Real-time computation of the TMS-induced electric field in a realistic head model

Research output: Contribution to journalArticleScientificpeer-review

Standard

Real-time computation of the TMS-induced electric field in a realistic head model. / Stenroos, Matti; Koponen, Lari M.

In: NeuroImage, 09.09.2019.

Research output: Contribution to journalArticleScientificpeer-review

Harvard

APA

Vancouver

Author

Bibtex - Download

@article{7c8cdf888eec465a8b9311f774833224,
title = "Real-time computation of the TMS-induced electric field in a realistic head model",
abstract = "Transcranial magnetic stimulation (TMS) is often targeted using a model of TMS-induced electric field (E). In such navigated TMS, the E-field models have been based on spherical approximation of the head. Such models omit the effects of cerebrospinal fluid (CSF) and gyral folding, leading to potentially large errors in the computed E-field. So far, realistic models have been too slow for interactive TMS navigation. We present computational methods that enable real-time solving of the E-field in a realistic five-compartment (5-C) head model that contains isotropic white matter, gray matter, CSF, skull and scalp. Using reciprocity and Geselowitz integral equation, we separate the computations to coil-dependent and -independent parts. For the Geselowitz integrals, we present a fast numerical quadrature. Further, we present a moment-matching approach for optimizing dipole-based coil models. We verified and benchmarked the new methods using simulations with over 100 coil locations. The new quadrature introduced a relative error (RE) of 0.3–0.6{\%}. For a coil model with 42 dipoles, the total RE of the quadrature and coil model was 0.44–0.72{\%}. Taking also other model errors into account, the contribution of the new approximations to the RE was 0.1{\%}. For comparison, the RE due to omitting the separation of white and gray matter was >11{\%}, and the RE due to omitting also the CSF was >23{\%}. After the coil-independent part of the model has been built, E-fields can be computed very quickly: Using a standard PC and basic GPU, our solver computed the full E-field in a 5-C model in 9000 points on the cortex in 27 coil positions per second (cps). When the separation of white and gray matter was omitted, the speed was 43–65 cps. Solving only one component of the E-field tripled the speed. The presented methods enable real-time solving of the TMS-induced E-field in a realistic head model that contains the CSF and gyral folding. The new methodology allows more accurate targeting and precise adjustment of stimulation intensity during experimental or clinical TMS mapping.",
keywords = "Coil model, Electric field calculation, Navigated transcranial magnetic stimulation, Transcranial magnetic stimulation (TMS), Volume conductor model",
author = "Matti Stenroos and Koponen, {Lari M.}",
year = "2019",
month = "9",
day = "9",
doi = "10.1016/j.neuroimage.2019.116159",
language = "English",
journal = "NeuroImage",
issn = "1053-8119",

}

RIS - Download

TY - JOUR

T1 - Real-time computation of the TMS-induced electric field in a realistic head model

AU - Stenroos, Matti

AU - Koponen, Lari M.

PY - 2019/9/9

Y1 - 2019/9/9

N2 - Transcranial magnetic stimulation (TMS) is often targeted using a model of TMS-induced electric field (E). In such navigated TMS, the E-field models have been based on spherical approximation of the head. Such models omit the effects of cerebrospinal fluid (CSF) and gyral folding, leading to potentially large errors in the computed E-field. So far, realistic models have been too slow for interactive TMS navigation. We present computational methods that enable real-time solving of the E-field in a realistic five-compartment (5-C) head model that contains isotropic white matter, gray matter, CSF, skull and scalp. Using reciprocity and Geselowitz integral equation, we separate the computations to coil-dependent and -independent parts. For the Geselowitz integrals, we present a fast numerical quadrature. Further, we present a moment-matching approach for optimizing dipole-based coil models. We verified and benchmarked the new methods using simulations with over 100 coil locations. The new quadrature introduced a relative error (RE) of 0.3–0.6%. For a coil model with 42 dipoles, the total RE of the quadrature and coil model was 0.44–0.72%. Taking also other model errors into account, the contribution of the new approximations to the RE was 0.1%. For comparison, the RE due to omitting the separation of white and gray matter was >11%, and the RE due to omitting also the CSF was >23%. After the coil-independent part of the model has been built, E-fields can be computed very quickly: Using a standard PC and basic GPU, our solver computed the full E-field in a 5-C model in 9000 points on the cortex in 27 coil positions per second (cps). When the separation of white and gray matter was omitted, the speed was 43–65 cps. Solving only one component of the E-field tripled the speed. The presented methods enable real-time solving of the TMS-induced E-field in a realistic head model that contains the CSF and gyral folding. The new methodology allows more accurate targeting and precise adjustment of stimulation intensity during experimental or clinical TMS mapping.

AB - Transcranial magnetic stimulation (TMS) is often targeted using a model of TMS-induced electric field (E). In such navigated TMS, the E-field models have been based on spherical approximation of the head. Such models omit the effects of cerebrospinal fluid (CSF) and gyral folding, leading to potentially large errors in the computed E-field. So far, realistic models have been too slow for interactive TMS navigation. We present computational methods that enable real-time solving of the E-field in a realistic five-compartment (5-C) head model that contains isotropic white matter, gray matter, CSF, skull and scalp. Using reciprocity and Geselowitz integral equation, we separate the computations to coil-dependent and -independent parts. For the Geselowitz integrals, we present a fast numerical quadrature. Further, we present a moment-matching approach for optimizing dipole-based coil models. We verified and benchmarked the new methods using simulations with over 100 coil locations. The new quadrature introduced a relative error (RE) of 0.3–0.6%. For a coil model with 42 dipoles, the total RE of the quadrature and coil model was 0.44–0.72%. Taking also other model errors into account, the contribution of the new approximations to the RE was 0.1%. For comparison, the RE due to omitting the separation of white and gray matter was >11%, and the RE due to omitting also the CSF was >23%. After the coil-independent part of the model has been built, E-fields can be computed very quickly: Using a standard PC and basic GPU, our solver computed the full E-field in a 5-C model in 9000 points on the cortex in 27 coil positions per second (cps). When the separation of white and gray matter was omitted, the speed was 43–65 cps. Solving only one component of the E-field tripled the speed. The presented methods enable real-time solving of the TMS-induced E-field in a realistic head model that contains the CSF and gyral folding. The new methodology allows more accurate targeting and precise adjustment of stimulation intensity during experimental or clinical TMS mapping.

KW - Coil model

KW - Electric field calculation

KW - Navigated transcranial magnetic stimulation

KW - Transcranial magnetic stimulation (TMS)

KW - Volume conductor model

UR - http://www.scopus.com/inward/record.url?scp=85072510776&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2019.116159

DO - 10.1016/j.neuroimage.2019.116159

M3 - Article

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

M1 - 116159

ER -

ID: 38293203