A Gaussian Process Emulator for Estimating the Volume of Tissue Activated During Deep Brain Stimulation

Pava, Iván De La; Gómez, V.; López, Mariano Jiménez; Henao, Oscar; Daza-Santacoloma, Genaro; Orozco, Álvaro A.

doi:10.1007/978-3-319-19390-8_77

Cited by 7 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep brain simulations use as building blocks single neuron cell models, which are based on differential equations and therefore require considerable computational time. (De La Pava et al, 2015) developed a new methodology to reduce the computation time required to estimate the volume of tissue activated during deep brain stimulation. At the heart of their method is an emulator that replaces the system of differential equations with combined multi-compartment axon models coupled to the stimulating electric field with a Gaussian process classifier.…”

Section: Deep Brain Stimulationmentioning

confidence: 99%

“…The emulator is constructed by fitting a Gaussian process to outcome data. The approach of (De La Pava et al, 2015) reduced by a factor of 10 the average computational runtime of Volume Tissue Activated estimation compared with the gold standard. (Melozzi, Woodman, Jirsa, & Bernard, 2017) describe the benefits of in silico experimentation with The Virtual Mouse Brain (TVMB), a platform that facilitates investigation of large-scale mouse brain dynamics.…”

Section: Deep Brain Stimulationmentioning

confidence: 99%

See 1 more Smart Citation

Statistical Methods for Computer Experiments with Applications in Neuroscience

Shapira¹,

Shaposhnik²,

Steinberg³

2019

Preprint

View full text Add to dashboard Cite

Many scientific and technological problems are studied with the help of computer codes that simulate the phenomena of interest rather than via traditional laboratory experiments. Such models play an important role in neuroscience where they are used to mimic brain function from the sub-cellular to the macroscopic level. Exploration with computer models carries with it a number of statistical challenges: where to sample the input space for the simulator, how to make sense of the data that is generated, how to estimate unknown parameters in the model, how to validate a model. The simulator setting also has some unique problems and possibilities. This review paper describes statistical research on these issues and how that work might be applied to neural simulations.

show abstract

Section: Deep Brain Stimulationmentioning

confidence: 99%

Section: Deep Brain Stimulationmentioning

confidence: 99%

Statistical Methods for Computer Experiments with Applications in Neuroscience

Shapira¹,

Shaposhnik²,

Steinberg³

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Our aim was then to train a Gaussian process classifier (GPC) to determine whether an axon at a given position in space was active due to DBS, and by doing so, to estimate the VTA. This work is an extended version of a study presented in the 7th Iberian Conference on Pattern Recognition and Image Analysis [16]. The present version contains a more detailed theoretical framework, and a larger experimental setup that includes the estimation of the VTA when realistic anisotropic brain tissue conductivity conditions are considered.…”

Section: Introductionmentioning

confidence: 99%

Accelerating the computation of the volume of tissue activated during deep brain stimulation using Gaussian processes

Panche

Gómez-Orozco

López³

et al. 2017

Rev. Fac. Ing. Univ. Antioquia

Self Cite

View full text Add to dashboard Cite

The volume of tissue activated (VTA) is a well-established approach to model the direct effect of deep brain stimulation (DBS) on neural tissue. Previous studies have pointed to its potential clinical applications. However, the elevated computational runtime required to estimate the VTA with standard techniques used in biological neural modeling limits its suitability for practical use. The goal of this study was to develop a novel methodology to reduce the computation time of VTA estimation. To that end, we built a Gaussian process emulator. It combines multicompartment axon models coupled to the stimulating electric field with a Gaussian process classifier (GPC), following the premise that computing the VTA from a field of axons is in essence a binary classification problem. We achieved a considerable reduction in the average time required to estimate the VTA, under both ideal isotropic and realistic anisotropic brain tissue conductive conditions, limiting the loss of accuracy and overcoming other drawbacks entailed by alternative methods. RESUMEN:El volumen de tejido activo (VTA) es un enfoque bien establecido para modelar los efectos directos de la estimulación cerebral profunda en el tejido neuronal. Estudios previos han señalado sus posibles aplicaciones clínicas. Sin embargo, el elevado costo computacional requerido para estimar el VTA con las técnicas estándar utilizadas en el modelado neuronal biológico limita su usabilidad a nivel práctico. El objetivo de este estudio fue desarrollar una metodología novedosa para reducir el tiempo de cálculo en la estimación del VTA. Con ese fin, se construyó un emulador basado en procesos gaussianos. Este combina modelos axonales de múltiples compartimientos acoplados al campo de estimulación eléctrica con un sistema de clasificación basado en procesos gaussianos, siguiendo la premisa de que calcular el VTA a partir de un campo axonal es en esencia un problema de clasificación binaria. Se logró una reducción considerable en el tiempo promedio requerido para estimar el VTA, tanto bajo condiciones de conductividad isotrópica idealizada como bajo condiciones realistas de conductividad anisotrópica, limitando la perdida de precisión y superando otros inconvenientes presentes en métodos alternativos.

show abstract