Transcranial magnetic stimulation and brain atrophy: a computer-based human brain model study
- First Online:
- Cite this article as:
- Wagner, T., Eden, U., Fregni, F. et al. Exp Brain Res (2008) 186: 539. doi:10.1007/s00221-007-1258-8
- 207 Views
This paper is aimed at exploring the effect of cortical brain atrophy on the currents induced by transcranial magnetic stimulation (TMS). We compared the currents induced by various TMS conditions on several different MRI derived finite element head models of brain atrophy, incorporating both decreasing cortical volume and widened sulci. The current densities induced in the cortex were dependent upon the degree and type of cortical atrophy and were altered in magnitude, location, and orientation when compared to healthy head models. Predictive models of the degree of current density attenuation as a function of the scalp-to-cortex distance were analyzed, concluding that those which ignore the electromagnetic field–tissue interactions lead to inaccurate conclusions. Ultimately, the precise site and population of neural elements stimulated by TMS in an atrophic brain cannot be predicted based on healthy head models which ignore the effects of the altered cortex on the stimulating currents. Clinical applications of TMS should be carefully considered in light of these findings.
KeywordsCortical atrophyFinite element modelTranscranial magnetic stimulationBrain stimulationNeuromodulationBrain plasticity
Transcranial magnetic stimulation (TMS) is a noninvasive procedure that utilizes pulsed magnetic fields to induce stimulating currents in cortical tissue (Barker et al. 1985). When this current is applied repetitively, repetitive transcranial magnetic stimulation (rTMS), it can modulate cortical excitability, decreasing or increasing it, depending on the parameters of stimulation. Since its inception, researchers have proposed the use of TMS and rTMS to study and treat neuropsychiatric diseases, such as major depression (Pascual-Leone et al. 1996; George et al. 2000; Martin et al. 2003; Holtzheimer et al. 2004; Rumi et al. 2005), schizophrenia (Hoffman et al. 2003; Lee et al. 2005), Parkinson’s disease (Mally and Stone 1999; de Groot et al. 2001; Khedr et al. 2003; Fregni et al. 2004; Lefaucheur et al. 2004), dystonia (Huang et al. 2004), epilepsy (Tergau et al. 1999; Menkes and Gruenthal 2000; Daniele et al. 2003; Fregni et al. 2005) and the acute or chronic sequels derived from stroke (Mansur et al. 2005). However, a fundamental question that needs to be addressed before the wide-spread use of TMS in clinical practice, is how the modification of brain anatomy and tissue properties caused by certain neuropsychiatric diseases can alter the effects of TMS.
With TMS, changes in the tissue anatomy and electromagnetic properties have been shown to alter the TMS induced stimulating currents in both phantom and modeling studies (Yunokuchi et al. 1998) (Liu and Ueno 2000; Wagner 2001; Miranda et al. 2003; Wagner et al. 2004). We have shown previously that the damaged areas in patients with stroke can perturb the location and magnitude of the stimulating cortical currents (Wagner et al. 2005). The main reason for this perturbation is that the altered distribution of cerebral spinal fluid (CSF) to brain tissue modifies the conductive tissue properties in the infarction region and effectively provides a path of lowered resistance for the stimulating currents to flow along. Several diseases explored or treated with TMS, such as depression (Pascual-Leone et al. 1996; George et al. 2000; Martin et al. 2003; Holtzheimer et al. 2004; Rumi et al. 2005), Alzheimer’s disease (Alagona et al. 2004; Pierantozzi et al. 2004), Huntington’s disease (Modugno et al. 2001), corticobasal degeneration (Kuhn et al. 2004) and Creutzfeldt–Jakob’s disease (Sakuishi et al. 2005), as well as normal aging (Sowell et al. 2003), show related anatomical changes that could have an impact on the TMS induced electric currents. All of these populations are characterized by varying degrees of cortical atrophy, where brain tissue is replaced by CSF, the CSF to cortical tissue volume ratios increase, and oftentimes the cerebral sulci widen. However to date, there has been no systematic study of the effects that cortical atrophy and the altered CSF to cortical tissue ratios can have on the currents induced by TMS in neural tissue.
Herein we systematically explored how the electrical and anatomical changes caused by cortical atrophy can perturb the TMS induced stimulating currents in the cortex through multiple MRI derived finite element TMS head models. We also discuss the potential clinical implications of current perturbations under TMS in patients with cortical atrophy.
Multiple MRI based finite element head models were constructed and evaluated for different stimulation orientations to address the effects of cortical atrophy on the TMS induced stimulating current.
MRI guided finite element head model
Mean conductivity values and relative permittivity schemes used
Mean conductivity (S/m)
Relative permittivity scheme (F/m)
1.2 × 104ε0
0.8 × 104ε0
Cerebral spinal fluid
0.60 × 104ε0
1.2 × 104ε0
1.2 × 104ε0
We implemented ten different atrophy models of increasing severity and varied cortical modifications to compare to the healthy head model under different stimulation conditions.
Six of the models were constructed by decreasing the overall volume of the gray and white matter symmetrically by 2.5% steps, from 100 to 85%, and filling the area with CSF as seen in in-vivo imaging and postmortem histology studies (Silbert et al. 2003; Sowell et al. 2003). These models will be referred to by their percent of atrophy (i.e., the 97.5% atrophy model). See Fig. 1b for a graphical representation of the models.
Four additional models were constructed to explore the effects of expanded cerebral sulci. These models were constructed by altering the healthy head model’s cortical geometry based on the MRIs of patients treated at Beth Israel Deaconess Medical Center (with atrophic regions cut with straight edges along their borders and with tissue regions clearly demarcated for use in the FEM solver). One model was constructed by removing the cortical tissue in the areas of the central sulcus and the sylvian fissure from the healthy head model and replacing the gray matter with CSF. This model is referred to as the base sulci model. Three further models were constructed based on this base sulci model, with symmetrically decreased cortical volumes, with 95, 90, and 85% of the brain volume of the base sulci model. See Fig. 1c for the base sulci model and its MRI basis.
The following specific tests were conducted to explore the effects of atrophy on the TMS induced stimulating currents:
Effects of the degree of atrophy on the induced stimulating currents
In order to investigate the effects that increasing atrophy can have on the induced stimulating currents, solutions were obtained and contrasted for both the healthy head model and the six symmetric atrophy models with the coil placed over the right dorsal lateral prefrontal (DLPF) cortex and the right primary (M1) motor cortex (with the coil center located at (43.8, 51.0, 48.0) and (43.1, 7.5, 51.0) respectively).
For each model and coil position, the magnitude and location of the maximum cortical current density were evaluated along with the surface area on the cortex where the current density was greater than 90% of its maximum value. We will refer to this as the maximum cortical current surface area. To examine the current behavior in the region of the expected current density maximum, the current density magnitude was determined along evaluation lines perpendicular to the coil at its center and at locations 10 mm anterior, posterior, rostral, and caudal to the center line (as defined in Fig. 1d). The distance from the skin surface to the cortical surface along each of these lines was determined for each model. Finally, the induced current density vector behavior was also analyzed in all the tissues.
Once the magnitude of the current density at the cortical interface and the distance from the scalp to the interface were determined for each model, plots were generated of the maximum magnitude on the surface of the cortex of each model relative to the healthy head model as a function of distance into the brain model along the aforementioned evaluation lines. We constructed a set of stochastic models of the relative magnitude of the cortical current density as a function of distance from the coil. In one embodiment, the current magnitude was modeled using a single decaying exponential function with independent identically distributed Gaussian noise (i.e., this model accounted for scalp to cortex distance alone and was represented by a single exponential function). An alternative model postulated that the current magnitude was described by separate distributions for each coil location, each with distinct exponential decay functions and noise variance parameters (i.e., this model accounted for both scalp-to-cortex distance and the coil position and was represented by two separate exponential functions, one for the DLPC and one for the motor strip coil locations—note that the coil position relative to the underlying electrical/anatomical distribution were constant for each line in this model embodiment). The exponential and variance parameters for the above models were fit with maximum likelihood methods. Using these models, we constructed a likelihood ratio test for the null hypothesis that the data from both coil positions was captured by the single exponential function versus the alternative hypothesis that the data was better explained by a distinct exponential function for each position. For this test, the log-likelihood ratio statistic should approximately follow a χ2 distribution with 3 degrees of freedom (Pawitan 2001).
Effects of widened sulci on the induced stimulating currents
Solutions were obtained with the coil placed above the dorsal lateral prefrontal cortex for the four widened sulci models and compared to the heterogeneous healthy head model and the analogous atrophy models with the coil in the same position. The analysis of the magnitude and location of the maximum cortical current density, the maximum cortical current surface area, the current density magnitude behavior along the evaluation lines, and the induced current density vector behavior detailed in the section above was completed for these models.
Effects of the degree of atrophy on the induced stimulating currents
Data for the maximum current density on the cortical surface
Model and position
% HH CDMax
MCDCS relative to coil
(29.8, 7.2, 35.3)
(−13.3, −0.3, −15.7)
(30.5, 9.9, 33.9)
(−12.6, 2.4, −17.1)
(29.7, 7.6, 33)
(−13.4, 0.1, −18.0)
(31.6, 7.4, 31)
(−11.5, −0.1, −20.0)
(28.5, 7.9, 31.1)
(−14.6, 0.4, −19.9)
(25.9, 12.2, 31)
(−17.2, 4.7, −20.0)
(22.4, 12.9, 31.1)
(−20.7, 5.4, −19.9)
(34.5, 45, 29.6)
(−9.3, −6, −18.4)
(34.5, 42.3, 28.86)
(−9.3, −8.7, −19.14)
(33.7, 41.1, 28.1)
(−10.1, −9.9, −19.9)
(33, 43.7, 26.5)
(−10.8, −7.3, −21.5)
(32, 42.7, 25.8)
(−11.8, −8.3, −22.2)
(31.1, 41.8, 24.9)
(−12.7, −9.2, −23.1)
(32.4, 42.9, 22.7)
(−11.4, −8.1, −25.3)
Data for the maximum current density on center evaluation line
Model and position
% HH CDMAx
LineDis from scalp
Exponential functions of the percentage of the maximum current density magnitude on the surface of the cortex of each model, relative to the healthy head model, as a function of distance into the brain model for each evaluation line
Evaluation line-coil position
Exponential trend line
y = 139.54e−0.0655x
y = 138.74e−0.0628x
y = 128.1e−0.0458x
y = 137.42e−0.0631x
Ventral line- MS
y = 112.07e−0.0225x
Center line -DLPC
y = 133.5e−0.0492x
y = 140.92e−0.0584x
y = 131.87e−0.0531x
y = 135.69e−0.054x
Ventral line- DLPC
y = 145.73e−0.0542x
Effects of widened sulci on the induced stimulating currents
In the widened sulci models, the behavior in the location of the current density maxima was less consistent. The behavior of the current density distribution directly under the coil center was very similar to the analogous atrophy models, but far less predictable along the widened sulci borders, similar to the results seen in stroke (Wagner 2001; Wagner et al. 2005) and heterogeneity studies (Wagner 2001; Miranda et al. 2003). There were locations directly below the most posterior portion of the figure-of-eight coil, along the widened central sulci border, where current density magnitudes were within ±15% of the maximums found under the coil center (3.59, −0.85, 6.61, and 14.35%, respectively for the base, 95, 90, and 85% widened sulci models).
Data for the maximum current density on the cortical surface for the widened sulci models
Model and position
MCDCS under coil
MCDCS location under coil
(32.9, 43.9, 28.1)
(34, 41.2, 28.0)
(37.9, 34.6, 25.3)
(37.8, 35.6, 22.1)
MCDCS on sulcal edges
MCDCS location on sulcal edges
(45.2, 4.1, 28.1)
(25.3, −2.2, 34.6)
(24.2, −3.8, 32.6)
Note: two other locations were within ±15% of the under the coil max: at (41.4, 1.83, 24.7) with 2.37 (8.59) magnitude and 3 mm2 area and at (45.2, −0.1, 20.4) with a 2.38 (8.62) magnitude and 4 mm2 area
(36.4, 2.06, 24.17)
Note: three other locations were within ±15% of the under the coil max: at (24.3, −5.3, 30.2) with a 2.05 (7.43) magnitude and 4 mm2 area, at (30.4, −7.2, 20.7) with a 2.21 (8.00) magnitude and 6 mm2 area and at (42.5, −0.4, 19.2) with a 2.09 (7.57) magnitude and a 4 mm2 area
The current density distribution directly under the coil center was very similar to the analogous atrophy models, and thus the behavior of the current densities along the evaluation lines were largely similar to those in the non-widened sulci models. Similarly, the current density vector behavior was consistent in the area under the figure-eight-center with the other atrophy models. However, the current density vector distributions deviated from predictable figure-of-eight distributions that were seen in the cortices of the models without the widened sulci to conform to the sulcal boundaries, such that the current vectors became more perpendicular to the borders. Additionally, in the CSF the current density vectors were directed into the widened sulci in the CSF surrounding the sulci (see Fig. 4b). Similar effects were seen in the current density behavior within the white matter along the widened sulcal boundaries.
This paper explores the effect that electrical and anatomical changes caused by cortical atrophy have on TMS induced electrical currents in the brain. The healthy head model was discussed in an earlier publication (Wagner et al. 2004), and herein we focus on the difference between healthy and atrophy head models. The results show that the alterations derived from cortical atrophy can modify the effect of TMS in several ways. As expected, the magnitude of the current densities induced on the cortex was dependent upon the degree of cortical atrophy and decreased in magnitude as the distance from the scalp to the cortex increased. However, the degree of attenuation did not solely depend on the distance, but also on the relative coil position and anatomical and electrical features of the tissues. The location of the maximum cortical current density varied with the degree and type of atrophy, whereas current density vectors were altered in the widened sulci models, conforming to the altered tissue geometries. In addition to analyzing the current density results for the individual models, exponential trend lines were fit to the data to predict how the current densities would decay at different locations in the models. Along the individual evaluation lines, the models that accounted for position and distance fit the data very well, but no single exponential model successfully captured the relation between cortical current density and the distance between the scalp and cortex. The clinical implications of some of our observations for TMS based tests and therapies are discussed.
Symmetric atrophy models
The magnitude of the maximum cortical current density decreased with increasing scalp to cortex distances in each of the models. While maintaining the coil location at a constant position and evaluating the current density as a function of the distance, the degree of attenuation could be well predicted by exponential functions, with the appropriate decay rate for a given coil position and evaluation line. Along a single evaluation line, these exponential models of distance alone explained over 97% of the variability in the cortical current density magnitude, suggesting that the distance between the scalp and cortex is an important predictor of the expected degree of attenuation when TMS is used in the settings of cortical atrophy. However, no one single function based on the distance alone could predict the degree of attenuation for the coil placed at different locations on the scalp (or for a single coil position at different locations). Our data strongly indicate that the functional form of the current magnitude attenuation differs between the two coil positions studied, suggesting that other factors such as the relative coil to scalp location and the electrical properties of the tissues significantly affect the final current density distribution.
In light of these findings, we conclude that the value of functions, based solely on static magnetic field measurements or simplified models as those adjusting the source magnetic field strength based on the scalp to cortex distance (Bohning 2000; Nahas et al. 2004; Knecht et al. 2005; Stokes et al. 2005) are insufficient and the implications of their clinical use need to be critically reevaluated.
Widened sulci models
For the TMS frequency spectrum and given tissue resistive values implemented in the model, the current density vector distribution is governed by the condition that the normal current density vector components must be continuous across boundaries of media with differing conductivities (Zahn 2003; Wagner et al. 2005). Thus, when one goes from the highly conductive CSF to the less conductive cerebral tissue at the widened sulci borders (or from the more conductive gray matter to the less conductive white matter), a jump in the normal component of the electrical field dictating final current density direction and magnitude along sulcal borders is expected. Thus, new small areas—less than 10 mm2—of maximum cortical current density magnitude found along the sulcal borders, are areas of least resistance indicative of these changes in current density distribution.
The extremes described in these widened sulci models (where current densities were seen in excess of 150% as compared to identical regions in the analogous atrophy models without altered sulci features) will most likely be rare in clinical cases of atrophy, since the linear edges will be more curvilinear in human brains. However, in large areas of perturbation surrounding these regions of extreme perturbation there was a consistent increase in the current density (see Fig. 4a). These current amplifications were dependent upon the geometry of the widened sulci and are basically indicative of regions with lowered resistance. In addition to the changes in magnitude, the current density vector orientation was altered along the sulcal borders. As stated above, the normal components of the current density must be continuous across tissue boundaries (CSF, white matter, gray matter) at sulcal borders, resulting in the alteration of the current density vector orientations (see Fig. 4b for an example).
It is important to note that many of those effects could also theoretically occur in normal sulcal regions to varying degrees. These will obviously depend on the geometry and electrical tissue distributions in the regions of the sulci, but earlier studies have clearly shown the effects of heterogeneities, as those operating at sulcal borders, can be extremely significant (Wagner 2001: Wagner et al. 2004, 2005; Miranda et al. 2003). As is supported by studies examining corner points in other electromagnetic models (Deeley 1990), we hypothesize that the effects will be most extensive at regions where the border geometries result in corner regions, such as could be seen along irregular edges of scar tissue along an infarction border. However, future studies need to be conducted to explore the subject since many other variables likely to have an influence (such as the effects of very tight sulcal folds) need further exploration (see Miranda et al. 2003 for further discussion).
In this study we examined the alterations in the pattern of TMS induced current density distributions in both symmetric atrophy and widened sulci models. In the clinical setting, such clear demarcation of both types of structural abnormalities is unlikely and to some degree, a combination of both will be always present. According to our predictions, under conditions of increasing atrophy or sulcal width alterations, current density distributions will be altered in magnitude, orientation, and location; all of which might significantly alter the location and manner by which a core population of neural elements will be practically affected by TMS-induced currents, in spite of the initial targeting and intended plan. Even though the scalp-brain distance alone is not the only variable accounting for the final current density distributions, it is clear that the use of TMS under conditions of increasing scalp to cortical distances requires special precautions. In particular, one should not make use of varied scalp to cortex distance to infer a single, general stimulus intensity versus depth relationship. In vitro experimental and modeling data suggests that the site of activation is predicted by the peak electric field magnitude (Amassian et al. 1992; Maccabee et al. 1993; Nagarajan et al. 1993; Nagarajan and Durand 1995). Moreover, in vivo TMS experiments in both the motor and visual cortex have provided evidence that stimulation occurs at the location of the peak electric field (Wassermann et al. 1996; Krings et al. 1997; Boroojerdi et al. 1999). Thus, even with minimal changes in the neural architecture (i.e., relative neural cell to current density orientations) current density magnitude attenuation seen with increasing atrophy should lead to an expected alteration in physiological (e.g., MEP, phosphene values) or behavioral (recovery from a given cognitive sequel) outputs in atrophic regions as compared to nonatrophic areas (see Fig. 3b). This effect is nonlinear and brain location dependent; and as such, it cannot be captured by a single function as the current modifications will be patient specific and dependent upon the coil position relative to a the specific tissue distribution.
In spite of the demonstrated decay of the current density magnitude at increasing scalp-to-coil distances, simple linear increases in the applied TMS intensity should not be pursued in the clinical setting, even if there is no perturbation in the current location. Oftentimes brain atrophy is also associated with a change in brain activity and therefore the remaining active neurons in the atrophic brain might have an altered excitability, as proven for example in patients with Alzheimer’s and Parkinson’s disease displaying motor cortex hyperexcitability (Ridding et al. 1995; de Carvalho et al. 1997; Liepert et al. 2001; Bhatia et al. 2003; Di Lazzaro et al. 2004; Lefaucheur 2005), and thus respond differently to the same patterns of induced electric current.
In addition to the magnitude changes, the vector current density orientation in the cortex was altered proximal to the widened sulci. With such changes, different neural elements as those initially targeted could also be activated proximal to the sulci as it is clear that the directionality of the induced currents plays a clear role in which neurons are stimulated (Chiappa 1994). Finally, this study also provides evidence that the focus of induced current from TMS can be appreciably diminished in the region of widened sulci with multiple disjoint areas, around the sulcal borders, where the current density is near its maximum or greatly increased as compared to the corresponding atrophy models without the widened sulci. In cases where the coil is placed more proximal or overlying the widened sulci, the current density distributions will be even less predictable than those accounted for with the coil positions that were implemented (dependent upon the individual geometry and electrical makeup of the sulci) as has been seen in the case of stroke (Wagner et al. 2005).
Many of these problems could be addressed by combining electromagnetic field solvers, based on individualized MRI based solutions, with stereotactic tracking technology to provide the clinician with a tool to predict the location, orientation, and magnitude of stimulation based on the patients unique electromagnetic tissue interactions which constrain the final stimulating currents. In the meantime, until such a solution is implemented, care must be taken, particularly in regions of pathologies, to not make oversimplifications based on distance alone.
Study limitations and areas of future research
Some limitations of this study suggest areas of possible future research. First, the resolution of the model is constrained by the CAD rendering of the human head, which unfortunately does not share the same resolution as the MRI that was used to derive it. The shape of the widened sulci was similarly limited in detail because of the same resolution restrictions. In order to cope with these limitations, the field solver must be modified to work on structural high spatial resolution MRI data sets directly. Additionally, further research needs to be completed on the low frequency tissue electrical properties to account for their dispersive properties, anisotropies, and heterogeneities. The model should be expanded in the future to account for these tissue properties as data becomes available. Finally, we limited our discussion to the cortical current densities and their perturbation by structural alterations during the “on line” local effects of TMS and in the future the models could be extended to address “off line” effects through the integration of physiological models of stimulation derived from human or animal studies (Valero-Cabre et al. 2005, 2006, 2007; Wagner et al. 2007).
We demonstrate that TMS induced currents in the cortex can be modified in magnitude, location, and orientation under conditions of brain atrophy and/or sulcus widening. These cortical current density perturbations could prove to be dangerous or at the very least lead to unreliable results as those initially intended if guided by computational TMS neuronavigation and current distribution models that neglect the crucial role of structural and tissue electromagnetic alterations induced by the underlying conditions of the treated patients. Future efforts and model refinements will help to further increase our understanding of the effects of atrophy on TMS.
We acknowledge the contribution of the Maxwell software from the Ansoft Corporation, Pittsburgh, PA that was used in this study. This work was supported in part by Center for Integration of Medicine and Innovative Technology (CIMIT) under Grant numbers 45335 and 20736, and by grants from the National Institutes of Health (K24 RR018875, RO1 EY12091, R01 EB005047, RO1 DC05672, RO1 NS47754, RO1 NS20068). F.F. received support under a grant from the Harvard Medical School Scholars in Clinical Science Program (NIH K30 HL04095-03). AV-C was supported in part by the National Institutes of Health (NS32137 and NS33975 to BP/MM-AVC, and grants from ‘La Caixa’ (Spain) and the Spanish Ministry of Education Culture and Sports, EX2002-041).