlations between respective fiber tracks and stimulation intensities were observed for fiber tracks using manual segmentation of the precentral gyrus as seeding ROIs and being generated by a probabilistic FT algorithm preoperatively. Thus, the prior hypothesis that preoperatively acquired DTI data and a probabilistic FT algorithm would deliver most accurate CST tracking results was confirmed. Interestingly, our expectation that nTMS results would be most appropriate for ROI selection was not fulfilled completely. The Effect of ROI Selection The selection of seeding ROIs for delineating the CST differs among scientists. Thus, Radmanesh et al59 compared different seeding methods, observing best results when selecting fibers passing through both, fMRI activation areas and the posterior limb of the internal capsule, in contrast to seeding from a single ROI. Analyzing fiber tracts obtained after ROI definition either according to nTMS results or according to anatomic motor cortex definition, the integration of nTMS data yielded more accurate tracking results, both, in tumor patients as well as in a tumor-free cohort undergoing deep brain stimulation.57,60 In the present study, manual segmentation of the precentral gyrus delivered slightly more accurate results than did seeding according to nTMS or fMRI data, reflected in a better, though low correlation. While there was hardly a difference in accuracy of FT based on nTMS and fMRI, former studies reported on a superiority of nTMS over fMRI when correlated to direct motor cortex stimulation.19,35 However, this comparison lacked methodological preciseness, since fMRI activation areas are diffuse regions of increased cerebral metabolism and are therefore not comparable to focal nTMS stimulation sites. While fMRI mostly reflects a number of cortical muscle representation areas that are involved in planning and executing a specific task, nTMS allows activation of pyramidal axons of individual muscles, resulting in smaller representational maps. As a consequence, several authors either outlined the area of nTMS data57,61 or concentrically enlarged nTMS results up to 6 mm36–38,62 with the intention to create confluent nTMS seeding ROIs. Although these approaches yielded fiber tracks that were more plausible37,62 and better overlapping the motor cortex,57 such artificial ROI enlargement might have biased study results when comparing FT based on nTMS seeding ROIs to other seeding methods. Based on our results and considering deterministic FT that still is the only algorithm implemented in neuronavigation systems, a concentric enlargement of 1 mm of nTMS data already increases FT accuracy compared to nTMS raw data, regardless of other variables such as TIME and ANATOMY. However, a further concentric augmentation by 2 mm of nTMS data does not lead to a significant additional accuracy improvement, indicating a possible ceiling effect. Nevertheless, FT using the segmented precentral gyrus as seeding ROI was more accurate than tractography using ROIs based on fMRI or nTMS results. A possible explanation for the inferiority of fMRI may be found in its constraints by hemodynamic changes caused by adjacent malignant tumors or by tumorous cortex infiltration.63,64 Besides, the clinical routine of manually adjusting fMRI data sets for rotational and translational alignment may have influenced the position and volume of fMRI activation areas, potentially influencing accuracy of tracts based on fMRI data. On the other side, fully automated alignment of different data sets involves the risk of fatal misregistration. Therefore, we first performed automated rotational and translational alignment which we then controlled for correctness and which we adjusted if necessary. Moreover, the BOLD signal is 3 to 7 times lower in the white matter,65–67 so that fMRI is mainly displayed in the grey matter. Since anisotropy in the cortex is strongly reduced, FT in the gray matter is generally less accurate and more prone to errors. This problem equally applies for nTMS. Although nTMS data can be displayed on the MPRAGE sequence at various depths from the head surface, nTMS results displayed in the white matter (or even projected into sulci below a thin winding gyrus) inevitably need to be only virtual and thus inaccurate, since TMS induces action potentials in cortical neurons. Moreover, an additional inaccuracy is encountered when defining the leg representation area since it is mainly located in the interhemispheric fissure. As a consequence, fiber tracts originating from the leg representation area defined by nTMS, but also by fMRI, are less plausible than arm-related fiber tracts, as could be shown recently by Weiss Lucas et al.62 When using nTMS data for further processing, moreover, its system-inherent errors should be assumed coming up to 5 mm as shown by Ruohonen and Karhu.68 Although enlargement of nTMS data could compensate this error, this enlargement might equally further increase the error; enlarging nTMS sites concentrically by 2 or 5 mm could eventually result in a combined error of √(52 + 22) = 5.38 mm or √(52 + 52) = 7.07 mm. Among these errors, the registration mismatch of nTMS data and the MPRAGE sequence is 2.5 mm, which might equally be found when fusing nTMS with the DWI sequence. Thus, nTMS data might not only be displayed in the motor area, but also in areas adjacent to them. In light of the fact that manual segmentation of the precentral gyrus yielded best FT accuracy and that nTMS allows for reliable functional mapping,19,69,70 filtering of the manually segmented motor area according to nTMS results might therefore be considered for achieving most accurate FT results. FT Algorithms in Brain Tumor Patients DTI tractography in brain tumor patients harbors the risk of inaccuracy due to histopathological white matter changes, caused by invading tumor cells and their interaction with the surrounding tissue, resulting in displaced, infiltrated, disrupted, or edematous fiber tracks.71 As a consequence, FA values have been correlated with tumor pathology and tumor invasion. While Stadlbauer et al72 reported on a decrease in FA values with the higher tumor cell infiltration determined histologically in grade II and grade III glioma patients, Beppu et al73 found a positive correlation between FA values and the cell density/MIB-1 proliferation index in patients suffering from glioblastoma. Mormina et al74 analyzed the effect of tumor infiltration on DTI parameters comparing healthy and tumor-affected hemispheres, and observed a decrease of FA and diffusion coefficients in close proximity to the tumor along with an increase in spherical diffusion. Additionally, Stadlbauer et al75 noticed that increasing FA thresholds in the presettings of DTI tractography went along with a reduction in number and length of fiber bundles, simultaneously increasing the distance of these fiber bundles to pathological signal changes on FA maps. As a consequence, a maximization of FT accuracy in proximity to invasive brain tumors demands for parameter adaptation. In order to differentiate malignant from healthy brain tissue, Kinoshita et al76 proposed an FA cut-off value of 0.26. With the aim to even render fiber tracts that have been invaded by tumor cells, an FA threshold between 0.15 and 0.2 was proposed.10,75 In the present study, all tumors were located both subcortically and cortically, with exception of 2 purely subcortically growing high-grade gliomas. On the cortical level, 5 of these tumors were in very close vicinity of the motor cortex (4 in postcentral and 1 in precentral location), but only one out of 11 tumors infiltrated the precentral gyrus. Thus, a potential confound based on infiltration of the motor cortex was unlikely. Taking all those findings into account, our deterministic algorithm used an FA threshold of 0.05, significantly lower compared to previous studies, with the aim to include more fibers in proximity to the tumor. In order to avoid a large number of misleading fibers not belonging to the CST, maximum fiber curvature was set more restrictively to 30°. In addition, the inferiority of postoperative tractography results could possibly be explained with the acquisition time of the underlying MRI, within 72 h after surgery. This time frame coincides with postoperative tumor edema decreasing FA values, thus negatively affecting tractography accuracy. A comparison of probabilistic vs deterministic DTI tractography demands for a strong and objective reference standard. Thus, Bucci et al49 delineated the CST with DTI and q-ball models of diffusion using deterministic and probabilistic FT algorithms, relating the resulting fiber tracks to DCS/DSS sites. Probabilistic algorithms showed to be significantly more sensitive in delineating the CST compared to deterministic approaches, with the q-ball model being significantly more sensitive than the DTI method. A similar approach was chosen by Mandelli et al50 relating probabilistic and deterministic DTI FT for delineating the CST on preoperatively acquired DTI to DCS/DSS sites. They observed significantly shorter distances between DCS/DSS sites to tracks delineated by the probabilistic algorithm compared to the deterministic method, but lacked a correlation of distances to the electric current applied. Moreover, both algorithms presented poor sensitivity to delineate the lateral parts of the CST, although the use of the probabilistic method improved this sensitivity significantly. In order to overcome this limitation, Mormina et al74 proposed a constrained spherical deconvolution algorithm that delivered more accurate tracking results of the lateral CST. Our study advanced these approaches in correlating the distances between fiber tracts and stimulation sites with the stimulation current, taking into consideration the dependence of neural activation on stimulation intensity. According to our results, FT accuracy was higher using the probabilistic parameter adoption algorithm first presented by Klein et al in 2015,46 independently of all other influencing factors (ROI, TIME, ANATOMY) analyzed. Our results are consistent with the results of former studies49,50,77,78 and underline the superiority of probabilistic tractography over deterministic methods in tumor-affected brains. Standardized preset tracking parameters of deterministic techniques barely match the highly variable tissue characteristics of tumor-infiltrated brain tissue. Obviously, this explains the advantage of atlas-based iterative parameter adoption by probabilistic FT.46,49,50 This superiority in accuracy of probabilistic FT should influence the selection of FT algorithms even in a clinical setting facing limited time and resources in order to maximize patient security. The probabilistic algorithm applied in this study calculated 200 iterations in 10 to 20 s processing time (vs 0,05-0,1 s for det. FT) at a total tracking time (including image registration, ROI setting, and segmentation) of approximately 30 min.24 Limitations and Future Prospects Intraoperative DWI or DTI data were not available for intraoperative tractography with the low-field iMRI scanner applied. All data analyzed were acquired both pre- and postoperatively and registered onto intraoperative structural MRI data. As a consequence, tractography based on preoperatively acquired DTI and referenced to postresection DSS was subject to systematic brain shift error caused by the resection itself. Therefore, further studies are needed for reassessing our findings using a modern high-field iMRI scanner producing intraoperative DTI data. Due to technical limitations in the neurosurgical operation room, only rigid image registration has been accessible for merging perioperatively acquired sequences (fMRI, nTMS) and intraoperative MRI sequences with the T1 MPRAGE sequence so far. Nonrigid image registration algorithms recently proposed by several authors are more flexible and better compensate for image deformations such as brain shift,79–81 but require high-performance computing capacities next to the operation room. In 2007, Archip et al82 presented a nonrigid registration algorithm, which overcomes these technical limitations, running on a workstation with 4 processors in less than 40 min. This nonrigid registration algorithm increased registration precision by factor 4.58 compared to the standard of rigid image registration (P < .001). Ongoing improvements in computing capacities and parallel processing will allow applying nonrigid image registration for future research purposes as well as routine medical procedures. Moreover, the well-known problem of fiber reconstruction in brain areas containing fiber crossings was not specifically addressed in this study.83 Several solutions such as diffusion spectrum magnetic resonance imaging, connectivity maps, or Spatial HARDI exist and could have further improved FT accuracy84–88 In addition, this study used diffusion-weighted image sequences with 20 diffusion directions. A higher number of DTI gradient directions would have probably increased tractography quality characterized by increased tract volume, median fiber density, and mean FA.89 Although tumor pathology differed among patients, this difference should not have affected results in this study, given that a minimal FA threshold of 0.05 for deterministic FT and an FA of 0.10 to 0.45 for probabilistic FT has been used. Previous studies reported on significantly higher FA in peritumoral edema of high-grade glioma than in edema surrounding metastasis, but FA values in peritumoral edema are >0.12 for both types of brain lesion.90,91 It should be noted that the results of this study are derived from a small cohort of patients (n = 11) operated on for brain tumors adjacent to the CST. Given tumor location and the complexity of necessary complete data sets for every patient (fMRI, nTMS, DTI pre- and postoperatively, iMRI and intraoperative DSS results evoking MEPs in the resection cavity), this number was a notable sample size. Before surgery, most patients suffered from motor deficits resulting in a reduction of both, motor cortex excitability by nTMS and performance of motor tasks for fMRI. Further studies on larger cohorts including patients not operated on for malignant tumors or undergoing stereotactic procedures are required for a systematic evaluation of FT results. CONCLUSION In summary, this study provides evidence that tractography may vary significantly in dependence of different influencing factors. Our findings indicate that best results of CST tractography can be achieved by using the segmented motor cortex as seeding ROI and applying a probabilistic algorithm. It has to be stressed that intraoperative subcortical neurophysiological mapping most reliably locates the CST.10,31,92,93 Although neuroimaging techniques including FT allow for a virtual estimation of functional areas, they should not replace neurophysiological monitoring and mapping. Disclosures This work was financed by a research grant from Medtronic Navigation Inc, Louisville, Colorado. Moreover, this study was supported by a grant of the Neurosurgery Research Foundation of the DGNC to MT Forster. Dr Senft has served as a consultant for Medtronic Navigation. REFERENCES 1. Sanai N , Berger MS . Extent of resection influences outcomes for patients with gliomas . Rev Neurol (Paris) . 2011 ; 167 ( 10 ): 648 – 654 . Google Scholar CrossRef Search ADS PubMed 2. Capelle L , Fontaine D , Mandonnet E et al. Spontaneous and therapeutic prognostic factors in adult hemispheric World Health Organization Grade II gliomas: a series of 1097 cases . J Neurosurg . 2013 ; 118 ( 6 ): 1157 – 1168 . Google Scholar CrossRef Search ADS PubMed 3. Bloch O , Han SJ , Cha S et al. Impact of extent of resection for recurrent glioblastoma on overall survival . J Neurosurg . 2012 ; 117 ( 6 ): 1032 – 1038 . Google Scholar CrossRef Search ADS PubMed 4. Brown TJ , Brennan MC , Li M et al. Association of the extent of resection with survival in glioblastoma . JAMA Oncol . 2016 ; 2 ( 11 ): 1460 – 1469 . Google Scholar CrossRef Search ADS PubMed 5. De Witt Hamer PC , Robles SG , Zwinderman AH , Duffau H , Berger MS . Impact of intraoperative stimulation brain mapping on glioma surgery outcome: a meta-analysis . J Clin Oncol . 2012 ; 30 ( 20 ): 2559 – 2565 . Google Scholar CrossRef Search ADS PubMed 6. Duffau H , Lopes M , Arthuis F et al. Contribution of intraoperative electrical stimulations in surgery of low grade gliomas: a comparative study between two series without (1985-96) and with (1996-2003) functional mapping in the same institution . J Neurol Neurosurg Psychiatry . 2005 ; 76 ( 6 ): 845 – 851 . Google Scholar CrossRef Search ADS PubMed 7. Prabhu SS , Gasco J, Tummala S, Weinberg JS, Rao G. Intraoperative magnetic resonance imaging-guided tractography with integrated monopolar subcortical functional mapping for resection of brain tumors . J Neurosurg . 2011 ; 114 ( 3) : 719 – 726 . Google Scholar CrossRef Search ADS PubMed 8. Senft C , Bink A , Franz K , Vatter H , Gasser T , Seifert V . Intraoperative MRI guidance and extent of resection in glioma surgery: a randomised, controlled trial . Lancet Oncol . 2011 ; 12 ( 11 ): 997 – 1003 . Google Scholar CrossRef Search ADS PubMed 9. Nimsky C , Ganslandt O , Merhof D , Sorensen AG , Fahlbusch R . Intraoperative visualization of the pyramidal tract by diffusion-tensor-imaging-based fiber tracking . Neuroimage . 2006 ; 30 ( 4 ): 1219 – 1229 . Google Scholar CrossRef Search ADS PubMed 10. Bello L , Gambini A , Castellano A et al. Motor and language DTI Fiber Tracking combined with intraoperative subcortical mapping for surgical removal of gliomas . Neuroimage . 2008 ; 39 ( 1 ): 369 – 382 . Google Scholar CrossRef Search ADS PubMed 11. Picht T , Frey D , Thieme S , Kliesch S , Vajkoczy P . Presurgical navigated TMS motor cortex mapping improves outcome in glioblastoma surgery: a controlled observational study . J Neurooncol . 2016 ; 126 ( 3 ): 535 – 543 . Google Scholar CrossRef Search ADS PubMed 12. Castellano A , Bello L , Michelozzi C et al. Role of diffusion tensor magnetic resonance tractography in predicting the extent of resection in glioma surgery . Neuro Oncol . 2012 ; 14 ( 2 ): 192 – 202 . Google Scholar CrossRef Search ADS PubMed 13. Nimsky C , Ganslandt O , Hastreiter P et al. Preoperative and intraoperative diffusion tensor imaging-based fiber tracking in glioma surgery . Neurosurgery . 2005 ; 56 ( 1 ): 130 – 138 ; discussion 138 . Google Scholar CrossRef Search ADS PubMed 14. Shahar T , Rozovski U , Marko NF et al. Preoperative imaging to predict intraoperative changes in tumor-to-corticospinal tract distance . Neurosurgery . 2014 ; 75 ( 1 ): 23 – 30 . Google Scholar CrossRef Search ADS PubMed 15. Feigl GC , Hiergeist W , Fellner C et al. Magnetic resonance imaging diffusion tensor tractography: evaluation of anatomic accuracy of different fiber tracking software packages . World Neurosurg . 2014 ; 81 ( 1 ): 144 – 150 . Google Scholar CrossRef Search ADS PubMed 16. Clark CA , Barrick TR , Murphy MM , Bell BA . White matter fiber tracking in patients with space-occupying lesions of the brain: a new technique for neurosurgical planning? Neuroimage . 2003 ; 20 ( 3 ): 1601 – 1608 . Google Scholar CrossRef Search ADS PubMed 17. Association WM. World Medical Association Declaration of Helsinki . JAMA . 2013 ; 310 ( 20 ): 2191 – 2194 . Google Scholar CrossRef Search ADS PubMed 18. Krishnan R , Raabe A , Hattingen E et al. Functional magnetic resonance imaging-integrated neuronavigation: correlation between lesion-to-motor cortex distance and outcome . Neurosurgery . 2004 ; 55 ( 4 ): 904 – 915 ; discusssion 914-915 . Google Scholar CrossRef Search ADS PubMed 19. Forster MT , Hattingen E , Senft C , Gasser T , Seifert V , Szelényi A . Navigated transcranial magnetic stimulation and functional magnetic resonance imaging: advanced adjuncts in preoperative planning for central region tumors . Neurosurgery . 2011 ; 68 ( 5 ): 1317 – 1325 ; discussion 1324-1325 . Google Scholar CrossRef Search ADS PubMed 20. Mills KR , Nithi KA . Corticomotor threshold to magnetic stimulation: normal values and repeatability . Muscle Nerve . 1997 ; 20 ( 5 ): 570 – 576 . Google Scholar CrossRef Search ADS PubMed 21. Danner N , Julkunen P , Könönen M , Säisänen L , Nurkkala J , Karhu J . Navigated transcranial magnetic stimulation and computed electric field strength reduce stimulator-dependent differences in the motor threshold . J Neurosci Meth . 2008 ; 174 ( 1 ): 116 – 122 . Google Scholar CrossRef Search ADS 22. Hahn HK , Peitgen H-O . The skull stripping problem in MRI solved by a single 3D watershed transform . In: Medical Image Computing and Computer-Assisted Intervention - MICCAI . Vol. 1935 . Berlin : Springer ; 2000 : 129 – 145 . Google Scholar CrossRef Search ADS 23. Hahn HK , Peitgen H-O. Interactive Watershed Transform: A Hierarchical Method for Efficient Interactive and Automated Segmentation of Multidimensional Grayscale Images . In : Sonka M Fitzpatrick JM , (eds) Proceedings of SPIE Medical Imaging . SPIE ; 2003 : 643 – 653 . 24. Klein J , Weiler F , Hahn HK . Probabilistic Parameter Adaptation for Fiber Tracking of the Corticospinal Tract. Proceedings of MICCAI 2014 - DTI Tractography Challenge. Boston , 2014 . 25. Andersen C , Haselgrove JC , Doenstrup S , Astrup J , Gyldensted C . Resorption of peritumoural oedema in cerebral gliomas during dexamethasone treatment evaluated by NMR relaxation time imaging . Acta Neurochir . 1993 ; 122 ( 3-4 ): 218 – 224 . Google Scholar CrossRef Search ADS PubMed 26. Coburger J , Hagel V , Wirtz CR , König R . Surgery for glioblastoma: impact of the combined use of 5-aminolevulinic acid and intraoperative MRI on extent of resection and survival . PLoS One . 2015 ; 10 ( 6 ): e0131872 . Google Scholar CrossRef Search ADS PubMed 27. Gessler F , Forster MT , Duetzmann S et al. Combination of intraoperative magnetic resonance imaging and intraoperative fluorescence to enhance the resection of contrast enhancing gliomas . Neurosurgery . 2015 ; 77 ( 1 ): 16 – 22 . Google Scholar CrossRef Search ADS PubMed 28. Senft C , Forster MT , Bink A et al. Optimizing the extent of resection in eloquently located gliomas by combining intraoperative MRI guidance with intraoperative neurophysiological monitoring . J Neurooncol . 2012 ; 109 ( 1 ): 81 – 90 . Google Scholar CrossRef Search ADS PubMed 29. Gasser T , Szelenyi A , Senft C et al. Intraoperative MRI and functional mapping . Acta Neurochir Suppl . 2011 ; 109 : 61 – 65 . Google Scholar CrossRef Search ADS PubMed 30. Szelényi A , Kothbauer KF , Deletis V . Transcranial electric stimulation for intraoperative motor evoked potential monitoring: stimulation parameters and electrode montages . Clin Neurophysiol . 2007 ; 118 ( 7 ): 1586 – 1595 . Google Scholar CrossRef Search ADS PubMed 31. Szelényi A , Senft C , Jardan M et al. Intra-operative subcortical electrical stimulation: a comparison of two methods . Clin. Neurophysiol . 2011 ; 122 ( 7 ): 1470 – 1475 . Google Scholar CrossRef Search ADS PubMed 32. Shiban E , Krieg SM , Haller B et al. Intraoperative subcortical motor evoked potential stimulation: how close is the corticospinal tract? J Neurosurg . 2015 ; 123 ( 3 ): 711 – 720 . Google Scholar CrossRef Search ADS PubMed 33. Cedzich C , Pechstein U , Schramm J , Schäfer S . Electrophysiological considerations regarding electrical stimulation of motor cortex and brain stem in humans . Neurosurgery . 1998 ; 42 ( 3 ): 527 – 532 . Google Scholar CrossRef Search ADS PubMed 34. Boehler T , van Straaten D , Wirtz S , Peitgen HO . A robust and extendible framework for medical image registration focused on rapid clinical application deployment . Comput Biol Med . 2011 ; 41 6 : 340 – 349 . Google Scholar CrossRef Search ADS PubMed 35. Coburger J , Musahl C , Henkes H et al. Comparison of navigated transcranial magnetic stimulation and functional magnetic resonance imaging for preoperative mapping in rolandic tumor surgery . Neurosurg Rev . 2013 ; 36 ( 1 ): 65 – 76 ; discussion 75-76 . Google Scholar CrossRef Search ADS PubMed 36. Frey D , Strack V , Wiener E , Jussen D , Vajkoczy P , Picht T . A new approach for corticospinal tract reconstruction based on navigated transcranial stimulation and standardized fractional anisotropy values . Neuroimage . 2012 ; 62 ( 3 ): 1600 – 1609 . Google Scholar CrossRef Search ADS PubMed 37. Weiss C Tursunova I Neuschmelting V et al. Improved nTMS- and DTI-derived CST tractography through anatomical ROI seeding on anterior pontine level compared to internal capsule . Neuroimage Clin. 2015 ; 7 : 424 – 437 . Google Scholar CrossRef Search ADS PubMed 38. Sollmann N , Wildschuetz N , Kelm A et al. Associations between clinical outcome and navigated transcranial magnetic stimulation characteristics in patients with motor-eloquent brain lesions: a combined navigated transcranial magnetic stimulation-diffusion tensor imaging fiber tracking approach . J Neurosurg . 2018 ;128(3): 800 – 810 . 39. Oishi K , Faria A , Jiang H et al. Atlas-based whole brain white matter analysis using large deformation diffeomorphic metric mapping: Application to normal elderly and Alzheimer's disease participants . Neuroimage . 2009 ; 46 ( 2 ): 486 – 499 . Google Scholar CrossRef Search ADS PubMed 40. Mori S , Oishi K , Jiang H et al. Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template . Neuroimage . 2008 ; 40 ( 2 ): 570 – 582 . Google Scholar CrossRef Search ADS PubMed 41. Oishi K , Zilles K , Amunts K et al. Human brain white matter atlas: identification and assignment of common anatomical structures in superficial white matter . Neuroimage . 2008 ; 43 ( 3 ): 447 – 457 . Google Scholar CrossRef Search ADS PubMed 42. Modersitzki J. Numerical Methods for Image Registration . Oxford : Oxford University Press ; 2004 . 43. Schlüter M . White matter lesion phantom for diffusion tensor data and its application to the assessment of fiber tracking . Med Imaging . 2005 ; 5746 : 835 – 844 . 44. Pan C , Peck KK , Young RJ , Holodny AI . Somatotopic organization of motor pathways in the internal capsule: a probabilistic diffusion tractography study . Am J Neuroradiol . 2012 ; 33 ( 7 ): 1274 – 1280 . Google Scholar CrossRef Search ADS PubMed 45. Holodny AI , Gor DM , Watts R , Gutin PH , Ulug AM . Diffusion-tensor MR tractography of somatotopic organization of corticospinal tracts in the internal capsule: initial anatomic results in contradistinction to prior reports . Radiology . 2005 ; 234 ( 3 ): 649 – 653 . Google Scholar CrossRef Search ADS PubMed 46. Klein J , Geisler B , Hahn HK . NeuroQLab DTI - Probabilistic Parameter Adaption for Efficient Fiber Tracking : Proceedings of MICCAI 2015 - DTI Tractography Challenge. Munich, Germany , 2015 . 47. Hattingen E , Rathert J , Jurcoane A et al. A standardised evaluation of pre-surgical imaging of the corticospinal tract: where to place the seed ROI . Neurosurg Rev . 2009 ; 32 ( 4 ): 445 – 456 . Google Scholar CrossRef Search ADS PubMed 48. Pinheiro JC , Bates DM . Mixed-Effects Models in S and S-PLUS . New York : Springer Verlag ; 2000 . Google Scholar CrossRef Search ADS 49. Bucci M , Mandelli ML , Berman JI et al. Quantifying diffusion MRI tractography of the corticospinal tract in brain tumors with deterministic and probabilistic methods . NeuroImage: Clinical . 2013 ; 3 : 361 – 368 . Google Scholar CrossRef Search ADS PubMed 50. Mandelli ML , Berger MS , Bucci M , Berman JI , Amirbekian B , Henry RG . Quantifying accuracy and precision of diffusion MR tractography of the corticospinal tract in brain tumors . J Neurosurg . 2014 ; 121 ( 2 ): 349 – 358 . Google Scholar CrossRef Search ADS PubMed 51. Javadi SA , Nabavi A , Giordano M , Faghihzadeh E , Samii A . Evaluation of diffusion tensor imaging-based tractography of the corticospinal tract: a correlative study with intraoperative magnetic resonance imaging and direct electrical subcortical stimulation . Neurosurgery . 2017 ; 80 ( 2 ): 287 – 299 . Google Scholar PubMed 52. Kamada K , Todo T , Ota T et al. The motor-evoked potential threshold evaluated by tractography and electrical stimulation . J Neurosurg . 2009 ; 111 ( 4 ): 785 – 795 . Google Scholar CrossRef Search ADS PubMed 53. Zolal A , Hejčl A , Vachata P et al. The use of diffusion tensor images of the corticospinal tract in intrinsic brain tumor surgery . Neurosurgery . 2012 ; 71 ( 2 ): 331 – 340 ; discussion 340 . Google Scholar CrossRef Search ADS PubMed 54. Maesawa S , Fujii M , Nakahara N , Watanabe T , Wakabayashi T , Yoshida J . Intraoperative tractography and motor evoked potential (MEP) monitoring in surgery for gliomas around the corticospinal tract . World Neurosurg . 2010 ; 74 ( 1 ): 153 – 161 . Google Scholar CrossRef Search ADS PubMed 55. Nossek E , Korn A , Shahar T et al. Intraoperative mapping and monitoring of the corticospinal tracts with neurophysiological assessment and 3-dimensional ultrasonography-based navigation . J Neurosurg . 2011 ; 114 ( 3 ): 738 – 746 . Google Scholar CrossRef Search ADS PubMed 56. Ohue S , Kohno S , Inoue A et al. Accuracy of diffusion tensor magnetic resonance imaging-based tractography for surgery of gliomas near the pyramidal tract . Neurosurgery . 2012 ; 70 ( 2 ): 283 – 294 ; discussion 294 . Google Scholar CrossRef Search ADS PubMed 57. Conti A , Raffa G , Granata F , Rizzo V , Germanò A , Tomasello F . Navigated transcranial magnetic stimulation for “somatotopic” tractography of the corticospinal tract . Neurosurgery . 2014 ; 10 ( Suppl 4) : 542 – 554 ; discussion 554 . Google Scholar CrossRef Search ADS PubMed 58. Ostrý S , Belšan T , Otáhal J , Beneš V , Netuka D . Is intraoperative diffusion tensor imaging at 3.0T comparable to subcortical corticospinal tract mapping? Neurosurgery . 2013 ; 73 ( 5 ): 797 – 807 ; discussion 806-897 . Google Scholar CrossRef Search ADS PubMed 59. Radmanesh A , Zamani AA , Whalen S , Tie Y , Suarez RO , Golby AJ . Comparison of seeding methods for visualization of the corticospinal tracts using single tensor tractography . Clin Neurol Neurosurg . 2015 ; 129 : 44 – 49 . Google Scholar CrossRef Search ADS PubMed 60. Forster MT , Limbart M , Seifert V , Senft C . Test-retest reliability of navigated transcranial magnetic stimulation of the motor cortex . Neurosurgery . 2014 ; 10 ( Suppl 1) : 51 – 55 ; discussion 55-56 . Google Scholar PubMed 61. Raffa G , Conti A , Scibilia A et al. Functional reconstruction of motor and language pathways based on navigated transcranial magnetic stimulation and DTI fiber tracking for the preoperative planning of low grade glioma surgery: a new tool for preservation and restoration of eloquent networks . Acta Neurochir Suppl . 2017 ; 124 : 251 – 261 . Google Scholar CrossRef Search ADS PubMed 62. Weiss Lucas C , Tursunova I , Neuschmelting V et al. Functional MRI vs. navigated TMS to optimize M1 seed volume delineation for DTI tractography. A prospective study in patients with brain tumours adjacent to the corticospinal tract . Neuroimage Clin . 2017 ; 13 : 297 – 309 . Google Scholar CrossRef Search ADS PubMed 63. Giussani C , Roux FE , Ojemann J , Sganzerla EP , Pirillo D , Papagno C . Is preoperative functional magnetic resonance imaging reliable for language areas mapping in brain tumor surgery? Review of language functional magnetic resonance imaging and direct cortical stimulation correlation studies . Neurosurgery . 2010 ; 66 ( 1 ): 113 – 120 . Google Scholar CrossRef Search ADS PubMed 64. Ulmer JL , Hacein-Bey L , Mathews VP et al. Lesion-induced pseudo-dominance at functional magnetic resonance imaging: implications for preoperative assessments . Neurosurgery . 2004 ; 55 ( 3 ): 569 – 581 ; discussion 580-581 . Google Scholar CrossRef Search ADS PubMed 65. Rostrup E , Law I , Blinkenberg M et al. Regional differences in the CBF and BOLD responses to hypercapnia: a combined PET and fMRI study . Neuroimage . 2000 ; 11 ( 2 ): 87 – 97 . Google Scholar CrossRef Search ADS PubMed 66. Preibisch C , Haase A . Perfusion imaging using spin-labeling methods: contrast- to-noise comparison in functional MRI applications . Magn Reson Med . 2001 ; 46 ( 1 ): 172 – 182 . Google Scholar CrossRef Search ADS PubMed 67. Helenius J , Perkiö J , Soinne L et al. Cerebral hemodynamics in a healthy population measured by dynamic susceptibility contrast MR imaging . Acta Radiol . 2003 ; 44 ( 5 ): 538 – 546 . Google Scholar CrossRef Search ADS PubMed 68. Ruohonen J , Karhu J . Navigated transcranial magnetic stimulation . Neurophysiol Clin . 2010 ; 40 ( 1 ): 7 – 17 . Google Scholar CrossRef Search ADS PubMed 69. Picht T , Mularski S , Kuehn B , Vajkoczy P , Kombos T , Suess O . Navigated transcranial magnetic stimulation for preoperative functional diagnostics in brain tumor surgery . Neurosurgery . 2009 ; 65 ( 6 Suppl ): 93 – 98 ; discussion 98-99 . Google Scholar PubMed 70. Krieg SM , Shiban E , Buchmann N et al. Utility of presurgical navigated transcranial magnetic brain stimulation for the resection of tumors in eloquent motor areas . J Neurosurg . 2012 ; 116 ( 5 ): 994 – 1001 . Google Scholar CrossRef Search ADS PubMed 71. Witwer BP , Moftakhar R , Hasan KM et al. Diffusion-tensor imaging of white matter tracts in patients with cerebral neoplasm . J Neurosurg . 2002 ; 97 ( 3 ): 568 – 575 . Google Scholar CrossRef Search ADS PubMed 72. Stadlbauer A , Nimsky C , Buslei R et al. Diffusion tensor imaging and optimized fiber tracking in glioma patients: Histopathologic evaluation of tumor-invaded white matter structures . Neuroimage . 2007 ; 34 ( 3 ): 949 – 956 . Google Scholar CrossRef Search ADS PubMed 73. Beppu T , Inoue T , Shibata Y et al. Fractional anisotropy value by diffusion tensor magnetic resonance imaging as a predictor of cell density and proliferation activity of glioblastomas . Surg Neurol . 2005 ; 63 ( 1 ): 56 – 61 ; discussion 61 . Google Scholar CrossRef Search ADS PubMed 74. Mormina E , Longo M , Arrigo A et al. MRI tractography of corticospinal tract and arcuate fasciculus in high-grade gliomas performed by constrained spherical deconvolution: qualitative and quantitative analysis . AJNR Am J Neuroradiol . 2015 ; 36 ( 10 ): 1853 – 1858 . Google Scholar CrossRef Search ADS PubMed 75. Stadlbauer A , Nimsky C , Gruber S et al. Changes in fiber integrity, diffusivity, and metabolism of the pyramidal tract adjacent to gliomas: a quantitative diffusion tensor fiber tracking and MR spectroscopic imaging study . AJNR Am J Neuroradiol . 2007 ; 28 ( 3 ): 462 – 469 . Google Scholar PubMed 76. Kinoshita M , Hashimoto N , Goto T et al. Fractional anisotropy and tumor cell density of the tumor core show positive correlation in diffusion tensor magnetic resonance imaging of malignant brain tumors . Neuroimage . 2008 ; 43 ( 1 ): 29 – 35 . Google Scholar CrossRef Search ADS PubMed 77. Li J , Chen X , Zhang J et al. Intraoperative diffusion tensor imaging predicts the recovery of motor dysfunction after insular lesions . Neural Regen Res . 2013 ; 8 ( 15 ): 1400 – 1409 . Google Scholar PubMed 78. Jenabi M , Peck KK , Young RJ , Brennan N , Holodny AI . Identification of the corticobulbar tracts of the tongue and face using deterministic and probabilistic DTI fiber tracking in patients with brain tumor . AJNR Am J Neuroradiol . 2015 ; 36 ( 11 ): 2036 – 2041 . Google Scholar CrossRef Search ADS PubMed 79. Garlapati RR , Roy A , Joldes GR et al. More accurate neuronavigation data provided by biomechanical modeling instead of rigid registration . J Neurosurg . 2014 ; 120 ( 6 ): 1477 – 1483 . Google Scholar CrossRef Search ADS PubMed 80. Clatz O , Delingette H , Talos IF et al. Robust nonrigid registration to capture brain shift from intraoperative MRI . IEEE Trans Med Imaging . 2005 ; 24 ( 11 ): 1417 – 1427 . Google Scholar CrossRef Search ADS PubMed 81. Ruiz-Alzola J , Westin CF , Warfield SK , Alberola C , Maier S , Kikinis R . Nonrigid registration of 3D tensor medical data . Med Image Anal . 2002 ; 6 ( 2 ): 143 – 161 . Google Scholar CrossRef Search ADS PubMed 82. Archip N , Clatz O , Whalen S et al. Non-rigid alignment of pre-operative MRI, fMRI, and DT-MRI with intra-operative MRI for enhanced visualization and navigation in image-guided neurosurgery . Neuroimage . 2007 ; 35 ( 2 ): 609 – 624 . Google Scholar CrossRef Search ADS PubMed 83. Mori S , van Zijl PC . Fiber tracking: principles and strategies - a technical review . NMR Biomed . 2002 ; 15 ( 7-8 ): 468 – 480 . Google Scholar CrossRef Search ADS PubMed 84. Glenn GR , Kuo LW , Chao YP , Lee CY , Helpern JA , Jensen JH . Mapping the orientation of white matter fiber bundles: a comparative study of diffusion tensor imaging, diffusional kurtosis imaging, and diffusion spectrum imaging . Am J Neuroradiol . 2016 ; 37 ( 7 ): 1216 – 1222 . Google Scholar CrossRef Search ADS PubMed 85. Barbieri S , Bauer MH , Klein J , Moltz J , Nimsky C , Hahn HK . DTI segmentation via the combined analysis of connectivity maps and tensor distances . Neuroimage . 2012 ; 60 ( 2 ): 1025 – 1035 . Google Scholar CrossRef Search ADS PubMed 86. Wedeen VJ , Wang RP , Schmahmann JD et al. Diffusion spectrum magnetic resonance imaging (DSI) tractography of crossing fibers . Neuroimage . 2008 ; 41 ( 4 ): 1267 – 1277 . Google Scholar CrossRef Search ADS PubMed 87. Raj A , Hess C , Mukherjee P . Spatial HARDI: improved visualization of complex white matter architecture with Bayesian spatial regularization . Neuroimage . 2011 ; 54 ( 1 ): 396 – 409 . Google Scholar CrossRef Search ADS PubMed 88. Wei PH , Cong F , Chen G , Li MC , Yu XG , Bao YH . Neuronavigation based on track density image extracted from deterministic high-definition fiber tractography . World Neurosurg . 2017 ; 98 : 880.e9 – 880.e15 . Google Scholar CrossRef Search ADS 89. Hoefnagels FWA , de Witt Hamer PC , Pouwels PJW , Barkhof F , Vandertop WP . Impact of gradient number and voxel size on diffusion tensor imaging tractography for resective brain surgery . World Neurosurg . 2017 ; 105 : 923 – 934.e2 . Google Scholar CrossRef Search ADS PubMed 90. Byrnes TJ , Barrick TR , Bell BA , Clark CA . Diffusion tensor imaging discriminates between glioblastoma and cerebral metastases in vivo . NMR Biomed . 2011 ; 24 ( 1 ): 54 – 60 . Google Scholar CrossRef Search ADS PubMed 91. Bauer AH , Erly W , Moser FG , Maya M , Nael K . Differentiation of solitary brain metastasis from glioblastoma multiforme: a predictive multiparametric approach using combined MR diffusion and perfusion . Neuroradiology . 2015 ; 57 ( 7 ): 697 – 703 . Google Scholar CrossRef Search ADS PubMed 92. Duffau H. Contribution of cortical and subcortical electrostimulation in brain glioma surgery: methodological and functional considerations . Neurophysiol Clin . 2007 ; 37 ( 6 ): 373 – 382 . Google Scholar CrossRef Search ADS PubMed 93. Seidel K , Beck J , Stieglitz L , Schucht P , Raabe A . The warning-sign hierarchy between quantitative subcortical motor mapping and continuous motor evoked potential monitoring during resection of supratentorial brain tumors . J Neurosurg . 2013 ; 118 ( 2 ): 287 – 296 . Google Scholar CrossRef Search ADS PubMed Acknowledgments The authors are grateful to Nadine Jahn and Stefanie Borchert for their excellent work in neuromonitoring. The authors also thank Marina Heibel for her help with figure and table design and acknowledge the contribution of Timothy Schaewe for his careful reading of the manuscript and providing language help. Copyright © 2018 by the Congress of Neurological Surgeons
Operative Neurosurgery – Oxford University Press
Published: Apr 14, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera