Conference PaperPDF Available

Construction and segmentation of pediatric head tissue atlases for electrical head modeling

Authors:
Construction and segmentation of pediatric head tissue atlases for
electrical head modeling
Presented During: Poster Session
Monday, June 26, 2017: 12:45 PM - 02:45 PM
Presented During: Poster Session
Monday, June 26, 2017: 12:45 PM - 02:45 PM
Stand-By Time
Monday, June 26, 2017: 12:45 PM - 2:45 PM
Submission No:
1648
Submission Type:
Abstract Submission
On Display:
Monday, June 26 & Tuesday, June 27
Authors:
David Hammond1, Nick Price2, Sergei Turovets2
Institutions:
1Oregon Institute of Technology - Wilsonville, Wilsonville, OR, 2Electrical Geodesics, Inc, Eugene, OR
First Author:
David Hammond - Lecture Information | Contact Me
Oregon Institute of Technology - Wilsonville
Wilsonville, OR
Introduction:
Accurate population models of head tissue geometry and conductivity are essential for accurate source localization in
electroencephalography (EEG) and precise targeting in transcranial electrical stimulation (TES). The work described
here is part of a larger effort to build a comprehensive set of age-specific pediatric head models, based on merging
multiple imaging modalities (structural MRI, CT, diffusion-weighted MRI, Electrical Impedance Tomography (EIT)) to
enable creation of atlases of soft tissues, skull mineralization density, and anisotropic electrical conductivity of white
matter (supported by NIH grant R44 MH106421).
Here we describe our construction of average T1 MRI atlas images, and compare two approaches for segmentation.
Using images obtained from the NIH Pediatric MR Database (Evans 2006, Almli et al 2007), we calculated a series of
atlases in 4 month age increments from 0 to 1 year, from 1 to 2 years, then in 2 year age increments (2-4 years, 4-6
years, etc.), from ages 2 to 18.
Our ultimate goal being the production of head conductivity models, we are especially interested in accurate
segmentation of these atlases; electrical head models (i.e. lead field matrices) can then be computed based on
assigning electrical conductivities for different tissue types consistent with known values from the literature (e.g.
Salman 2005). We use BrainK (Li et al, 2016), which computes a segmentation based on a single MRI image, using
internal skull templates for skull segmentation. For an atlas computed from a collection of individual images, such a
tool can be used in two distinct ways: a) directly segmenting the average atlas ("direct atlas segmentation"), or b)
segmenting each individual MRI image, and then aggregating the individual segmentations ("aggregate atlas
segmentation").
Methods:
Our average MRI atlases were based on a set of 736 T1-weighted MRI images obtained from the Pediatric MRI Data
Repository, created as part of the NIH MRI Study of Normal Brain Development
(https://pediatricmri.nih.gov/nihpd/info/index.html). For each age group, we computed atlas images by iterating the
following : 1) Given a current template, compute transformations (from a set of transformations allowed at this
iteration) registering each individual image onto the the template; 2) Average these transformed images to give the
template for the next iteration. We organized our calculations in stages, the first allowing only rigid rotations, the
second only affine transformations, and the third allowing nonlinear transformations. All transformations were
computed using the ANTs software (Tustison et al, 2014), using symmetric normalization (SyN) based on the
neighborhood cross correlation (CC) metric.
As an example, for the 12-14 age group (63 original images) we computed the aggregate segmentation by
segmenting the original images, transforming these segmentations into the atlas space using the same nonlinear
transformations computed during the atlas construction, then aggregating them into a single "maximum probability"
aggregate segmentation by "voting": i.e. by choosing at each voxel the tissue type that was present most often
amongst the individual transformed segmentations.
Results:
Axial slices of selected atlases are shown in Fig 1. Fig 2 shows axial slices of the aggregate atlas segmentation and
of the direct atlas segmentation, and two representative individual segmentations (out of 63 total) mapped onto the
atlas space.
·Figure 1, Selected average atlas images
·Figure 2, Selected individual segmentation images, and atlas segmentation images (aggregate and direct)
Conclusions:
The direct atlas segmentation and the aggregate atlas segmentation are much smoother than the individual
segmentations. Both are quite similar, and give reasonable segmented images. For the ultimate purpose of computing
electrical head models, the probability maps (i.e. images with probabilities of each tissue type) can be used to
interpolate tissue conductivities in a manner consistent with the inferred uncertainty in the tissue segmentation, an
advantage not available from the direct segmentation method.
Imaging Methods:
Anatomical MRI
Informatics:
Brain Atlases 1
Modeling and Analysis Methods:
Image Registration and Computational Anatomy
Segmentation and Parcellation 2
Poster Session:
Poster Session - Monday
Keywords:
Atlasing
MRI
Segmentation
STRUCTURAL MRI
1|2Indicates the priority used for review
Would you accept an oral presentation if your abstract is selected for an oral session?
Yes
I would be willing to discuss my abstract with members of the press should my abstract be marked
newsworthy:
Yes
Please indicate below if your study was a "resting state" or "task-activation” study.
Other
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to
distribute the presentation in any format, including video, audio print and electronic text through OHBM
OnDemand, social media channels or other electronic media and on the OHBM website.
I accept
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Healthy subjects
Internal Review Board (IRB) or Animal Use and Care Committee (AUCC) Approval. Please indicate approval
below. Please note: Failure to have IRB or AUCC approval, if applicable will lead to automatic rejection of
abstract.
Not applicable
Please indicate which methods were used in your research:
Structural MRI
For human MRI, what field strength scanner do you use?
1.5T
Which processing packages did you use for your study?
Other, Please list - ANTs, BrainK
Provide references in author date format
C.R. Almli et al, "The NIH MRI study of normal brain development
(Objective-2): Newborns, infants, toddlers, and preschoolers",
NeuroImage, Volume 35, Issue 1, March 2007, Pages 308-325, ISSN
1053-8119, http://dx.doi.org/10.1016/j.neuroimage.2006.08.058
A.C. Evans (2006). "The NIH MRI study of normal brain development". NeuroImage 30 (2006) 184-202
Kai Li et al, "BrainK for Structural Image Processing: Creating Electrical Models of the Human Head", Computational
Intelligence and Neuroscience, vol. 2016, Article ID 1349851, 25 pages, 2016. doi:10.1155/2016/1349851
N. J. Tustison et al, "Large-scale evaluation of ANTs and FreeSurfer cortical thickness measurements", NeuroImage,
Volume 99, 1 October 2014, Pages 166-179, ISSN 1053-8119, http://dx.doi.org/10.1016/j.neuroimage.2014.05.044.
A. Salman et al (2005), "Computational modeling of human head conductivity", Proceedings of
the 5th International Conference on Computational Science, pp 631-638

Supplementary resource (1)

Conference Paper
Electrical impedance tomography (EIT) is a medical imaging modality that has the potential to benefit diagnosing, monitoring, and understanding several pathological conditions. However, some regions of the body, such as the brain, are more challenging to reconstruct, demanding improvements before the technique can be used in clinical practice. In this study, we implemented and evaluated an algorithm for 2D static EIT image reconstruction based on the Deep Image Prior (DIP) method. The method was tested in measurements calculated from a computational human head model, where we included a region representing the occurrence of a stroke. The results showed that the DIP-based algorithm had some advantages compared to a more classical method, such as robustness to noise and independence of an initial solution. Therefore, this method could be better suited for real-life EIT image reconstructions.
Conference Paper
Time-Difference Electrical Impedance Tomography (TDEIT) is an imaging technique to visualize resistivity changes over time in a region of interest. Regularization is necessary because TDEIT is an ill-posed problem. In this work, we use Regularization by Denoising (RED) with four different denoisers to reconstruct brain images in a simplified 2D head model. We compared the RED results to two traditional reconstruction methods, generalized Tikhonov regularization and total variation regularization. In both the noiseless and the noisy scenarios, we achieved the best results using RED with non-local means as the denoiser in relation to figures of merit such as ringing, resolution and shape deformation.
Chapter
Continuous monitoring of brain hemodynamics is important to quickly detect changes in healthy cerebral blood flow, helping physician decision-making in the treatment of the patient. Resistivity changes in the brain happen as a result of the pulsatile characteristic of the blood in the arteries or pathological conditions such as ischemia. We developed a dynamic model of cerebral circulation capable of portraying variations in resistivities in arteries within a cardiac cycle. From the hypothesis that the resistivity changes in the brain can be detected by Electrical Impedance Tomography (EIT), we included this model as prior information in time-difference image reconstruction algorithm. With this prior information, image reconstruction of the brain with pre-existing ischemia was possible, showing that EIT is a potential technique for brain hemodynamic monitoring.KeywordsBlood flow modelElectrical impedance tomographyDifference imagingStroke
Chapter
Electrical impedance tomography (EIT) is a technique that can be used to estimate resistivity distribution from the inside of a domain based on surface measurements. This could be useful, for example, in the diagnosis of cerebral strokes. However, a method to acquire EIT images of the head with enough quality to achieve this task is still needed. In this work, an automated method for the calculation of a statistical atlas of the human head is presented to be used as prior information for the ill-posed inverse problem associated to EIT. Fifty magnetic resonance images of healthy subjects were used for this purpose. Numerical simulations using a realistic head model with hemorrhagic and ischemic stroke were used to evaluate its effect. The results show that, when the atlas was used, there was a decrease in the root mean square error of the images obtained. Also, some artifacts observed in the image generated without the use of the atlas were eliminated or diminished. These findings hint to the possibility of using a statistical atlas of the head to improve the quality of EIT images.KeywordsAnatomical atlasElectrical impedance tomographyImage processingStroke
Article
Full-text available
BrainK is a set of automated procedures for characterizing the tissues of the human head from MRI, CT, and photogrammetry images. The tissue segmentation and cortical surface extraction support the primary goal of modeling the propagation of electrical currents through head tissues with a finite difference model (FDM) or finite element model (FEM) created from the BrainK geometries. The electrical head model is necessary for accurate source localization of dense array electroencephalographic (dEEG) measures from head surface electrodes. It is also necessary for accurate targeting of cerebral structures with transcranial current injection from those surface electrodes. BrainK must achieve five major tasks: image segmentation, registration of the MRI, CT, and sensor photogrammetry images, cortical surface reconstruction, dipole tessellation of the cortical surface, and Talairach transformation. We describe the approach to each task, and we compare the accuracies for the key tasks of tissue segmentation and cortical surface extraction in relation to existing research tools (FreeSurfer, FSL, SPM, and BrainVisa). BrainK achieves good accuracy with minimal or no user intervention, it deals well with poor quality MR images and tissue abnormalities, and it provides improved computational efficiency over existing research packages.
Conference Paper
Full-text available
The computational environment for estimation of unknown regional electrical conductivities of the human head, based on reali stic geometry from seg- mented MRI up to2563 resolution, is described. A finite difference alternating d i- rection implicit (ADI) algorithm, parallelized using OpenMP, is used to solve the forward problem describing the electrical field distributi on throughout the head given known electrical sources. A simplex search in the multi-dimensional para- meter space of tissue conductivities is conducted in parall el using a distributed system of heterogeneous computational resources. The theoretical and computa- tional formulation of the problem is presented. Results from test studies are pro- vided, comparing retrieved conductivities to known solutions from simulation. Performance statistics are also given showing both the scaling of the forward problem and the performance dynamics of the distributed search.
Article
Many studies of the human brain have explored the relationship between cortical thickness and cognition, phenotype, or disease. Due to the subjectivity and time requirements in manual measurement of cortical thickness, scientists have relied on robust software tools for automation which facilitate the testing and refinement of neuroscientific hypotheses. The most widely used tool for cortical thickness studies is the publicly available, surface-based FreeSurfer package. Critical to the adoption of such tools is a demonstration of their reproducibility, validity, and the documentation of specific implementations that are robust across large, diverse imaging datasets. To this end, we have developed the automated, volume-based Advanced Normalization Tools (ANTs) cortical thickness pipeline comprising well-vetted components such as SyGN (multivariate template construction), SyN (image registration), N4 (bias correction), Atropos (n-tissue segmentation), and DiReCT (cortical thickness estimation). In this work, we have conducted the largest evaluation of automated cortical thickness measures in publicly available data, comparing FreeSurfer and ANTs measures computed on 1205 images from four open data sets (IXI, MMRR, NKI, and OASIS), with parcellation based on the recently proposed Desikan-Killiany-Tourville (DKT) cortical labeling protocol. We found good scan-rescan repeatability with both FreeSurfer and ANTs measures. Given that such assessments of precision do not necessarily reflect accuracy or ability to make statistical inferences, we further tested the neurobiological validity of these approaches by evaluating thickness-based prediction of age and gender. ANTs is shown to have a higher predictive performance than FreeSurfer for both of these measures. In promotion of open science, we make all of our scripts, data, and results publicly available which complements the use of open image data sets and the open source availability of the proposed ANTs cortical thickness pipeline.
Article
MRI is increasingly used to study normal and abnormal brain development, but we lack a clear understanding of "normal". Previous studies have been limited by small samples, narrow age ranges and few behavioral measures. This multi-center project conducted epidemiologically based recruitment of a large, demographically balanced sample across a wide age range, using strict exclusion factors and comprehensive clinical/behavioral measures. A mixed cross-sectional and longitudinal design was used to create a MRI/clinical/behavioral database from approximately 500 children aged 7 days to 18 years to be shared with researchers and the clinical medicine community. Using a uniform acquisition protocol, data were collected at six Pediatric Study Centers and consolidated at a Data Coordinating Center. All data were transferred via a web-network into a MYSQL database that allowed (i) secure data transfer, (ii) automated MRI segmentation, (iii) correlation of neuroanatomical and clinical/behavioral variables as 3D statistical maps and (iv) remote interrogation and 3D viewing of database content. A population-based epidemiologic sampling strategy minimizes bias and enhances generalizability of the results. Target accrual tables reflect the demographics of the U.S. population (2000 Census data). Enrolled subjects underwent a standardized protocol to characterize neurobehavioral and pubertal status. All subjects underwent multi-spectral structural MRI. In a subset, we acquired T1/T2 relaxometry, diffusion tensor imaging, single-voxel proton spectroscopy and spectroscopic imaging. In the first of three cycles, successful structural MRI data were acquired in 392 subjects aged 4:6-18:3 years and in 72 subjects aged 7 days to 4:6 years. We describe the methodologies of MRI data acquisition and analysis, using illustrative results. This database will provide a basis for characterizing healthy brain maturation in relationship to behavior and serve as a source of control data for studies of childhood disorders. All data described here will be available to the scientific community from July, 2006.
Article
The Magn. Reson. Imaging (MRI) study of normal brain development currently conducted by the Brain Development Cooperative Group represents the most extensive MRI study of brain and behavioral development from birth through young adulthood ever conducted. This multi-center project, sponsored by four Institutes of the National Institutes of Health, uses a combined longitudinal and cross-sectional design to characterize normal, healthy brain and behavioral development. Children, ages newborn through 18-plus years of age, receive comprehensive behavioral, neurological and multimodal MRI evaluations via Objective-2 (birth through 4-years 5-months of age) and Objective-1 (4-years 6-months through 18 years of age and older). This report presents methods (e.g., neurobehavioral assessment, brain scan) and representative preliminary results (e.g., growth, behavior, brain development) for children from newborn through 4-years 5-months of age. To date, 75 participants from birth through 4-years 5-months have been successfully brain scanned during natural sleep (i.e., without sedation); most with multiple longitudinal scans (i.e., 45 children completing at least three scans, 22 completing four or more scans). Results from this younger age range will increase our knowledge and understanding of healthy brain and neurobehavioral development throughout an important, dynamic, and rapid growth period within the human life span; determine developmental associations among measures of brain, other physical characteristics, and behavior; and facilitate the development of automated, quantitative MR image analyses for neonates, infants and young children. The correlated brain MRI and neurobehavioral database will be released for use by the research and clinical communities at a future date.