Accepted Manuscript Title: Non-invasive mapping of bilateral motor speech areas using navigated transcranial magnetic stimulation and functional magnetic resonance imaging Author: Mervi K¨on¨onen Niko Tamsi Laura S¨ais¨anen Samuli Kemppainen Sara M¨aa¨ tt¨a Petro Julkunen Leena Jutila Marja ¨ a Reetta K¨alvi¨ainen Eini Niskanen Ritva Vanninen Pasi Aiki¨ Karjalainen Esa Mervaala PII: DOI: Reference:
S0165-0270(15)00131-4 http://dx.doi.org/doi:10.1016/j.jneumeth.2015.03.030 NSM 7197
To appear in:
Journal of Neuroscience Methods
Received date: Revised date: Accepted date:
13-4-2014 26-3-2015 27-3-2015
Please cite this article as: K¨on¨onen M, Tamsi N, S¨ais¨anen L, Kemppainen S, M¨aa¨ tt¨a ¨ a M, K¨alvi¨ainen R, Niskanen E, Vanninen R, Karjalainen S, Julkunen P, Jutila L, Aiki¨ P, Mervaala E, Non-invasive mapping of bilateral motor speech areas using navigated transcranial magnetic stimulation and functional magnetic resonance imaging, Journal of Neuroscience Methods (2015), http://dx.doi.org/10.1016/j.jneumeth.2015.03.030 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ip t cr
Ac c
ep te
d
M
an
us
Highlights rTMS disturbs both the speech process and the motor execution of speech. Reviewers verified speech problems induced by rTMS on both hemispheres. Locations of rTMS trials were overlaid with fMRI activations of word generation task. rTMS results overlapped with the areas of fMRI activation only on left hemispheres. By combining rTMS and fMRI, motor speech areas can be separated from language areas.
1
Page 1 of 31
ip t cr
us
Non-invasive mapping of bilateral motor speech areas using navigated transcranial magnetic stimulation and functional magnetic resonance imaging
an
Mervi Könönen1,2,† ,*, Niko Tamsi3,† , Laura Säisänen1,3, Samuli Kemppainen1, Sara Määttä1, Petro Julkunen1, Leena Jutila4, Marja Äikiä4, Reetta Kälviäinen4,5, Eini Niskanen6, Ritva Vanninen2,7, Pasi Karjalainen6, Esa Mervaala1,3 Department of Clinical Neurophysiology, Kuopio University Hospital, P.O.Box 100, 70029 Kuopio, Finland
2
Department of Radiology, Kuopio University Hospital, P.O.Box 100, 70029 Kuopio, Finland
3
Institute of Clinical Medicine, Clinical Neurophysiology, University of Eastern Finland, 70211 Kuopio, Finland
4
Department of Neurology, Kuopio University Hospital, P.O.Box 100, 70029 Kuopio, Finland
5
Institute of Clinical Medicine, Neurology, University of Eastern Finland, 70211 Kuopio, Finland
6
Department of Applied Physics, University of Eastern Finland, 70211 Kuopio, Finland
7
Institute of Clinical Medicine, Clinical Radiology, University of Eastern Finland, 70211 Kuopio, Finland
†
ep te
d
M
1
These authors contributed equally to this work
*Correspondence: Mervi Könönen P.O. Box 100
Ac c
Department of Clinical Neurophysiology, Kuopio University Hospital FI-70029 Kuopio, Finland phone: +358 44 7113253 fax: +358-17-173244
e-mail: [email protected] Words: 233 (abstract), 4862 (excluding acknowledgements and references)
2
Page 2 of 31
ip t cr
Abstract
us
Background: Navigated transcranial magnetic stimulation (nTMS) is a modern precise method to activate and study cortical functions noninvasively. We hypothesized that a combination of rTMS and functional magnetic resonance imaging (fMRI) could clarify the localization of functional areas involved with motor control and production of speech.
an
New Method: Repetitive nTMS with short bursts was used to map speech areas on both hemispheres by inducing speech disruption during number recitation tasks in healthy volunteers. Two experienced video reviewers, blinded to the stimulated area, graded each trial offline according to possible speech disruption. The locations of speech disrupting nTMS trials were overlaid with fMRI activations of word generation task.
M
Results, Comparison with Existing methods: Speech disruptions were produced on both hemispheres by nTMS, though there were more disruptive stimulation sites on the left hemisphere. Grade of the disruptions varied from subjective sensation to mild objectively recognizable disruption up to total speech arrest. The distribution of locations in which speech disruptions could be elicited varied among individuals. On the left hemisphere the locations of disturbing rTMS bursts with reviewers’ verification followed the areas of fMRI activation. Similar pattern was not observed on the right hemisphere.
d
Conclusions: The reviewer-verified speech disruptions induced by nTMS provided clinically relevant information, and fMRI might explain further the function of the cortical area. nTMS and fMRI complement each other, and their combination should be advocated when assessing individual localization of speech network.
DCS, Direct cortical stimulation; fMRI, Functional magnetic resonance imaging; MRI, Magnetic resonance imaging; MT, Motor threshold; NBS, Navigated brain stimulation; NRS, numeric rating scale; nTMS, Navigated transcranial magnetic stimulation; rTMS, Repetitive transcranial magnetic stimulation; TMS, Transcranial magnetic stimulation; WGEN, Word generation task
Ac c
Abbreviations
ep te
Keywords: navigated transcranial magnetic stimulation, nTMS, motor speech area, preoperative mapping, bilateral speech area mapping, functional magnetic resonance imaging
4
Page 3 of 31
ip t cr us
Introduction
The reliable identification of eloquent areas of motor, sensory, and language functions is crucial when performing neurosurgical resections. The gold
an
standard for this kind of mapping is intraoperative awake craniotomy with direct cortical stimulation (DCS) (Szelenyi et al., 2010). For motor cortex, recent noninvasive techniques such as navigated transcranial magnetic stimulation (nTMS) (Forster et al., 2011; Krieg et al., 2012; Picht et al., 2011;
M
Vitikainen et al., 2012) and functional magnetic resonance imaging (fMRI) (Bizzi et al., 2008; Lehericy et al., 2000) enable reliable preoperative mapping as well. The nTMS has been proven to be more accurate than fMRI when compared to DCS (Forster et al., 2011; Mangraviti et al., 2013).
d
However, the mapping of cortical areas responsible for speech and language functions is more challenging. Previous DCS studies of cortical areas
ep te
critical for language functions have revealed a combination of several discrete important cortical locations, although with substantial inter-individual variability (Corina et al., 2010; Ojemann et al., 1989).
Ac c
Conventional TMS has been applied to speech area identification for years, and different stimulation techniques and responses (i.e., types of linguistic errors) induced by repetitive TMS (rTMS) have been explored extensively (Aziz-Zadeh et al., 2005; Devlin and Watkins, 2007; Epstein et al., 1999; Flitman et al., 1998; Stewart et al., 2001). The results, however, have varied. rTMS has been shown to produce speech arrest or naming errors in patients with epilepsy (Pascual-Leone et al., 1991; Wassermann et al., 1999) with lateralizing findings parallel to the Wada test (also known as the intracarotid amobarbital test), but also non-uniform results between rTMS and Wada test have been reported (Epstein et al., 2000; Jennum et al., 1994). In some epilepsy patients, rTMS has favored the right hemisphere dominance at a rate significantly greater than the Wada test. This difference may
5
Page 4 of 31
ip t cr
relate to the specific features of TMS induced speech arrest, which appears to occur also at the motor area in addition to classic language areas (Epstein
us
et al., 2000). In these TMS studies, the source of the speech arrest, i.e. motor arrest or classic language areas, has not been further clarified.
an
Furthermore, the functional cortical areas were localized according to external cranial landmarks, not based on individual brain anatomy. Therefore,
M
the exact stimulated brain area has neither been certain.
In nTMS, a stereotactic MRI-based on-line navigation system is integrated with the TMS device, allowing the operator to accurately stimulate specific
d
anatomical brain locations within an error margin of about 5-6 mm (Ruohonen and Karhu, 2010). nTMS has proven reliable for the presurgical
ep te
mapping of primary motor areas in tumor patients (Forster et al., 2011; Krieg et al., 2014; Picht et al., 2009; Picht et al., 2011). Recent study on patients with lesions in language-eloquent regions have shown that also language function mapping with navigated rTMS correlates well with intraoperative DCS (Picht et al., 2013b). They suggest that rTMS is a lesion-based approach like DCS, and therefore allows targeted analysis of
Ac c
circumscribed cortical areas essential for function. However, lesion-based approaches may not provide a complete view of the complexity of the language. Interobserver and intraobserver comparisons only corresponded partially, when the left hemisphere language mapping by navigated TMS was studied using object naming task and categorizing the error types in seven different types (Sollmann et al., 2013a). When the right hemisphere's contribution to language processing in the healthy human brain was investigated by categorizing the naming errors, the middle and ventral parts of preand postcentral gyri were related to speech-motor control functions and the middle and posterior middle frontal gyri were supporting language task performance (Sollmann et al., 2014).
6
Page 5 of 31
ip t cr
us
fMRI is a widely used method for preoperative assessment of the language areas and the hemispheric dominance of language function. Word
an
generation (WGEN) paradigms especially have been shown to provide robust and reproducible localization of language areas and to generate high degree of laterality (Brannen et al., 2001; Mahdavi et al., 2011; Partovi et al., 2012; Zaca et al., 2013). The high degree of laterality seen in WGEN
M
tasks has been thought to be related in language production. Since the pure motor functions related in overt speech i.e. articulation and phonation as well as orofacial movements have been shown to produce bilateral activation in the human motor cortex (Brown et al., 2009; Grabski et al., 2012). The
d
advantage of fMRI is the ability to illustrate the activated areas of the whole network (Rutten and Ramsey, 2010), including activations in more deeply
ep te
located cortical areas as well, while direct brain stimulation methods, DCS and TMS, provide very local and discrete functional location affecting only a specific part of a neuronal network involved in speech production. However, the feature of fMRI to locate the contributory areas as well impedes the
Ac c
reliable localization of the most essential areas in language processing.
Our study focused on stimulating the eloquent motor speech areas in both hemispheres with navigated rTMS to explore its effects on aloud speech. The locations of speech-disrupting rTMS bursts were compared to areas of fMRI activation for WGEN task to differentiate whether rTMS induced speech disruption in language production or in motor execution of speech. If the disruptive stimulation locations were on the area with fMRI activation, the area was interpreted to represent language production and not the motor speech, whereas if there was no fMRI activation the area was interpreted to
7
Page 6 of 31
ip t cr
represent the motor area of orofacial muscles participating in language processing. Furthermore, for applications with clinical purpose, we also
an
us
evaluated the possible subjective complaints and side-effects i.e., pain experience and facial muscle contractions induced by nTMS.
M
2. Material and methods 2.1. Study participants
d
Originally, twenty healthy adults volunteered to a language fMRI study to form a control group for epilepsy surgery candidates. For clinical purposes
ep te
the age range was chosen wide, and both left and right handed subjects were accepted to achieve as good a resemblance as possible between controls and patients. The results of the fMRI data of the twenty healthy subjects have been published previously (Niskanen et al., 2012). One year later ten of these subjects (six females and four males) further volunteered to participate in the current nTMS study. Two participants were left-handed and one was ambidextrous, according to the 20-item modified Edinburgh Handedness Inventory (Oldfield 1971). One left-handed participant had to be
Ac c
excluded from the study because the resting motor threshold (MT) of her abductor pollicis brevis muscle was too high to be determined, even at the maximum stimulus intensity. The mean age of the remaining nine subjects was 38 years (range 21–49 years). According to the fMRI results, all volunteers, except S2, had left hemispheric dominance of language functions (Table 2). To ensure safety, all subjects completed questionnaire forms including the issues described in the most recent TMS safety guidelines to exclude any contraindications for TMS (Rossi et al., 2009). The study was approved by the local Ethical Committee.
8
Page 7 of 31
ip t cr
2.2. Magnetic resonance imaging
us
To obtain an anatomical reference for nTMS and fMRI, MR images were acquired using a 1.5T scanner (Siemens Magnetom Avanto, Siemens AG, Germany) with the following parameters: a three-dimensional T1-weighted Magnetization Prepared Rapid Acquisition Gradient Echo sequence,
2.3. Functional magnetic resonance imaging
M
an
repetition time 1980 ms, echo time 3.09 ms, flip angle 15°, matrix 256 × 256, 176 contiguous sagittal slices with isotropic 1 mm3 voxels.
d
Functional images consisted of 36 contiguous slices with a voxel size of 3×3×3 mm3 acquired parallel to the anterior commissure –posterior
ep te
commissure plane with a gradient-echo echo-planar imaging sequence (repetition time 3100ms, echo time 50 ms, flip angle 90º, matrix 64×64, slice acquisition order interleaved). The verbal fluency task is a classical neuropsychological test to assess verbal abilities and constitutes a widely used task for functional imaging of speech related areas. Therefore, a similar fMRI task, a word generation task (WGEN) was chosen as the fMRI correlate to the
Ac c
nTMS results. The task paradigm consisted of blocks of active condition in which the task was to covertly generate different words starting with the visually given alternating letter (one letter per block). In the control condition, subjects were shown a fixation cross and were instructed to stop generating words and instead to perform a simple self-paced finger tapping task alternating with the right and left index fingers. The active baseline task instead of passive one was chosen, because activity during a passive resting baseline may vary unpredictably (Gusnard et al., 2001; Peck et al., 2004). The task was practiced carefully before scanning, and silent word generation was emphasized; as well as mechanical button pressing as baseline
9
Page 8 of 31
ip t cr
without thinking or counting. The fMRI paradigm consisted of five control blocks separated with four blocks of active condition, 10 scans per block,
an
us
block duration 33 seconds, and total acquisition time 5 minutes.
2.4. Navigated transcranial magnetic stimulation (nTMS)
M
The Magstim Rapid stimulator (Magstim Company Ltd, Whitland, Wales, UK) and figure-of-eight coil (Double 70mm, PN9925) were connected to a stereotactic on-line Navigated Brain Stimulation (NBS) system (eXimia software version 3.1., Nexstim Ltd, Helsinki, Finland). Prior to the
d
stimulations, the occurrence of any possible adverse effects or sensations (e.g., pain, distractive muscle contractions, and speech disruption of different
ep te
types) during stimulations was thoroughly explained to the subjects. The resting MT of the right abductor pollicis brevis muscle was determined using the Rossini-Rothwell method on the left hemisphere (Rossini et al., 1994; Säisänen et al., 2008). Mapping of the motor speech area was done using the
Ac c
number reciting task without knowledge of the fMRI results. The number recitation task was chosen as an easy, fluently speech-producing task.
The mapping was initiated by focusing nTMS at the posterior frontal lobe areas of the left hemisphere. Participants were instructed to continuously recite numbers aloud and to try to continue reciting even if their speech was disturbed. Furthermore, they were instructed to take their head off the coil if stimulation was too uncomfortable. Stimulation bursts (intensity 100–150% of MT, frequency 5 Hz and burst duration 4 s) were always delivered from one to three seconds after the voluntary recitation of numbers had started. The stimulus intensity was initially 120% MT, and was elevated in 10% MT increments to a maximum of 150% MT, or until subjective speech disruption was reported or observed. Stimulation parameters were chosen
10
Page 9 of 31
ip t cr
according to Epstein et al (1996; 1999). At that intensity, the stimulation area was extended concentrically until there was no disruption in speech. The
us
stimulated area covered posterior frontal lobe, anterior parietal lobe and in some subject even superior temporal lobe. For two subjects stimulation
an
intensities below 120% MT (100% or 110% MT) were used as well. Bursts of stimuli were applied with spacing of 5-10 mm. Next, the corresponding posterior frontal lobe area of the right hemisphere was stimulated similarly. Both hemispheres were examined in the same session in five subjects,
M
whereas four subjects were examined in two separate sessions on different days because of practicalities. All stimulation sessions were also recorded
d
for offline analysis (video and audio), and the recordings were segmented to samples of trials for blinded reviewer evaluation.
ep te
Each participant assessed his/her subjective experience immediately after each stimulation burst. First, they judged whether speech was disrupted: (1) “definitely”, (2) “possibly” or (3) “definitely not”. Secondly, they were asked to describe how the speech was disrupted. Thirdly, they were asked to assess the level of pain using a numeric rating scale (NRS): 0 meaning no pain and 10 meaning the worst imaginable pain. Also distracting muscle
Ac c
contractions were enquired. After the nTMS examinations were completed, the participants were requested to pay attention to possible adverse effects which were asked from them on the following day.
2.5. Analysis
Functional MRI data was analysed with the SPM5 software package (the Wellcome Trust Center for Neuroimaging, UCL, London, UK, www.fil.ion.ucl.ac.uk/spm/) running under Matlab R2007a (The Mathworks Inc., Natick, MA, USA). Pre-processing of the images included motion
11
Page 10 of 31
ip t cr
correction, correction of the acquisition time difference between slices, and spatial smoothing with a Gaussian kernel of 6mm FWHM. Statistical
us
analysis was performed on a voxel-by-voxel basis using the General Linear Model (Worsley and Friston, 1995), comparing the active condition and the
an
control condition with a T-test. The T-maps in individual space with threshold of uncorrected p-value less than 0.001 were generated for TMS comparison. For the laterality index calculation, the pre-processing step with normalization to the standard ICBM152-space was added; and the regions
M
of interest were defined using the atlases in WFU PickAtlas. The laterality index was calculated as LI = [(L-R)/(L+R)], where L and R were the number of voxels surviving the threshold of 80% of the maximum T-value in predefined region of interest (Broca area, Wernicke area, Heschl gyrus
ep te
d
and the hippocampus) on the left and the right hemisphere, respectively (Chlebus et al., 2007).
A neurologist (L.J.) and a neuropsychologist (M.Ä.), clinically experienced in the evaluation of speech disruption during the Wada test, analyzed the video samples separately and blinded to the location of stimuli. Both reviewers graded each trial for three clinically relevant classifications: (Class 1)
Ac c
words are unrecognizable, or there is a speech arrest or aphasia (speech disruptions, as in the Wada test); (Class 2) pronunciation is significantly disrupted or imprecise, but words are recognizable; or (Class 3) there is no detectable disruption.
The evaluations of the two reviewers and volunteers’ subjective experience of definitive speech disruption together categorized further the effects of stimuli in different locations (Table 1). Category 1: both reviewers graded that words were unrecognizable or there was arrest or aphasia (Class 1). Category 2: one reviewer evaluated significant disruption in speech or in pronunciation (Class 1 or 2) and the other reviewer evaluated significant
12
Page 11 of 31
ip t cr
disruption in pronunciation (Class 2). Category 3: only one reviewer judged that at least a significant disruption in pronunciation occurred (Class 1 or
an
us
2). Category 4: no objective finding judged by the reviewers.
The nTMS sessions and T-value maps of fMRI were imported to eXimia software version 3.2 which allowed visualization of both methods
M
simultaneously. The rTMS locations were presented over each subjects’ T-map and examined visually. Statistical analyses were performed using SPSS (version 19.0, IBM Corporation, Somers, NY, USA). Non-parametric tests (Kruskal-Wallis test and
3. Results
ep te
Smirnov test. Significance level was set at P < 0.05.
d
Spearman’s rho correlation) were used because the response variables were not normally distributed, as verified using the one-sample Kolmogorov-
3.1. Speech disruptions with nTMS
Ac c
On average, 60 stimulation bursts per subject were delivered (range 41–82), 33 stimulus bursts (range 25–47) to the left and 27 (11–41) to the right hemisphere. The individual numbers of stimulus bursts, subjective experiences, and reviewers’ judgements are summarized in Table 2. The assessment of subjective experience immediately after each TMS burst with verbal comments is presented in Table 3, which shows that rTMS elicited sensorimotor activations to all subjects and both hemispheres, when elicited speech arrest was dominant during stimulations on left hemisphere.
13
Page 12 of 31
ip t cr
The level of experienced pain or the stimulation intensity did not affect the subjective evaluations of speech disturbance, although the level of pain
us
correlated significantly with the stimulation intensity (ρ=0.196, p
S0165-0270(15)00131-4 http://dx.doi.org/doi:10.1016/j.jneumeth.2015.03.030 NSM 7197
To appear in:
Journal of Neuroscience Methods
Received date: Revised date: Accepted date:
13-4-2014 26-3-2015 27-3-2015
Please cite this article as: K¨on¨onen M, Tamsi N, S¨ais¨anen L, Kemppainen S, M¨aa¨ tt¨a ¨ a M, K¨alvi¨ainen R, Niskanen E, Vanninen R, Karjalainen S, Julkunen P, Jutila L, Aiki¨ P, Mervaala E, Non-invasive mapping of bilateral motor speech areas using navigated transcranial magnetic stimulation and functional magnetic resonance imaging, Journal of Neuroscience Methods (2015), http://dx.doi.org/10.1016/j.jneumeth.2015.03.030 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ip t cr
Ac c
ep te
d
M
an
us
Highlights rTMS disturbs both the speech process and the motor execution of speech. Reviewers verified speech problems induced by rTMS on both hemispheres. Locations of rTMS trials were overlaid with fMRI activations of word generation task. rTMS results overlapped with the areas of fMRI activation only on left hemispheres. By combining rTMS and fMRI, motor speech areas can be separated from language areas.
1
Page 1 of 31
ip t cr
us
Non-invasive mapping of bilateral motor speech areas using navigated transcranial magnetic stimulation and functional magnetic resonance imaging
an
Mervi Könönen1,2,† ,*, Niko Tamsi3,† , Laura Säisänen1,3, Samuli Kemppainen1, Sara Määttä1, Petro Julkunen1, Leena Jutila4, Marja Äikiä4, Reetta Kälviäinen4,5, Eini Niskanen6, Ritva Vanninen2,7, Pasi Karjalainen6, Esa Mervaala1,3 Department of Clinical Neurophysiology, Kuopio University Hospital, P.O.Box 100, 70029 Kuopio, Finland
2
Department of Radiology, Kuopio University Hospital, P.O.Box 100, 70029 Kuopio, Finland
3
Institute of Clinical Medicine, Clinical Neurophysiology, University of Eastern Finland, 70211 Kuopio, Finland
4
Department of Neurology, Kuopio University Hospital, P.O.Box 100, 70029 Kuopio, Finland
5
Institute of Clinical Medicine, Neurology, University of Eastern Finland, 70211 Kuopio, Finland
6
Department of Applied Physics, University of Eastern Finland, 70211 Kuopio, Finland
7
Institute of Clinical Medicine, Clinical Radiology, University of Eastern Finland, 70211 Kuopio, Finland
†
ep te
d
M
1
These authors contributed equally to this work
*Correspondence: Mervi Könönen P.O. Box 100
Ac c
Department of Clinical Neurophysiology, Kuopio University Hospital FI-70029 Kuopio, Finland phone: +358 44 7113253 fax: +358-17-173244
e-mail: [email protected] Words: 233 (abstract), 4862 (excluding acknowledgements and references)
2
Page 2 of 31
ip t cr
Abstract
us
Background: Navigated transcranial magnetic stimulation (nTMS) is a modern precise method to activate and study cortical functions noninvasively. We hypothesized that a combination of rTMS and functional magnetic resonance imaging (fMRI) could clarify the localization of functional areas involved with motor control and production of speech.
an
New Method: Repetitive nTMS with short bursts was used to map speech areas on both hemispheres by inducing speech disruption during number recitation tasks in healthy volunteers. Two experienced video reviewers, blinded to the stimulated area, graded each trial offline according to possible speech disruption. The locations of speech disrupting nTMS trials were overlaid with fMRI activations of word generation task.
M
Results, Comparison with Existing methods: Speech disruptions were produced on both hemispheres by nTMS, though there were more disruptive stimulation sites on the left hemisphere. Grade of the disruptions varied from subjective sensation to mild objectively recognizable disruption up to total speech arrest. The distribution of locations in which speech disruptions could be elicited varied among individuals. On the left hemisphere the locations of disturbing rTMS bursts with reviewers’ verification followed the areas of fMRI activation. Similar pattern was not observed on the right hemisphere.
d
Conclusions: The reviewer-verified speech disruptions induced by nTMS provided clinically relevant information, and fMRI might explain further the function of the cortical area. nTMS and fMRI complement each other, and their combination should be advocated when assessing individual localization of speech network.
DCS, Direct cortical stimulation; fMRI, Functional magnetic resonance imaging; MRI, Magnetic resonance imaging; MT, Motor threshold; NBS, Navigated brain stimulation; NRS, numeric rating scale; nTMS, Navigated transcranial magnetic stimulation; rTMS, Repetitive transcranial magnetic stimulation; TMS, Transcranial magnetic stimulation; WGEN, Word generation task
Ac c
Abbreviations
ep te
Keywords: navigated transcranial magnetic stimulation, nTMS, motor speech area, preoperative mapping, bilateral speech area mapping, functional magnetic resonance imaging
4
Page 3 of 31
ip t cr us
Introduction
The reliable identification of eloquent areas of motor, sensory, and language functions is crucial when performing neurosurgical resections. The gold
an
standard for this kind of mapping is intraoperative awake craniotomy with direct cortical stimulation (DCS) (Szelenyi et al., 2010). For motor cortex, recent noninvasive techniques such as navigated transcranial magnetic stimulation (nTMS) (Forster et al., 2011; Krieg et al., 2012; Picht et al., 2011;
M
Vitikainen et al., 2012) and functional magnetic resonance imaging (fMRI) (Bizzi et al., 2008; Lehericy et al., 2000) enable reliable preoperative mapping as well. The nTMS has been proven to be more accurate than fMRI when compared to DCS (Forster et al., 2011; Mangraviti et al., 2013).
d
However, the mapping of cortical areas responsible for speech and language functions is more challenging. Previous DCS studies of cortical areas
ep te
critical for language functions have revealed a combination of several discrete important cortical locations, although with substantial inter-individual variability (Corina et al., 2010; Ojemann et al., 1989).
Ac c
Conventional TMS has been applied to speech area identification for years, and different stimulation techniques and responses (i.e., types of linguistic errors) induced by repetitive TMS (rTMS) have been explored extensively (Aziz-Zadeh et al., 2005; Devlin and Watkins, 2007; Epstein et al., 1999; Flitman et al., 1998; Stewart et al., 2001). The results, however, have varied. rTMS has been shown to produce speech arrest or naming errors in patients with epilepsy (Pascual-Leone et al., 1991; Wassermann et al., 1999) with lateralizing findings parallel to the Wada test (also known as the intracarotid amobarbital test), but also non-uniform results between rTMS and Wada test have been reported (Epstein et al., 2000; Jennum et al., 1994). In some epilepsy patients, rTMS has favored the right hemisphere dominance at a rate significantly greater than the Wada test. This difference may
5
Page 4 of 31
ip t cr
relate to the specific features of TMS induced speech arrest, which appears to occur also at the motor area in addition to classic language areas (Epstein
us
et al., 2000). In these TMS studies, the source of the speech arrest, i.e. motor arrest or classic language areas, has not been further clarified.
an
Furthermore, the functional cortical areas were localized according to external cranial landmarks, not based on individual brain anatomy. Therefore,
M
the exact stimulated brain area has neither been certain.
In nTMS, a stereotactic MRI-based on-line navigation system is integrated with the TMS device, allowing the operator to accurately stimulate specific
d
anatomical brain locations within an error margin of about 5-6 mm (Ruohonen and Karhu, 2010). nTMS has proven reliable for the presurgical
ep te
mapping of primary motor areas in tumor patients (Forster et al., 2011; Krieg et al., 2014; Picht et al., 2009; Picht et al., 2011). Recent study on patients with lesions in language-eloquent regions have shown that also language function mapping with navigated rTMS correlates well with intraoperative DCS (Picht et al., 2013b). They suggest that rTMS is a lesion-based approach like DCS, and therefore allows targeted analysis of
Ac c
circumscribed cortical areas essential for function. However, lesion-based approaches may not provide a complete view of the complexity of the language. Interobserver and intraobserver comparisons only corresponded partially, when the left hemisphere language mapping by navigated TMS was studied using object naming task and categorizing the error types in seven different types (Sollmann et al., 2013a). When the right hemisphere's contribution to language processing in the healthy human brain was investigated by categorizing the naming errors, the middle and ventral parts of preand postcentral gyri were related to speech-motor control functions and the middle and posterior middle frontal gyri were supporting language task performance (Sollmann et al., 2014).
6
Page 5 of 31
ip t cr
us
fMRI is a widely used method for preoperative assessment of the language areas and the hemispheric dominance of language function. Word
an
generation (WGEN) paradigms especially have been shown to provide robust and reproducible localization of language areas and to generate high degree of laterality (Brannen et al., 2001; Mahdavi et al., 2011; Partovi et al., 2012; Zaca et al., 2013). The high degree of laterality seen in WGEN
M
tasks has been thought to be related in language production. Since the pure motor functions related in overt speech i.e. articulation and phonation as well as orofacial movements have been shown to produce bilateral activation in the human motor cortex (Brown et al., 2009; Grabski et al., 2012). The
d
advantage of fMRI is the ability to illustrate the activated areas of the whole network (Rutten and Ramsey, 2010), including activations in more deeply
ep te
located cortical areas as well, while direct brain stimulation methods, DCS and TMS, provide very local and discrete functional location affecting only a specific part of a neuronal network involved in speech production. However, the feature of fMRI to locate the contributory areas as well impedes the
Ac c
reliable localization of the most essential areas in language processing.
Our study focused on stimulating the eloquent motor speech areas in both hemispheres with navigated rTMS to explore its effects on aloud speech. The locations of speech-disrupting rTMS bursts were compared to areas of fMRI activation for WGEN task to differentiate whether rTMS induced speech disruption in language production or in motor execution of speech. If the disruptive stimulation locations were on the area with fMRI activation, the area was interpreted to represent language production and not the motor speech, whereas if there was no fMRI activation the area was interpreted to
7
Page 6 of 31
ip t cr
represent the motor area of orofacial muscles participating in language processing. Furthermore, for applications with clinical purpose, we also
an
us
evaluated the possible subjective complaints and side-effects i.e., pain experience and facial muscle contractions induced by nTMS.
M
2. Material and methods 2.1. Study participants
d
Originally, twenty healthy adults volunteered to a language fMRI study to form a control group for epilepsy surgery candidates. For clinical purposes
ep te
the age range was chosen wide, and both left and right handed subjects were accepted to achieve as good a resemblance as possible between controls and patients. The results of the fMRI data of the twenty healthy subjects have been published previously (Niskanen et al., 2012). One year later ten of these subjects (six females and four males) further volunteered to participate in the current nTMS study. Two participants were left-handed and one was ambidextrous, according to the 20-item modified Edinburgh Handedness Inventory (Oldfield 1971). One left-handed participant had to be
Ac c
excluded from the study because the resting motor threshold (MT) of her abductor pollicis brevis muscle was too high to be determined, even at the maximum stimulus intensity. The mean age of the remaining nine subjects was 38 years (range 21–49 years). According to the fMRI results, all volunteers, except S2, had left hemispheric dominance of language functions (Table 2). To ensure safety, all subjects completed questionnaire forms including the issues described in the most recent TMS safety guidelines to exclude any contraindications for TMS (Rossi et al., 2009). The study was approved by the local Ethical Committee.
8
Page 7 of 31
ip t cr
2.2. Magnetic resonance imaging
us
To obtain an anatomical reference for nTMS and fMRI, MR images were acquired using a 1.5T scanner (Siemens Magnetom Avanto, Siemens AG, Germany) with the following parameters: a three-dimensional T1-weighted Magnetization Prepared Rapid Acquisition Gradient Echo sequence,
2.3. Functional magnetic resonance imaging
M
an
repetition time 1980 ms, echo time 3.09 ms, flip angle 15°, matrix 256 × 256, 176 contiguous sagittal slices with isotropic 1 mm3 voxels.
d
Functional images consisted of 36 contiguous slices with a voxel size of 3×3×3 mm3 acquired parallel to the anterior commissure –posterior
ep te
commissure plane with a gradient-echo echo-planar imaging sequence (repetition time 3100ms, echo time 50 ms, flip angle 90º, matrix 64×64, slice acquisition order interleaved). The verbal fluency task is a classical neuropsychological test to assess verbal abilities and constitutes a widely used task for functional imaging of speech related areas. Therefore, a similar fMRI task, a word generation task (WGEN) was chosen as the fMRI correlate to the
Ac c
nTMS results. The task paradigm consisted of blocks of active condition in which the task was to covertly generate different words starting with the visually given alternating letter (one letter per block). In the control condition, subjects were shown a fixation cross and were instructed to stop generating words and instead to perform a simple self-paced finger tapping task alternating with the right and left index fingers. The active baseline task instead of passive one was chosen, because activity during a passive resting baseline may vary unpredictably (Gusnard et al., 2001; Peck et al., 2004). The task was practiced carefully before scanning, and silent word generation was emphasized; as well as mechanical button pressing as baseline
9
Page 8 of 31
ip t cr
without thinking or counting. The fMRI paradigm consisted of five control blocks separated with four blocks of active condition, 10 scans per block,
an
us
block duration 33 seconds, and total acquisition time 5 minutes.
2.4. Navigated transcranial magnetic stimulation (nTMS)
M
The Magstim Rapid stimulator (Magstim Company Ltd, Whitland, Wales, UK) and figure-of-eight coil (Double 70mm, PN9925) were connected to a stereotactic on-line Navigated Brain Stimulation (NBS) system (eXimia software version 3.1., Nexstim Ltd, Helsinki, Finland). Prior to the
d
stimulations, the occurrence of any possible adverse effects or sensations (e.g., pain, distractive muscle contractions, and speech disruption of different
ep te
types) during stimulations was thoroughly explained to the subjects. The resting MT of the right abductor pollicis brevis muscle was determined using the Rossini-Rothwell method on the left hemisphere (Rossini et al., 1994; Säisänen et al., 2008). Mapping of the motor speech area was done using the
Ac c
number reciting task without knowledge of the fMRI results. The number recitation task was chosen as an easy, fluently speech-producing task.
The mapping was initiated by focusing nTMS at the posterior frontal lobe areas of the left hemisphere. Participants were instructed to continuously recite numbers aloud and to try to continue reciting even if their speech was disturbed. Furthermore, they were instructed to take their head off the coil if stimulation was too uncomfortable. Stimulation bursts (intensity 100–150% of MT, frequency 5 Hz and burst duration 4 s) were always delivered from one to three seconds after the voluntary recitation of numbers had started. The stimulus intensity was initially 120% MT, and was elevated in 10% MT increments to a maximum of 150% MT, or until subjective speech disruption was reported or observed. Stimulation parameters were chosen
10
Page 9 of 31
ip t cr
according to Epstein et al (1996; 1999). At that intensity, the stimulation area was extended concentrically until there was no disruption in speech. The
us
stimulated area covered posterior frontal lobe, anterior parietal lobe and in some subject even superior temporal lobe. For two subjects stimulation
an
intensities below 120% MT (100% or 110% MT) were used as well. Bursts of stimuli were applied with spacing of 5-10 mm. Next, the corresponding posterior frontal lobe area of the right hemisphere was stimulated similarly. Both hemispheres were examined in the same session in five subjects,
M
whereas four subjects were examined in two separate sessions on different days because of practicalities. All stimulation sessions were also recorded
d
for offline analysis (video and audio), and the recordings were segmented to samples of trials for blinded reviewer evaluation.
ep te
Each participant assessed his/her subjective experience immediately after each stimulation burst. First, they judged whether speech was disrupted: (1) “definitely”, (2) “possibly” or (3) “definitely not”. Secondly, they were asked to describe how the speech was disrupted. Thirdly, they were asked to assess the level of pain using a numeric rating scale (NRS): 0 meaning no pain and 10 meaning the worst imaginable pain. Also distracting muscle
Ac c
contractions were enquired. After the nTMS examinations were completed, the participants were requested to pay attention to possible adverse effects which were asked from them on the following day.
2.5. Analysis
Functional MRI data was analysed with the SPM5 software package (the Wellcome Trust Center for Neuroimaging, UCL, London, UK, www.fil.ion.ucl.ac.uk/spm/) running under Matlab R2007a (The Mathworks Inc., Natick, MA, USA). Pre-processing of the images included motion
11
Page 10 of 31
ip t cr
correction, correction of the acquisition time difference between slices, and spatial smoothing with a Gaussian kernel of 6mm FWHM. Statistical
us
analysis was performed on a voxel-by-voxel basis using the General Linear Model (Worsley and Friston, 1995), comparing the active condition and the
an
control condition with a T-test. The T-maps in individual space with threshold of uncorrected p-value less than 0.001 were generated for TMS comparison. For the laterality index calculation, the pre-processing step with normalization to the standard ICBM152-space was added; and the regions
M
of interest were defined using the atlases in WFU PickAtlas. The laterality index was calculated as LI = [(L-R)/(L+R)], where L and R were the number of voxels surviving the threshold of 80% of the maximum T-value in predefined region of interest (Broca area, Wernicke area, Heschl gyrus
ep te
d
and the hippocampus) on the left and the right hemisphere, respectively (Chlebus et al., 2007).
A neurologist (L.J.) and a neuropsychologist (M.Ä.), clinically experienced in the evaluation of speech disruption during the Wada test, analyzed the video samples separately and blinded to the location of stimuli. Both reviewers graded each trial for three clinically relevant classifications: (Class 1)
Ac c
words are unrecognizable, or there is a speech arrest or aphasia (speech disruptions, as in the Wada test); (Class 2) pronunciation is significantly disrupted or imprecise, but words are recognizable; or (Class 3) there is no detectable disruption.
The evaluations of the two reviewers and volunteers’ subjective experience of definitive speech disruption together categorized further the effects of stimuli in different locations (Table 1). Category 1: both reviewers graded that words were unrecognizable or there was arrest or aphasia (Class 1). Category 2: one reviewer evaluated significant disruption in speech or in pronunciation (Class 1 or 2) and the other reviewer evaluated significant
12
Page 11 of 31
ip t cr
disruption in pronunciation (Class 2). Category 3: only one reviewer judged that at least a significant disruption in pronunciation occurred (Class 1 or
an
us
2). Category 4: no objective finding judged by the reviewers.
The nTMS sessions and T-value maps of fMRI were imported to eXimia software version 3.2 which allowed visualization of both methods
M
simultaneously. The rTMS locations were presented over each subjects’ T-map and examined visually. Statistical analyses were performed using SPSS (version 19.0, IBM Corporation, Somers, NY, USA). Non-parametric tests (Kruskal-Wallis test and
3. Results
ep te
Smirnov test. Significance level was set at P < 0.05.
d
Spearman’s rho correlation) were used because the response variables were not normally distributed, as verified using the one-sample Kolmogorov-
3.1. Speech disruptions with nTMS
Ac c
On average, 60 stimulation bursts per subject were delivered (range 41–82), 33 stimulus bursts (range 25–47) to the left and 27 (11–41) to the right hemisphere. The individual numbers of stimulus bursts, subjective experiences, and reviewers’ judgements are summarized in Table 2. The assessment of subjective experience immediately after each TMS burst with verbal comments is presented in Table 3, which shows that rTMS elicited sensorimotor activations to all subjects and both hemispheres, when elicited speech arrest was dominant during stimulations on left hemisphere.
13
Page 12 of 31
ip t cr
The level of experienced pain or the stimulation intensity did not affect the subjective evaluations of speech disturbance, although the level of pain
us
correlated significantly with the stimulation intensity (ρ=0.196, p
