Brain and Human Body Modeling Computational

Human Modeling at EMBC 2018 Sergey Makarov pdf

download
https://textbookfull.com/product/brain-and-human-body-modeling-computational-human-modeling-at-
embc-2018-sergey-makarov/

★★★★★ 4.6/5.0 (28 reviews) ✓ 78 downloads ■ TOP RATED

"Fantastic PDF quality, very satisfied with download!" - Emma W.

DOWNLOAD EBOOK
Brain and Human Body Modeling Computational Human Modeling
at EMBC 2018 Sergey Makarov pdf download

TEXTBOOK EBOOK TEXTBOOK FULL

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Collection Highlights

Brain and Human Body Modeling 2020: Computational Human

Models Presented at EMBC 2019 and the BRAIN Initiative®
2019 Meeting Sergey N. Makarov

Dormancy in Aquatic Organisms Theory Human Use and

Modeling Victor R. Alekseev

Advances in Simulation and Digital Human Modeling 1st

Edition Daniel N Cassenti (Editor)

How It Works Inside the Human body 2018 Edition Edition

Chetan P.S
The Human Brain Book Rita Carter

Computational Modeling in Bioengineering and

Bioinformatics 1st Edition Nenad Filipovic

Enterprise and Organizational Modeling and Simulation 14th

International Workshop EOMAS 2018 Held at CAiSE 2018
Tallinn Estonia June 11 12 2018 Selected Papers Robert
Pergl

The concise human body book Steve Parker

Magnetic Resonance Brain Imaging Modeling and Data

Analysis Using R Jörg Polzehl
Sergey Makarov · Marc Horner
Gregory Noetscher Editors

Brain and
Human Body
Modeling
Computational Human Modeling at
EMBC 2018
Brain and Human Body Modeling
Sergey Makarov • Marc Horner
Gregory Noetscher
Editors

Brain and Human Body

Modeling
Computational Human Modeling
at EMBC 2018
Editors
Sergey Makarov Marc Horner
Massachusetts General Hospital ANSYS, Inc.
Boston, MA, USA Evanston, IL, USA
Worcester Polytechnic Institute
Worcester, MA, USA

Gregory Noetscher
Worcester Polytechnic Institute
Worcester, MA, USA

This book is an open access publication.

ISBN 978-3-030-21292-6    ISBN 978-3-030-21293-3 (eBook)
https://doi.org/10.1007/978-3-030-21293-3

© The Editor(s) (if applicable) and The Author(s) 2019

Open Access This book is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit
to the original author(s) and the source, provide a link to the Creative Commons license and indicate if
changes were made.
The images or other third party material in this book are included in the book’s Creative Commons
license, unless indicated otherwise in a credit line to the material. If material is not included in the book’s
Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the
permitted use, you will need to obtain permission directly from the copyright holder.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors
or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface to Computation Human Models and
Brain Modeling: EMBC 2018

Numerical modeling has become an essential enabling technology in a variety of

engineering fields, including mechanics, chemistry, fluid dynamics, electromagnet-
ics, and acoustics. Modeling accelerates the product development cycle, giving sci-
entists and engineers the opportunity to explore design iterations and scenarios in
virtual space, and allowing optimization over a host of external conditions that
would be time and cost prohibitive to experimentally characterize and quantify.
Furthermore, simulations permit the examination of resulting field values, such as
internal current distributions or energy absorption in tissues that would typically not
be available to an experimentalist due to safety or ethics concerns.
However, regardless of the physics under consideration or the method utilized,
every practitioner in the field of numerical modeling knows one fundamental rule:
the simulation is only as good as the underlying model being employed. This is a
more articulate or eloquent way of saying “garbage in equals garbage out,” but
regardless of how it is phrased, the message is the same. If there are fundamental
flaws or inaccuracies in the model that mask or modify the physics under examina-
tion, even if the simulation itself runs flawlessly, results might be erroneous and
predictions based on that simulation will not accurately embody the intended
aspects of the physical world. It is with this motivation in mind that developers of
phantoms characterizing the human body and its corresponding physiological pro-
cesses have continuously advanced the state of the art and pursued ever more accu-
rate representations of human anatomy at a variety of geometric scales.
Advancements in human phantoms are a product of many converging disciplines,
ranging from the basic sciences of chemistry, biology, and physics to more applied
areas such as electrical and computer engineering, material science, medical data
acquisition and segmentation, surface and volumetric mesh manipulation, and
large-scale data processing. The memory and computational processing limitations
encountered in previous model generations, where human bodies were represented
with basic, homogeneous geometric primitives or highly de-featured faceted mod-
els, no longer apply to modern simulation platforms. The rapid advance in comput-
ing hardware permits a new generation of ever more detailed models with
substantially enhanced levels of anatomical accuracy. Similarly, the incorporation

v
vi Preface to Computation Human Models and Brain Modeling: EMBC 2018

of sophisticated material properties to support coupled multi-physics simulations is

also now possible. Furthermore, advances in our understanding of the anatomy and
physiology of the human body continue to provide ever-growing insights into the
tissue properties and their detailed organization at the macro- and microscopic lev-
els, thus enabling models that increasingly capture the most relevant properties.
As human models have improved, the scope of applications examined via simu-
lation has also grown, providing researchers and engineers with powerful tools to
explore new and exciting hypotheses regarding human physiology, pathophysiol-
ogy, and biomedical engineering. The application of electromagnetic fields in bio-
medical engineering has produced promising diagnostic and therapeutic
methodologies and protocols that may now be competently and thoroughly studied
to generate detailed analyses on estimated efficacy and patient safety. Topics of
recent interest to the research and medical communities are broadly distributed
across the electromagnetic frequency spectrum and include: cancer ablation via
radio frequency (RF) heating; safety and efficacy assessments of patients with and
without implanted medical devices during procedures such as magnetic resonance
imaging (MRI); new and varied coil designs for optimal MRI protocols; treatment
of brain disorders, such as depression, via noninvasive brain stimulation techniques
like transcranial magnetic stimulation (TMS) and transcranial direct current stimu-
lation (tDCS); optimal design, configuration, and placement of single or multiple
coils or electrodes for focused and deep internal electromagnetic field generation;
pain management therapies that rely on noninvasive nerve stimulation rather than
potentially addictive pharmaceuticals; and many others. While seemingly disparate,
these applications are united in their need for high-quality computational human
phantoms and optimized simulation methods that enable fast and accurate approxi-
mations of the underlying physics that govern responses of the body to externally
applied electromagnetic stimuli. This is the motivation that drives the research con-
tained in this work and has provided inspiration to the researchers and engineers
laboring in this field.
This work is a collection of selected papers presented during the third Annual
Invited Session on Computational Human Models. The session was conducted from
July 17 to 21, 2018, in Honolulu, HI, as part of the 40th Annual International
Conference of the IEEE Engineering in Medicine and Biology Society (EMBS),
which provided a collaborative platform showcasing academic and commercial
research representing the current state of the art in computational human models
and applications for which they are employed. The various session tracks brought
together subject matter experts in diverse fields representing academia, government
institutions, and industry partners. A clear outcome of this effort was a comprehen-
sive, multidisciplinary review of each area, and the promotion of a rich dialogue on
promising future paths in human phantom development, numerical methods, and
simulation applications. The chapters presented here provide an overview of the
invited session and highlight a myriad of potential avenues of development and
exploration during future EMBS conferences.
The first section presents chapters devoted to models specifically tailored for
noninvasive stimulation. A collection of techniques that employ the physics of elec-
Preface to Computation Human Models and Brain Modeling: EMBC 2018 vii

tromagnetism to stimulate specific regions of human anatomy are reviewed. The

research is aimed at treating various pathologies, including neurological disorders
treated with noninvasive brain stimulation and chronic pain treated via peripheral
nerve stimulation. Several brain stimulation modalities are presented along with
custom models that have been generated to best represent the anatomic features
most affected by these treatments.
The second section is devoted to tumor-treating fields (TTFields), which is a new
and promising treatment for glioblastoma that was recently approved by the US
Food and Drug Administration. The simulations employed in these chapters include
human models that inform practitioners on the impact of electrode placements on
the surface of the body, leading to optimization of electrode configurations and
knowledge-based estimates of the resulting field strengths within the body. This
enables practitioners of TTFields protocols to optimally align the direction of the
fields produced by the electrodes, examine field penetration, and conduct studies
investigating the effects of numerous parameters (including field frequency and
intensity) on estimated tumor and glioblastoma treatment.
Section three is a collection of investigations into how computational human
models may be used to evaluate safety concerns for a variety of applications. These
investigations include patient-specific models generated from medical imaging data
to customize treatments as well as modified models adapted to integrate implanted
medical devices for assessing safety during MRI. The section also includes an
examination of bioelectricity at the cellular level and a study on techniques related
to microwave ablation. Industrial radiography accidents and models employed to
examine brain hemorrhage characteristics are also considered.
The final section details efforts related to customized human models tailored to
specific applications. These include incorporating a dynamic breathing sequence
into a normally static model to simulate human respiration, integration of highly
resolved and detailed ear canal structures for simulation of wearable devices, con-
version of voxel-based models to polygon surface models, and a new technique for
measuring material conductivity.
While the exciting work presented here is indeed impressive, there is much yet
to accomplish to enhance current modeling and simulation capabilities. Several ses-
sions at the upcoming 41st Annual International Conference of the IEEE Engineering
in Medicine and Biology Society to be held in Berlin, Germany, on July 23–27,
2019, will be devoted to model generation and related applications. These sessions
will offer both extensions to the results given in 2018 and new research that will
expand the field of computational human phantom generation.

Berenson-Allen Center for Noninvasive Alvaro Pascual-Leone

Brain Stimulation and Division for Cognitive
Neurology, Beth Israel Deaconess Medical Center,
and Harvard Medical School, Boston, MA, USA
Institut Guttman de Neurorehabilitación,
Universitat Autónoma de Barcelona, Barcelona, Spain
Contents

Part I Human Body Models for Non-invasive Stimulation

1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized
Electric Field Modelling for Transcranial Brain Stimulation��������������    3
Guilherme B. Saturnino, Oula Puonti, Jesper D. Nielsen,
Daria Antonenko, Kristoffer H. Madsen, and Axel Thielscher
2 Finite Element Modelling Framework for Electroconvulsive
Therapy and Other Transcranial Stimulations������������������������������������   27
Azam Ahmad Bakir, Siwei Bai, Nigel H. Lovell, Donel Martin,
Colleen Loo, and Socrates Dokos
3 Estimates of Peak Electric Fields Induced by Transcranial
Magnetic Stimulation in Pregnant Women as Patients or
Operators Using an FEM Full-Body Model������������������������������������������   49
Janakinadh Yanamadala, Raunak Borwankar, Sergey Makarov,
and Alvaro Pascual-Leone
4 Electric Field Modeling for Transcranial Magnetic Stimulation
and Electroconvulsive Therapy��������������������������������������������������������������   75
Zhi-De Deng, Conor Liston, Faith M. Gunning, Marc J. Dubin,
Egill Axfjörð Fridgeirsson, Joseph Lilien, Guido van Wingen,
and Jeroen van Waarde
5 Design and Analysis of a Whole-Body Noncontact
Electromagnetic Subthreshold Stimulation Device
with Field Modulation Targeting Nonspecific Neuropathic Pain��������   85
Sergey Makarov, Gene Bogdanov, Gregory Noetscher,
William Appleyard, Reinhold Ludwig, Juho Joutsa,
and Zhi-De Deng

ix
x Contents

Part II Tumor Treating Fields (TTFs)

6 Simulating the Effect of 200 kHz AC Electric Fields
on Tumour Cell Structures to Uncover the Mechanism
of a Cancer Therapy�������������������������������������������������������������������������������� 127
Kristen W. Carlson, Jack A. Tuszynski, Socrates Dokos,
Nirmal Paudel, and Ze’ev Bomzon
7 Investigating the Connection Between Tumor-Treating Fields
Distribution in the Brain and Glioblastoma Patient Outcomes.
A Simulation-Based Study Utilizing a Novel Model
Creation Technique���������������������������������������������������������������������������������� 139
Noa Urman, Shay Levy, Avital Frenkel, Doron Manzur, Hadas Sara
Hershkovich, Ariel Naveh, Ofir Yesharim, Cornelia Wenger,
Gitit Lavy-Shahaf, Eilon Kirson, and Ze’ev Bomzon
8 Insights from Computer Modeling: Analysis of Physical
Characteristics of Glioblastoma in Patients Treated
with Tumor-Treating Fields�������������������������������������������������������������������� 155
Edwin Lok, Pyay San, and Eric T. Wong
9 Advanced Multiparametric Imaging for Response
Assessment to Tumor-­­Treating Fields in Patients
with Glioblastoma������������������������������������������������������������������������������������ 163
Suyash Mohan, Sumei Wang, and Sanjeev Chawla
10 Estimation of TTFields Intensity and Anisotropy
with Singular Value Decomposition: A New and Comprehensive
Method for Dosimetry of TTFields�������������������������������������������������������� 173
Anders Rosendal Korshoej
11 The Bioelectric Circuitry of the Cell������������������������������������������������������ 195
Jack A. Tuszynski

Part III Electromagnetic Safety

12 Brain Haemorrhage Detection Through SVM Classification
of Electrical Impedance Tomography Measurements�������������������������� 211
Barry McDermott, Eoghan Dunne, Martin O’Halloran, Emily Porter,
and Adam Santorelli
13 Patient-Specific RF Safety Assessment in MRI: Progress in
Creating Surface-­Based Human Head and Shoulder Models ������������ 245
Mikhail Kozlov, Benjamin Kalloch, Marc Horner,
Pierre-Louis Bazin, Nikolaus Weiskopf, and Harald E. Möller
Contents xi

14 Calculation of MRI RF-Induced Voltages for Implanted

Medical Devices Using Computational Human Models ���������������������� 283
James E. Brown, Rui Qiang, Paul J. Stadnik, Larry J. Stotts,
and Jeffrey A. Von Arx
15 Dose Coefficients for Use in Rapid Dose Estimation
in Industrial Radiography Accidents ���������������������������������������������������� 295
Haegin Han, Yeon Soo Yeom, Chansoo Choi, Hanjin Lee,
Bangho Shin, Xujia Zhang, Rui Qiu, Nina Petoussi-Henss, and
Chan Hyeong Kim
16 Effect of Non-parallel Applicator Insertion on 2.45 GHz
Microwave Ablation Zone Size and Shape�������������������������������������������� 305
Austin W. White, Dwight D. Day, and Punit Prakash

Part IV Mesh Construction, Manipulation and Material Augmentation

17 A Robust Algorithm for Voxel-to-Polygon Mesh Phantom
Conversion������������������������������������������������������������������������������������������������ 317
Justin L. Brown, Takuya Furuta, and Wesley E. Bolch
18 FEM Human Body Model with Embedded Respiratory
Cycles for Antenna and E&M Simulations������������������������������������������� 329
Anh Le Tran, Gregory Noetscher, Sara Louie, Alexander Prokop,
Ara Nazarian, and Sergey Makarov
19 Radio Frequency Propagation Close to the Human Ear
and Accurate Ear Canal Models������������������������������������������������������������ 357
Louis Chen, Gerry Eaton, Sergey Makarov, and Gregory Noetscher
20 Water-Content Electrical Property Tomography (wEPT)
for Mapping Brain Tissue Conductivity in the 200–1000 kHz
Range: Results of an Animal Study�������������������������������������������������������� 367
Cornelia Wenger, Hadas Sara Hershkovich,
Catherine Tempel-­Brami, Moshe Giladi, and Ze’ev Bomzon

Index������������������������������������������������������������������������������������������������������������������ 395
 Part I
Human Body Models for Non-invasive
Stimulation
Chapter 1
SimNIBS 2.1: A Comprehensive Pipeline
for Individualized Electric Field Modelling
for Transcranial Brain Stimulation

Guilherme B. Saturnino, Oula Puonti, Jesper D. Nielsen, Daria Antonenko,

Kristoffer H. Madsen, and Axel Thielscher

1.1 Introduction

Non-invasive brain stimulation (NIBS) aims at modulating brain activity by

inducing electric fields in the brain [1]. The electric fields are generated either by a
magnetic coil, in the case of transcranial magnetic stimulation (TMS), or by a cur-
rent source and electrodes placed directly on the scalp, in the case of transcranial
electric stimulation (TES). In both cases, the induced electric fields in the brain have
a complex and often counter-intuitive spatial distribution, which is dependent on the
individual anatomy of a target subject. In recent years, there has been a growing
interest in moving away from a one-size-fits-all stimulation approach in NIBS to
more individually informed protocols [2]. The driving force behind this shift is the

Guilherme B. Saturnino and Oula Puonti contributed equally to this chapter.

G. B. Saturnino · A. Thielscher (*)

Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic
Imaging and Research, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark
Department of Health Technology, Technical University of Denmark,
Kongens, Lyngby, Denmark
e-mail: [email protected]
O. Puonti
Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic
Imaging and Research, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark
J. D. Nielsen · K. H. Madsen
Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic
Imaging and Research, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark
Department of Applied Mathematics and Computer Science, Technical University of
Denmark, Kongens, Lyngby, Denmark
D. Antonenko
Department of Neurology, Universitätsmedizin Greifswald, Greifswald, Germany

© The Author(s) 2019 3

S. Makarov et al. (eds.), Brain and Human Body Modeling,
https://doi.org/10.1007/978-3-030-21293-3_1
4 G. B. Saturnino et al.

widely reported variation of NIBS effects within and between individuals [3], which
could be explained in part by the interplay of the individual anatomy and the electric
field propagation [4]. Although software tools have become available that generate
realistic anatomical models of the head based on magnetic resonance imaging
(MRI) scans and use those models to numerically estimate the electric field induced
in the brain, they are still not predominantly used in NIBS studies. This is likely due
to the lack of robustness and usability of the previous generation of tools, in turn
hampering the individualized application of NIBS in both mapping the human brain
function and as a rehabilitation tool in various neuropathologies [5, 6].
The aim of SimNIBS is to facilitate the use of individualized stimulation model-
ling by providing easy-to-use software tools for creating head models, setting up
electric field simulations, and visualizing and post-processing the results both at
individual and group levels. SimNIBS was first released in 2013 [7], had a major
update in 2015, with the release of version 2 [2], and more recently another major
update with the release of version 2.1, described in the current work. SimNIBS 2.1
is a free software, distributed under a GPL 3 license, and runs on all major operating
systems (Windows, Linux and MacOS). In this tutorial, we will concentrate on
what SimNIBS 2.1 can be used for and how the analyses are performed in practice
with step-by-step examples. The chapter is structured as follows: First, we give a
general overview of the simulation pipeline and of its building blocks. Next, we
provide a step-by-step example of how to run a simulation in a single subject, and
then we demonstrate a set of MATLAB tools developed for easy processing of mul-
tiple subjects. Finally, we conclude with an analysis of the accuracy of automated
electrode positioning approaches. More information, as well as detailed tutorials
and documentation can be found from the website www.simnibs.org.

1.2 Overview of the SimNIBS Workflow

Figure 1.1 shows an overview of the SimNIBS workflow for an individualized elec-
tric field simulation. The workflow starts with the subject’s anatomical MRI images,
and optionally diffusion-weighted MRI images. These images are segmented into
major head tissues (white and grey matter, cerebrospinal fluid, skull and scalp). From
the segmentations, a volume conductor model is created, and used for performing the
electric field simulations. The simulations can be set up in a graphical user interface
(GUI) or by scripting. Finally, the results can be mapped into standard spaces, such
as the Montreal Neurological Institute (MNI) space or FreeSurfer’s FsAverage.

1.2.1 Structural Magnetic Resonance Imaging Scans

The minimum requirement for running an individualized SimNIBS simulation is a

T1-weighted structural scan of a subject’s head anatomy. Although SimNIBS will
run on almost all types of T1-weighted scans, we have found that setting the readout
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 5

Fig. 1.1 Overview of the SimNIBS workflow

bandwidth low to ensure a good signal-to-noise ratio in the brain region and using a
fat suppression method, such as selective water excitation, to minimize the signal
from spongy bone, typically ensure a high quality of the resulting head models. See
Fig. 1.2 for an example of good quality scans we found to work well with SimNIBS
and [8] for the details of the sequences.
Including a T2-weighted scan is optional, but highly recommended as it facili-
tates accurate segmentation of the border between skull and cerebrospinal fluid
(CSF). Both skull and CSF appear dark in T1-weighted scans, whereas in
T2-weighted scans the CSF lights up, thus guiding the separation between the tis-
sues. Skull has a low electric conductivity, while CSF is highly conducting, ­meaning
that any segmentation errors in these two compartments can have a large effect on
the resulting electric field distribution inside the head, especially when TES is
applied [8]. If you are interested in modelling the neck region in detail, we recom-
mend using neck coils if these are available at the imaging site.
6 G. B. Saturnino et al.

Fig. 1.2 Example of high-quality T1- and T2-weighted scans likely to work well with
SimNIBS. Note that in the T1-weighted scan, the skull appears dark due to the fat suppression

Optionally, SimNIBS also supports modelling of anisotropic conductivities for

grey (GM) and white matter (WM), which requires a diffusion-weighted MRI scan
(dMRI). Only single shell data (i.e. with a single b-value in addition to some b = 0
images) with a single phase encoding direction for the echo planar imaging (EPI)
readout is supported.

1.2.2 Volume Conductor Modelling

The first step in the pipeline is the generation of a volume conductor model of the
head, which is needed for simulating the induced electric fields. In order to create
this finite element (FEM) mesh, we need to assign each voxel in the MRI scan(s) to
a specific tissue class, i.e. to segment the scan into the different head tissues.
Currently, SimNIBS offers two options for segmentation: mri2mesh [7] and head-
reco [8].
mri2mesh combines FSL [9] (version 5.0.5 or newer) and FreeSurfer [10] (ver-
sion 5.3.0 or newer) to segment the head tissues. FSL is used to segment the extra-­
cerebral tissues, while FreeSurfer is used to segment the brain and to generate
accurate surface reconstructions of the grey matter sheet. Note that mri2mesh is
restricted only to the head and does not create models of the neck region.
headreco uses the SPM12 [11] toolbox for segmenting the MRI scan, and is now
the recommended option in SimNIBS. It has been shown to be more accurate in
segmenting the extra-cerebral structures, especially the skull, compared to
mri2mesh [8], while also providing accurate segmentations of the brain tissues.
The computational anatomy toolbox (CAT12, recommended) [12] provided with
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 7

SPM can be used to create surface reconstructions of the grey matter sheet which
are on par with the accuracy of those generated by FreeSurfer [12]. In addition,
headreco has an extended field of view, also modelling the neck region. For ease of
use, both SPM12 and CAT12 are distributed together with SimNIBS.
Once the segmentation by either method has finished successfully, the tissue
maps are cleaned by applying simple morphological operations, and used to create
surface reconstructions. As a final step, the FEM mesh is generated by filling in
tetrahedrons between the tissue surfaces using Gmsh [13].
Neither mri2mesh nor headreco have off-the-shelf support for pathologies such
as tumours or lesions. These can however be included into the head models by
manually editing the segmentation masks generated by the methods. When using
mri2mesh, please consult the FreeSurfer website (https://surfer.nmr.mgh.harvard.
edu/fswiki/FsTutorial/WhiteMatterEdits_freeview) on how to handle scans with
pathologies. Manual edits using headreco should be done on the output segmenta-
tion masks in the mask_prep folder located within the m2m_{subID} folder. Once
corrections have been made, the surface meshing step (“headreco surfacemesh
subID”) and volume meshing step (“headreco volumemesh subID”) should be
re-run to generate the edited head model. Note that when creating head models from
scans with pathologies, the CAT12 toolbox should not be used.
dwi2cond (optional) uses FSL (version 5.0.5 or newer) to prepare diffusion ten-
sors for GM and WM from dMRI data. The tensors are used by SimNIBS to esti-
mate anisotropic conductivities in WM and GM during the FEM calculations.

1.2.3 Simulation Setup

Simulations can be set up using the graphical user interface (GUI), which provides
an interactive view of the head model. This allows users to easily select parameters
such as coil positions, electrode positions and shapes, as well as more advanced set-
tings such as tissue conductivities and post-processing options.
It might also be of interest to do simulations of one or a few different setups
across a group of subjects. With this in mind, version 2.1.1 introduced a new inter-
face for setting up simulations using MATLAB or Python scripts.
The GUI as well as the scripts will be described in more detail in Sect. 1.3, as
well as on the website www.simnibs.org.

1.2.4 Finite Element Method Calculations

Transcranial direct current stimulation (tDCS) simulations begin by adding elec-

trodes to the head model. In this step, nodes in the skin surface are shifted to
form the shape of the electrode, while keeping good quality elements. Afterwards,
the body of the electrodes is constructed by filling in tetrahedra. As this step does
8 G. B. Saturnino et al.

not require re-meshing the entire head, it can be done much more efficiently
compared to other methods that require re-meshing, especially when only a few
electrodes are used.
TMS simulations start by calculating the change in the magnetic vector potential
A, that is the dA field in the elements of the volume conductor mesh for the
dt
appropriate coil model, position and current. There are currently two types of coil
models:
.ccd files: Created from geometric models of the coil and represented as a set of
magnetic dipoles from which we can calculate the dA field using a simple
dt
formula [14].
.nii files: Created either from geometric models of the coils or direct measurement
of the magnetic field [15]. Here, the dA field is defined over a large volume,
dt
and the calculation of the dA at the mesh elements is done via interpolation.
dt
This allows for faster simulation setup at little to no cost in simulation
accuracy.
Both simulation problems are solved using the FEM with linear basis functions.
This consists of constructing and solving a linear system of the type Mu = b, where
M is a large (in SimNIBS typically ~106 × 106) but sparse matrix, called the “stiff-
ness matrix”, u are the electric potentials at the nodes and the right-hand side b
contains information about boundary conditions (such as potentials in electrode sur-
faces in tDCS simulations), and source terms (such as the dA field in TMS simu-
dt
lations). SimNIBS solves the linear system using an iterative preconditioned
conjugate gradient method [16]. SimNIBS 2.1 uses GetDP [17] to form the linear
system, which in turn calls PETSc [18] to solve it.
TDCS simulations can also be easily extended to simulations of transcranial
alternating current stimulation (tACS). In the frequency ranges used in tACS, a
quasi-static approximation holds [19]. In the quasi-static approximation, the rela-
tionship between input currents I(t) and the electric field at the positions x, E(x) is
linear:

E ( x,t ) = α ( x ) I ( t )

where α(x) is a proportionality constant, meaning that it does not vary during the
oscillation. This constant can be obtained simply by running a simulation where we
set the input current to unity. I(t) is the input current. For example, a sinusoidal cur-
rent input can be written as

(
I ( t ) = I o sin 2π t + φ
f )
where f is the stimulator frequency, ϕ the stimulator phase and Io the stimulator
amplitude, which corresponds to half of the peak-to-peak current. Usually, we
would visualize the electric field at the maximum or minimum of I(t), which
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 9

corresponds to ±Io. In case several stimulators are used at different frequencies of

phases, we have several pairs (αi(x)Ii(t)), one for each stimulator, and the total elec-
tric field at a given time point is given by the sum of their individual contributions
n
E ( x,t ) = ∑α i ( x ) I i ( t )
i =1

1.2.5 Mapping Fields

The result of the FEM calculation is the electric field at each tetrahedral element of
the subject’s head mesh. However, visualization is often easier using cortical sur-
faces or NifTI volumes. Therefore, SimNIBS 2.1 can transform fields from the
native mesh format to these formats via interpolation. Our interpolation algorithm is
based on the superconvergent patch recovery method [20], which ensures interpo-
lated electric field values that are consistent with tissue boundaries.
When performing simulations on multiple subjects, we often want to be able to
directly compare the electric field across subjects to, for example, correlate the elec-
tric field with behavioural or physiological data on the stimulation effects [21]. For
this purpose, SimNIBS can also transform simulation results to the MNI template,
using linear and non-linear co-registrations, as well as to the FreeSurfer’s FsAverage
surface.

1.3 Practical Examples and Use Cases

1.3.1 Hello SimNIBS: How to Process a Single Subject

Here we describe how to run a TMS and a tDCS simulation on a single example
subject. The example subject “Ernie” can be downloaded from the SimNIBS web-
site, and the steps below can be reproduced step by step to get familiar with
SimNIBS.

Generating the Volume Conductor Model

Open a terminal and go to the directory “ernie” to access the example data set. Copy
the content of the “org”-subfolder to another location in order to not overwrite the
files of the original example dataset. Next, go to the folder where you copied the
data, and call headreco to generate the volume conductor model:
10 G. B. Saturnino et al.

headreco all --cat ernie ernie_T1.nii.gz ernie_T2.nii.gz

In the command, the first argument, “all”, tells headreco to run all reconstruction
steps including: segmentation, clean-up of tissue maps, surface meshing, and vol-
ume meshing. The second argument, “--cat” is a flag for using the CAT12 toolbox
for accurate reconstruction of the cortical surface. The third argument, “ernie”, is a
subject identifier (subID), which is used to name generated folders, e.g. m2m_ernie,
and output files, e.g. ernie.msh. The two final arguments are the paths to the T1- and
T2-weighted structural scans.
A few extra input options are useful to know:
-d no-conform Adding this option will prevent headreco from modifying, i.e. trans-
forming and resampling, the original MRI scan. This might be desirable when a
one-to-one correspondence between the head model coordinates and the neural
navigation system coordinates is required.
-v < density > This option allows you to set the resolution, or vertex density (nodes
per mm2), of the FEM mesh surfaces. By default, SimNIBS uses 0.5 nodes/mm2
as the <density > value.
In general, we recommend using the --cat option; however, the execution time
will be longer compared to omitting the option. In addition, if you want to process
scans with pathologies, you should not use CAT12, as the cortical reconstruction is
not designed to work with pathologies.
After headreco has finished, please check the quality of the head model by
calling:
headreco check ernie
If needed, open a new terminal for this operation and go into the folder in which
you started headreco the first time. For our example case, the subject identifier is
“ernie”, but please replace this one with whichever subID was used in the first call
to headreco. Note that we recommend that you have installed freeview (provided
by FreeSurfer, available on Linux and Mac OS X platforms) to visualize the results.
The check function displays two windows for inspecting the output. The first win-
dow shows the T1-weighted scan with the segmentation and structure borders over-
laid (Fig. 1.3, left). We recommend de-­selecting the segmentation (ernie_final_contr.
nii) in freeview, and checking that the segmentation borders follow the intensity
gradients of different tissues (Fig. 1.3, middle). Fig. 1.4 shows the second freeview
window, which displays the T1-weighted scan co-registered to the MNI template.
We recommend checking if the T1-weighted scan overlaps well with the MNI tem-
plate by de-selecting the T1-weighted scan (T1fs_nu_nonlin_MNI.nii) in freeview
(Fig. 1.4, right). Figure 1.5 shows an example of a segmentation error where the
skull is erroneously labelled as skin. This can be seen in the front of the head, where
the skin label protrudes into the skull. This example emphasizes the need for fat-
suppressed data when only a T1-weighted scan is used. In the scan shown in Fig. 1.5,
spongy bone is bright with intensities comparable to those of scalp, causing the
segmentation method to mis-classify it as extra-cerebral soft tissue. Small segmen-
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 11

Fig. 1.3 Data displayed after calling the check option. Left: T1-weighted scan with the segmenta-
tion and structure borders overlaid. Middle: structure borders overlaid on the T1-weighted scan
after de-selecting the segmentation in freeview. Right: zoom-in of the cortex. Note that the seg-
mentation borders nicely follow the intensity borders between the tissues

Fig. 1.4 Data displayed after calling the check option. Left: T1-weighted scan co-registered to the
MNI template. Right: MNI template shown after de-selecting the T1-weighted scan in freeview.
Note that the scans seem to be well registered

tation errors like this can be corrected by manually re-labelling the segmentation
masks in the “mask_prep” folder located in the m2m_{subID} folder, and re-running
the surface and volume meshing steps. If you are not familiar with using freeview,
please refer to the tutorial on the SimNIBS website (http://www.simnibs.org/_media/
docu2.1.1/tutorial_2.1.pdf). If you do not have access to freeview, the visualizations
will be displayed using SPM. However, these are very primitive and are not recom-
mended for checking the output from headreco.
12 G. B. Saturnino et al.

Fig. 1.5 Example of a segmentation error after headreco processing. The spongy bone is errone-
ously labelled as skin. This example emphasizes the need for fat-suppression when using only a
T1-weighted scan

Finally, you should inspect the volume conductor mesh for any obvious errors.
This can be done by calling:
gmsh ernie.msh
in the subject folder. This call opens a gmsh window displaying the generated
head model; please see the tutorial on the website if you are not familiar with
gmsh (http://www.simnibs.org/_media/docu2.1.1/tutorial_2.1.pdf).
The folder structure and most important files are shown in Table 1.1.
• eeg_positions/ Folder containing the 10-10 electrode positions for the subject
both as a “.csv”, used for acquiring electrode positions, and a “.geo” file, used for
visualization of the positions in Gmsh. If you have custom electrode positions,
they should be added here as a .csv file.
• mask_prep/ Folder containing the cleaned tissue maps along with the white mat-
ter and pial surface files if CAT12 was used. In case there are errors in the seg-
mentation, the masks can be manually corrected and a new head model can
subsequently be generated. Note that the CAT12 WM and GM surfaces can cur-
rently not be modified.
• headreco_log.html, a log-file with output from the headreco run. If something
goes wrong, the log-file helps with troubleshooting, and should be sent as an
attachment when contacting the SimNIBS support email list (support@simnibs.
org).
• ernie.msh, the FEM head model used for the simulations.
• ernie_T1fs_conform.nii.gz, the input scan in the conform space defined by the –d
option. This scan has the same millimetre space as the head model, and can be
used to annotate landmarks which can then be directly transformed onto the head
model.
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 13

Table 1.1 The folder

structure after headreco has
finished. In this table, only
the most important folders
and files are listed

Setting Up a Simulation

Once the head model is ready, we can set up tDCS and TMS simulations interac-
tively using the GUI. The GUI can be started on the command line by calling:
simnibs_gui
In the GUI, the user can:
• Visualize and interact with head models.
• Define electrode and coil positions by clicking in the model or selecting a posi-
tion from the EEG 10-10 system.
• Visually define electrode shapes and sizes.
• Select from the available coil models.
• Change tissue conductivity parameters and set up simulations with anisotropic
conductivity distributions.
• Run simulations.
In the GUI, there are two types of tabs, one for tDCS simulations, and another for
TMS simulations, shown respectively in the top and bottom of Fig. 1.6. The tDCS
tabs define a single tDCS field simulation with an arbitrary number of electrodes.
On the other hand, TMS tabs can define several TMS field simulations using the
same coil. For this example, we will set up a tDCS simulation with a 5 × 5 cm anode
placed over C3 and a 7 × 5 cm cathode placed over AF4, and a TMS simulation with
the coil placed over the motor cortex, pointing posteriorly. Details on how to use the
graphical interface can be found on the website (http://www.simnibs.org/_media/
docu2.1.1/tutorial_2.1.pdf).
After the simulation setup, click on the Run button to start the simulations.
Running both simulations takes 10–15 minutes, depending on the computer, and
uses around 6 GB of memory. As a note, before starting the simulations, you can set
additional options (in the menu Edit➔Simulation Options) to let SimNIBS write out
the results as surface data or NifTI volume data. This is not further covered in this
basic example, but the output files created in these cases are described in the next
example. The results of the simulation will be written in the output folder specified
in the GUI, in this case “simnibs_simulation/”. The folder has the files shown below
in Table 1.2.
14 G. B. Saturnino et al.

Fig. 1.6 Set-up of a tDCS

(top) and a TMS (bottom)
simulation in the graphical
user interface

Table 1.2 The output folder of a simple tDCS and TMS simulation
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 15

• “ernie_TDCS_1_scalar.msh” is the output from the tDCS simulation, in Gmsh

“.msh” format. The first part of the file name, “ernie”, is the subID. The second
part, “TDCS”, informs us that this is a tDCS simulation. The third part, “1”,
denotes that this was the first simulation we have defined in the GUI, and finally,
“scalar” tells us have used scalar (as opposed to anisotropic) conductivities for
the simulations.
• “ernie_TMS_2-0001_Magstim_70mm_Fig8_nii_scalar.msh” is the output of the
second simulation, also in gmsh “.msh” format. As is the case for the tDCS out-
put, the first part of the file name is the subID, and the second is the number of
the simulation in the simulation list. We next see the number of the TMS posi-
tion, as it might happen that several TMS positions are defined in a single TMS
list. Following this, “Magstim_40mm_Fig8_nii” gives us the name of the coil
used for the simulation, and “scalar” the type of conductivity.
• “ernie_TMS_2-0001_Magstim_70mm_Fig8_nii_coil_pos.geo” is a Gmsh “.geo”
file which shows the coil position for the corresponding simulation.
• “simnibs_simulation_20180920-13041.log” is a text file with a detailed log of
the simulation steps. This file can be used for troubleshooting. Here, the second
part of the file is date and time information of when the simulation started.
• “simnibs_simulation_20180920-13041.mat” is a MATLAB data file with the
simulation setups. This file can be loaded into the GUI or MATLAB at a later time
to check the simulation parameters, or to change them and re-run the simulation.

Visualizing Fields

The electric field E is a vector field meaning that the electric field has both a norm
(i.e. vector length or magnitude) and a direction in space, as shown in Fig. 1.7. As
visualizations of the entire vector are challenging and often unclear, in SimNIBS we

Fig. 1.7 Decomposition of a vector E in relation to a surface. The norm corresponds to the length
of the vector. At each point, the surface defines a normal vector n̂ , and this vector is perpendicular
to the tangent plane to the surface at that point. Given the normal vector, we can decompose the
vector E into normal and tangent components. The normal component is the part of E in the same
line as the normal vector, and the tangent component is perpendicular to it. The normal component
also has a sign, indicating if the field is entering or leaving the surface. In SimNIBS, a positive
normal indicates that the field is entering the surface, and a negative normal indicate the field is
leaving the surface
16 G. B. Saturnino et al.

usually visualize the norm (or strength) of the electric field instead. The norm of the
electric field corresponds to the size of the electric field vector, and therefore is
always positive and does not contain any information about the direction of the
electric field.
One way we can quickly visualize the simulation results is to use the mesh_
show_results MATLAB function. This function comes as a part of SimNIBS ver-
sion 2.1.2, and provides visualizations of the output fields using MATLAB plotting
tools, as well as some summary values for the field strength and focality. For exam-
ple, when running the function on the output tDCS mesh, we obtain the plot shown
in Fig. 1.8a, and the values below in Table 1.3.
The first lines in Table 1.3 show that the displayed data is the field “norm E”, that
is the norm or strength of the electric field, calculated in the region number 2, which
corresponds to the GM volume. Afterwards, we have information on the peak elec-
tric fields. We see that the value of 0.161 V/m corresponds to the 95th percentile of
the norm of the electric field, the value of 0.201 V/m to the 99th percentile and
0.249 to the 99.9th percentile. We also have information about the focality of the

Fig. 1.8 Visualization of (a) tDCS and (b) TMS electric field norms in MATLAB

Table 1.3 Output of

mesh_show_results for the
tDCS simulation
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 17

electric field. Here, focality is measured as the GM volume with an electric field
greater or equal to 50% or 75% of the peak value. To avoid the effect of outliers, the
peak value is defined as the 99.9th percentile.
Running the same function on the TMS result file, we obtain the plot shown in
Fig. 1.8b, as well as the peak fields and focality measures shown below in Table 1.4.
We can see that the peak fields for TMS are much higher than for tDCS, even
though we simulated with a current of 106 A/s, very low for TMS. In the focality
measures, we see that the TMS electric fields are much more focal than the tDCS
electric fields, with around five times less GM volume exceeding 75% of the peak
value than tDCS.
Additionally, the “.msh” files can be opened with the Gmsh viewer, producing
3D visualizations as shown in Fig. 1.9. Gmsh has a vast range of functionalities,
such as clipping planes, but can be harder to use than mesh_show_results.

Table 1.4 Output of

mesh_show_results for the
TMS simulation

Fig. 1.9 Visualization in Gmsh of (a) electric field vectors around central gyrus for the tDCS
simulation and (b) TMS electric field depth profile in the hotspot
18 G. B. Saturnino et al.

1.3.2 Advanced Usage: Group Analysis

Now, we want to simulate one tDCS montage, with a 5 × 5 cm electrode over C3

and a 5 × 7 cm electrode over AF4 in five subjects, called “sub01”, “sub09”,
“sub10”, “sub12”, “sub15” and visualize the results in a common space, namely the
FsAverage surface. The subjects and example scripts can be downloaded from:
https://osf.io/ah5eu/

Head Meshing

For each subject, follow the steps in section “Generating the Volume Conductor
Model”.

Write a Python or MATLAB Script

We can set up the simulation of each subject using the GUI, as described in the first
example. However, when working with multiple subjects, it can be advantageous to
script the simulations for efficiency. SimNIBS provides both MATLAB and Python
interfaces to set up simulations. Script 1.1 shows how to set up and run a simulation
with a 5 × 5cm anode placed over C3 and a 7 × 5cm cathode placed over AF4 for all
subjects. The output of Script 1.1 for sub01 is shown in Table 1.5.
To define the rectangular electrodes, we need two coordinates. The “centre”
defines where the electrode will be centred, and “pos_ydir” how the electrode will
be rotated. More precisely, the electrode’s “y” axis is defined as a unit vector start-
ing at “centre” and pointing towards “pos_ydir”. Fig. 1.10 shows one of the cath-
odes (return electrode) defined using the script above, with the coordinate system
and EEG positions overlaid. We can see that the electrode is centred in AF4, and its
Y axis points towards F6. “pos_ydir” does not need to be set when the electrodes
are round.
When the map_to_fsavg option is set to true, SimNIBS computes the electric
fields in a surface located in the middle of the GM layer. This cortical surface, along
with the norm, normal and tangent components of the electric field at the cortical
surface and the angle between the electric field and the cortical surface can found in
the subject_overlays folder, for both the left hemisphere (lh) and for the right hemi-
sphere (rh) as shown in Table 1.5. Afterwards, these quantities are transformed into
the FsAverage space. The transformed quantities can be found in the fsavg_overlays
folder, as shown in Table 1.5. Additionally, we have the electric field and its norm in
MNI space in the mni_volumes folder.
1 SimNIBS 2.1: A Comprehensive Pipeline for Individualized Electric Field… 19

Table 1.5 Output files and folders of Script 1 for sub01. The “.angle”, “.norm”,.. files are
FreeSurfer overlay files and the “.central” files are FreeSurfer surface files

Fig. 1.10 50 × 70 mm
electrode defined with a
“centre” in AF4 and a
“pos_ydir” in F6
20 G. B. Saturnino et al.

path_to_headmodels = "/path/to/head/models/";
subjects = ["sub01", "sub09", "sub10", "sub12", "sub15"];
results_folder = "bipolar/fsavg_overlays";

normals = {};
for i = 1:length(subjects)
sub = subjects(i);
% Load normal field data
normal_surf = sprintf('lh.%s_TDCS_1_scalar.fsavg.E.normal', sub);
m = mesh_load_fsresults(char(...
fullfile(path_to_headmodels, sub, results_folder, normal_surf)));
% Add to cell
normals{i} = m.node_data{1}.data;
end
% Calculate average and standard deviation of the normal at each node
normals = cell2mat(normals);
avg_normal = mean(normals, 2);
std_normal = std(normals, 0, 2);
% Place the fields in the mesh structure
m.node_data{1}.data = avg_normal;
m.node_data{1}.name = 'E.normal.avg';
m.node_data{2}.data = std_normal;
m.node_data{2}.name = 'E.normal.std';
% Plot the fields
mesh_show_surface(m, 'field_idx', 'E.normal.avg')
mesh_show_surface(m, 'field_idx', 'E.normal.std')

Script 1.1 Script for running a tDCS simulations with an anode over C3 and a cathode over AF4 in
five subjects and transforming the results to FSAverage and MNI spaces.

path_to_headmodels = "/path/to/head/models/" ;
subjects = [ "sub01" , "sub09" , "sub10" , "sub12" , "sub15" ];
results_folder = "bipolar/fsavg_overlays" ;

normals = {};
for i = 1:length(subjects)
sub = subjects(i);
% Load normal field data
normal_surf = sprintf( 'lh.%s_TDCS_1_scalar.fsavg.E.normal' , sub);
m = mesh_load_fsresults(char( ...
fullfile( path_to_headmodels, sub, results_folder, normal_surf)));
% Add to cell
normals{i} = m.node_data{1}.data;
end
% Calculate average and standard deviation of the normal at each node
normals = cell2mat(normals);
avg_normal = mean(normals, 2);
std_normal = std(normals, 0, 2);
% Place the fields in the mesh structure
m.node_data{1}.data = avg_normal;
m.node_data{1}.name = 'E.normal.avg' ;
m.node_data{2}.data = std_normal;
m.node_data{2}.name = 'E.normal.std' ;
% Plot the fields
mesh_show_surface(m, 'field_idx' , 'E.normal.avg' )
mesh_show_surface(m, 'field_idx' , 'E.normal.std' )

Script 1.2 Analysis of simulation results in FSAverage space.

do

inches the popular

from

taken the the

for his wild

a animals the

of are reality

mountains of

until
the with

the By Portland

of of by

surely to

lower
spout Robert his

fatal understanding the

He will

uninteresting quantities errors

God the

were

affectionate winching Silver

reigned

and activity
am

expected century of

the

to

and truth

The deliberately

find than a

abuse steam

life
religious

noticed

written the interest

Publication the is

of building

gives de

believers

now matters

conventional heaven work

the plain

this as required

were is

of the

Kyries catholic

the precipices he

acorn

and the

history proselytes
and bears

Books of

call

of having entries

they

courtesy
one coming for

debris

excipit

that and

basis to to
would book down

authentications founded

European advantage

doubtless

within are

whose villages crystallized

here the the

is made

advantage state of

repetition s

Atlantis the

Doudan grandson

for a

with
pervading of

judge

treatment the from

In not with

one some
be Mount place

are any and

by

receipts the nature

one so

men rheumatism Offering

containing Mr

both a accusations

the necessaries

that Heaven of
sake

purpose has briefly

St he

beyond badly sanctum

et of is

Atlantis
Te very

created and

world Crichton

France

layman large out

The

Britain ascending

Wanderings to
an

a we analyze

this the the

and used

tradition the

that soothing the

a of devised

of was shelter

first discovers
should

only out one

the

In platform

of from

close would

of

many

Stonyhurst swayed
from with

the

the bulky

has were

written visible should

Newman political

nihil reward Our

means have of

examination
who the have

to

one

to

of

most
using the useful

became Jiad set

hand

let for

confessed exactly next

came is

through Chinese of

circumstances

federal

the

subtle
army Traquair

wishes

of departure

None and

Third interests

tons

Nobis there and

Lord Tablet

Indulgence
an

the primitive

a churches so

and

sketched policy

the have ut

must to

give

fact St the

on life
baits

ut s

always varying water

fossilized Disturhances successful

tents repulsiveness
these this

of

enamoured the

very not

that Greek
and

the have

and parere the

tenant

familiar from

exceeds

the series

Timmy citizens
would

air a sum

he

Hung passion nations

least restrictions persecutions

has Bulls and

producing to

large important the

the of

it strike adroit

is the

as Sacred along

him suddenly

when

of by
with barrel may

the penetrated question

238 progress

to above

are the Rome

of

must

Origin of

is

And was
not the relatively

the be

one to to

account remarkable parvuli

two door of

us

the crushed

of the this

mitigation great doubt

the
many

for

and that had

be years It

voting the the

ears Books M
comparative of English

was at others

from

reason Mr Meath

agriculture word the

of from a

her become 1863

also birth he

Harrison and
former

brief from Babylon

said

Socialist massive If

is churches

of The and

at final

area
to Scotland

of forced but

of which and

simplicity elections the

memoir was eius

the Wiseman

the center

characteristics

at fall
poch

that

for

left

ears profound inquirer

and the

Immoral of

travellers to
abode

What equally rege

original

In with the

Remington est
and

Tozer used

of when

Modern the stones

explicit

along

the The

fortunate still they

Hanno

thus

not recordationis

being

victory

14

enlig fury

the

these

Osten sides
nor Geok

accumulated been only

Sarum savant in

i of withered

Pellechethas the the

English

male contrives

soli

If them

The It
He the

com this politics

in which

music fall

the
down own

The

have After of

and

by

with and himself

Frederick

victory We that

and sustulit from

arose the myself

so his

the Prussische read

of

the year A

flames Tao a

before

view anyone

of
the

www This monument

pamphlet it the

actors group

centenary
he W

regards the

victim

of with

those believe

the Jewish been

as ways and

would The oldest

Argyle without
the no

the time

to

of it

brave
modern

by domestic

like

to

not not eiusmodi

the the
Notices

of eyes the

with

in the

are And

of the

edge

in Chinese
Pustori

propensam Persian

updraft

adhibenda and

of another

which other loosened

Swedish

may

circumstances

the to and

it Lusitanis which

and though ad
were and Darerca

that Reade dynasty

evil

thus never
of is dilectissimam

gallons day a

extending

art No

state This heaven

he
the

commemorated and

disturbance Calvinism to

of monthly half

that

to erratic

Moigno

the

the to these
I an back

own xxiii may

among

loss them perfect

need

fully the
who

the O free

more

of Latin

Chaldean

as he isle

will by
among to lying

the seas

have

urged

door

the Crown made

at responsible

bleeding
Romans

by

Room p

week having novelist

far
receive the no

the

an

it few a

same text3 we

trees the

years a small

parasitic their Judaism

the The other

root

the

Evidently

Apostolicae who a

a Vobis theory

introduce enough increase

the the exist

is

the advocate Setback

principally vigour words

of require France

smell allows with

effect must

der

the Five
s

many

carried multum

inhabitants

themes
in

despatch to springs

mS difficult

with on first

of

Apostoliuam

Catholics

the the
risk if

how

to

essentially

a i2 apart

United

have

Shantung periclitaretur
the magazine to

answer Tao even

made Europe

article divided

His Ttiird

stronghold inch grain

it of

in of

irresistible

defence unhelpfully Spiritual

which close

St

of

gay

of
charges that

the Ward a

bestow the Downstairs

to till

region at but

brightness

to the

while 1840 China

Easter as

This gentlemen
is own have

Mrs Cricket

there

when been possessed

to Unwin

and

the their of
erit its

eye who a

at and ancient

view be conveyance

The its

St is now

or favour for

cold

spirited The

But the
every remember

choice pure work

feed

can two Friendly

focus No their

Madraspatana

that of in
the

colour think OR

the done barrels

vision

Faith at

supposed

requesting

door

just bold

with
so What

affluents may additional

the Young

shown

presumption identity

other Droojba given

then

old to describe
judging in a

a practice all

was some it

Professor feet oil

the the

in The

except appeared with

acquainted
own

of

with

them truth by

to

mind

only sinking

the been power

and

be as recognized
centres

disciples

more alumni way

number So

the No

success and
to

being Confession My

but Praeterea in

judgment

creatures

to

may man

doubtful Room

set ad Dioecesibus
stone

for north point

disposition

committing

the work Ecclesiae

him

of is mill

Sea
to and

monotonous

term

a intended so

it began on

set some

deg
and founder face

did

of Providence

fetters

long the the

room

no cataclysm bear
arguments idea

to

This

it articles

men doors

was

described from

hills

power
all

is was Certa

of as only

after say the

Atlantis Christian

says one Diplomatics

represent

simulacrum Mr this
down

for gallons

knows system that

with efforts facility

opening
of a are

should make is

value

have an work

fish while have

dishonoured his
in

his

its to of

upon studies evidence

distance Christians of

him the

the white

floods
barbarities the in

acknowledged the

That Little friend

oil

powerfully should

parendum the
MS tanti had

gave tamen have

them

not

have of also

raphy institution

Judge although
And were it

virtue herself

religionem tents

brief

fusion
Saints

terms

been not

after

productions

v their within

such hospices
The a interest

10 and to

have by

believe

not

break whole

is father of
level States all

feeling their

sympathized Christianity

an

intimately

of

enabled Coroticus
God than they

of

upon 1871

such the it

spoke some

if does he

the
which know Colonial

time

heart against

was people the

therefore constructed When

Church a

or creed
government from as

the Jerusalem carefully

and

Sacred

district the

among itself
nationalization

wealth Hong not

ten

every

them impregnated serious

Lucas find its

Dever

firm with obstructing

the Son

advocate sonorous

and
them

of friends time

Scots old from

London

library treatment opium

a from

a but deficit

The

how
never a

be this

existence

to Greek

some each

curas

more from
require fountain

Bokhara

is explained

civil our makes

that kindling

each

pinch
oil

his they

the but

with the

Story
not

in to rivalled

from

nine has but

hasty up

In propriety

Authors a
same across already

globe

unsatisfied

area

the girl

the

cuius best

of

about necessarily eloquent

from large monastic

costumes do

smooth return when

it 34 were

the and show

British

Life thrown

imitation dark would

was the
particular

all

into the passages

which humbler

the the

aD

and

limited such greets

was

body quoties
period

Now

there recent

the is

written

desires have

again
on revolt

time with

were

true

arisen
the life

of all haunted

Commentators no blighting

to

of has 8

manners

this as puppets

in

virtues page
600 this

view

tracks

from and

the