A Comparative Study of Brain Connectivity Visualization and Clustering Techniques

International Journal of Computer Application
Issue 4, Volume 1 (January-February 2014)

Available online on http://www.rspublication.com/ijca/ijca_index.htm
ISSN: 2250-1797
_______________________
A Comparative Study of Brain Connectivity Visualization
and Clustering Techniques
Jenish Lavji#1, Bhumika Shah#2
Sarvajanik
college
of engg.
technology,
Sarvajanik
college
of engg.
AndAnd
technology,
Surat,Surat,
India India
ABSTRACT
Present a novel approach of brain connectivity visualization of white matter pathways
and gray matter areas using diffusion tensor imaging and functional magnetic resonance
imaging. Clustering is an important tool for creating abstraction and filtration in general and is
central representation in particular. While abstraction involves simplification and
generalization, filtration here entails clustering. To visualization of brain volumetric using
stream tube, stream surface, streamlines or variation of streamlines of nerve bundles of fibers
pathways of brain. Different visualization techniques present here that is weighted graph
comparison method, three dimensional neural maps represent in web-interface stand-alone
approach similar to Google maps that shown with associated names, measurements and link
outs of brain neural maps of different regions like corpus callosum, left cingulum and arcuate
and a small portion of right arcuate, both cortico-spinal tracts and inferior frontooccipital
fasciculi. Different clustering algorithm that is K-means, Spectral Clustering, Normalized Cut
Expectation maximization and fuzzy c-means apply on nerve fiber cluster.
Key words: Brain connectivity,Diffusion tensor Imaging,Visualization,Clustering.
Corresponding Author: Jenish Lavji
INTRODUCTION
Brains structure is very quiet and complex in nature. The whole brain comprise of
thousand and ten thousand number of cortex neurons. The entire central core part is known as
white matter and cortical surface of brain is called gray matter [7] as shown in fig 1. Brain
connectivity comprise of two things: one related to white matter and another related to gray
matter. Brain connectivity refers to different aspects of brain organization including anatomical
connectivity consisting of axonal fibers across cortical region and functional connectivity
defined as the observed statistical correlation between regions of interest. White matter fiber
tracts so called anatomical connectivity derived through apply tractography algorithm to
diffusion tensor imaging data which represents an isotropic diffusion of water through bundles of
neural axons. For gray matter function connectivity is derived from functional magnetic
resonance imaging. It depends upon the blood oxygen consumption level; functional connectivity
can be seen as relatedness of specific region of interest [1].
Diffusion tensor imaging is a non-invansive tool to extract the features of white matter
pathways by using magneto resonance tractography algorithm apply on nerve fiber bundles.
Graphical representation of magneto resonance tractography is well suited to the issue of
figuring out white matter structure. It suggests conceivable association between end points of
pathways from one place to another in brain [7]. For the visualization of tractography nerve
bundles using stream tube, stream surface, line jittering, hyper stream lines or variation of stream
lines [10]. Similarity of fiber cluster can be seen using different color coding scheme. The
_______________________________
R S. Publication (http://rspublication.com), rspublicationhouse@gmail.com
Page 186

ISSN: 2250-1797
_______________________
complex structure of tractography data sets has motivated earlier work to apply clustering
techniques to tractograms [9, 10, 12]. The success of tract clustering is often determined by the
degree to which the similarity measure used in the process can captured anatomical features that
are the main area of interest to a specific user. Clustering is an important tool for creating
abstraction and filtration in general and is central representation in particular [3].
Fig1. The two dissection view of the brain that is one is side view (left) and other is top view
(right) [6, 7].
BRAIN CONNECTIVITY VISUALIZATION AND CLUSTERING TECHNIQUES
A.
Visualization Techniques
Main theme of fiber trajectories is to visualize by using stream tube, stream surface and
stream lines or variation of stream lines. Stream tube represents linear dissemination which
has tendency to transform by fibrous structure. The cross segment and shade of stream tube
represent extra part of dissemination along trajectories of fiber tracts. Stream surface represent
structure in which dissemination is principally planer [10]. Different visualize method
mentioned below.
1.
Visualize using Graph Method
The analysis of brain connectivity tasks is a vast field in recent study; increase the use
of visual representation using visual analysis tool. Until now, no comparison of graph
techniques is to be shown; for two techniques: node-links and matrices. For visualization,
matrices perform this tasks well and node link performs outperforms [1].
2.
Visualize using fMRI guided Tractography
Diffusion tensor imaging (DTI) and functional magnetic resonance imaging (fMRI)
investigate two different aspects of brain networks: white matter anatomical connectivity and
gray matter functional connectivity. Individual fMRI driven tractography is usually applied
and only few studies address on group analysis. fMRI driven tractography gives more
effective result by apply to the group study and creation of tractography atlas on gray matter
_______________________________
Page 187

ISSN: 2250-1797
_______________________
areas. fMRI guided tractography atlas gives more accurate and precise result compared to
probabilistic atlas [2].
3.
Visualize using Neural Maps
Create a standard stream tube model that display the diffusion weighted cerebrum
imaging information plus neural path representation into the plane. To visualize planer neural
maps gives more visual clarity and simplicity tract of interest. A stand alone visualization
light weighted web-interface approach; representation in such a way that precomputed
pathway of brain can be similar way of representation of Google maps with particular with
associated names, measurement, link outs [3].
4.
Visualize of Brain in Three dimensional view
Disentangle and analyze neural pathway estimate from magnetic resonance imaging
data, scientist need an interface three dimensional pathways. For that use of pen and mice that
gives only two degree of opportunity. For that CINCH solve that problem to provide bimanual
interface employ pen and trackball and marking language to select three dimensional
pathways [6].
5.
Visualize using Diffusion tensor imaging and Magneto resonance
Diffusion tensor imaging (DTI) plus Magneto resonance (MR) tractography techniques
used for desirable properties of white matter pathways. For that precomputed pathways and its
statistical properties (length, fractional anisotropy, average curvature path) to extract the
desirable feature of white matter. It saves some time by precomputing pathways. For selecting
pathway use of query language and Boolean operation like AND & OR [7].
6.
Visualize using Stream tube and Stream surface
A new method for visualizing three dimensional volumetric diffusion tensor magneto
resonances images; distinguish between linear anisotropy (stream tube model) and planar
anisotropy (stream surface model). An expert studying the white matter after gamma
capsulotomy and preoperative making arrangement for cerebrum tumour surgery indicates
that stream tube associate well suite with major neural structures, the two dimensional area
and geometric historic points are vital in comprehension the visualization [10].
7.
Visualize using Diffusion tensor imaging tractography
The state of the art of reconstruction of the axonal tracts in the central nervous system
(CNS) using diffusion tensor imaging (DTI) is reviewed. This moderately new system has
created much excitement and high desires on the grounds that it in a matter of seconds is the
main approach accessible to non-intrusively study the three-dimensional structural planning
of white matter tracts. While there is no doubt that DTI fiber tracking is providing exciting
new opportunities to study CNS anatomy. This technique can be used only for macroscopic
analysis of white matter architecture, but not to address connectivity questions at the cellular
_______________________________
Page 188

ISSN: 2250-1797
_______________________
level. DTI tractography can indeed delineate the core of large white matter tracts as judged
from the encouraging results from initial validation studies. At present, there are no other noninvasive techniques that can provide equivalent information and, as a consequence, DTI
tractography is expected to be a powerful technique to investigate white matter anatomy and
diseases [11].
B.
Clustering Techniques
Clustering is an important tool for creating abstractions and filtrations [3]. A clustering
method that propagates cluster labels from fiber to neighbouring fiber. It assigns each
unlabelled fiber to the cluster of its closest neighbour, if the closest neighbour is below a
threshold. A partition of the data with a specific number of clusters can be acquired by setting
a threshold on the maximal accepted distance [13, 14]. Use a spectral embedding technique
called Laplacian Eigen maps in which the high dimensional fibers are reduced to points in a
low dimensional Euclidean space and these positions are mapped into a RGB color space,
such a way that similar fibers are assigned similar colors [15]. A clustering method based on
normalized cuts is used to group fibers [9].
In the event that fibers are remade and envisioned independently through the complete
white matter, the showcase gets effortlessly jumbled making it troublesome to get
understanding in the information. Various clustering techniques have been proposed to
automatically obtain bundles that should represent anatomical structures. Found that the use
of hierarchical clustering using single-link and a fiber similarity measure based on the mean
distance between fibers gave the best results [8]. Different cluster methods mentioned below.
1.
Clustering via K-Means and Spectral Clustering
The utilization of information driven bunching techniques for utilitarian connectivity

examination in fMRI. K-Means and Spectral Clustering calculations as plan to the ordinarily
utilized Seed-Based Analysis. K-Means (KM) Clustering Algorithm is based on minimizing
the Euclidean distance in such a way that to maximizing correlation. Spectral Clustering (SC)
utilizes the Eigen deterioration of a couple shrewd partiality grid built from information
focuses. SC can identify cluster with complex signal geometries. Seed based Correlation
investigation recognizes voxel associated with interim course in a client specified seed locale,
yet it requires priori learning. Independent component analysis isolates spatial sources
account for activity variation across the brain & can be used to delineate different functional
network [4].
2.
Clustering using Expectation Maximization Algorithm
A novel based approach for joint clustering and point by point correspondence.
Knowledge of point by point correspondence gives accurate and precise clustering and also
gives tract-oriented quantitative analysis. Employ an expectation maximization algorithm in
gamma mixture model with parameter i.e. spatial mean, variance and standard deviation.
Point by point correspondence obtained by constructing distance map and label map for each
_______________________________
Page 189

ISSN: 2250-1797
_______________________
iteration of expectation maximization algorithm. Result gives more effectiveness in terms of
time using expectation maximization algorithm applies on fiber bundles of white matter
pathways of brain [5].
3.
Clustering using Normalized Cut Criterion
A structure for unsupervised division of white matter strand follow got from
dissemination weighted MRI information. Fiber follow are contrasted pair wise with make a
weighted undirected chart which is divided into rational sets utilizing the standardized cut
basis. A straightforward then again successful strategy for pair wise examination of fiber
follow is displayed which in blend with the Cut measure is demonstrated to generate
conceivable divisions of both engineered and genuine fiber follow information. Divisions are
pictured as hued stream-tubes or changed to a division of voxel space, uncovering structures
in a manner that looks guaranteeing for future explorative investigations of dispersion
weighted MRI information. Determining Fiber similarities: Split the computation of similarity
into following steps: 1) Mapping fiber traces to the Euclidean Feature space that preserve
some information, but not information about fiber shape and connectivity. 2) Gaussian kernel
for comparison of points in the Euclidean feature space. 3) Combining mapping to a
Euclidean feature space with Gaussian kernel [9].
4.
Clustering using Fuzzy C-Means Algorithm
Diffusion Tensor tracking provides line trajectory data representing neural fiber
pathways throughout the brain. It is difficult to display and analyze the thousands of resulting
tracks. An automated clustering approach was devised to segment the data into major
component pathways. The fuzzy c-means (FCM) clustering algorithm was used, which
provides probabilistic values for cluster membership based on distance measures between
tracks. Several distance metrics were tested that included functions of distance between tracks
and shape information. Using this algorithm several hundred human subcortical tracks were
classified into pathways with high accuracy, compared with manual selection using spatial
volumes. Diffusion tensor MRI has provided data with sufficient signal to noise ratio and
resolution to perform accurate tracking of nerve fiber bundles in the brain. These algorithms
can produce thousands of tracks using whole brain data. Automatically segmenting this large
amount of data into more manageable clusters of tracks that represent the known major fiber
bundles in the brain would be extremely useful for future applications of this technique, both
for research and clinical purposes. An automatic technique would not be biased in patients
with brain lesions or distorted anatomy. The FCM algorithm is a robust clustering algorithm
that has been used in numerous engineering and medical applications. The FCM clustering
algorithm can be applied to DT-MRI tracks and can produce reliable groupings of tracks [12]
_______________________________
Page 190

ISSN: 2250-1797
_______________________
C.
Comparison of Visualization Techniques
Table-1. Comparison of Visualization Techniques

Paper Title
Visualization
Techniques
Methods
Tasks
Selection
Type
Application
Weighted Graph
Comparison
Techniques for Brain
Connectivity Analysis
Weighted
graph
Node-link
and Matrices
Analysis of
Brain
Connectivity
Task
Region
Of
Interest
Biological
Pathway
N/w, Social
N/w,
Commercial
Trade N/w,
Air Traffic
N/w.
A novel approach of
fMRI-guided
tractography analysis
within a group:
construction of an
fMRI-guided
tractographic atlas.
fMRI driven
Tractography
and
Probabilistic
Tractography
DTI and
fMRI used
for Gray
Matter and
White matter
Group
Analysis of
task carried
out with 11
healthy
subjects
Region
of
Interest
Construction
of 3D
trajectory
and White
Matter
detection.
Exploring Brain
connectivity with two dimensional Neural
maps
Stream tube
Quantitative
and
Anecdotal
Selection of
fiber bundles
Stand
Alone
Web
Interface
Application,
Digital Map
Interface.
Inventing an Interface
for 3D Selection of
Neural Pathway
Estimates
Stream tube
Shape
matching
and Touch
Round
formed set of
language
identified
pathways
3-D
Region
of
Interest
Selection
and
refinement of
gestures from
neural maps.
Exploring Connectivity
of the Brains White
Matter with Dynamic
Queries
Stream tube
Diffusion
Tensor
Imaging
Estimate the
precomputed
pathways
Region
of
Interest
Determinatio
n of fiber
cluster using
Boolean
drivers
_______________________________
Page 191

ISSN: 2250-1797
_______________________
D.
Comparison of Clustering Techniques.
Table-2. Comparison of Clustering Techniques.

Paper Title
Algorithm
Selection Type
Application
Exploring Functional Connectivity in

FMRI via Clustering
K-Means and
Spectral Clustering
Region of Interest
Approximate
Spectral
Decomposition
A unified framework for clustering and

quantitative analysis of white matter
fiber tracts
ExpectationMaximization
Region of Interest
Curve
matching.
Clustering Fiber Traces Using

Normalized Cuts.
Normalized cut
Criterion
Automatic
segmentation
Surgical
planning.
Automated Fuzzy Clustering of

Neuronal Pathways in Diffusion Tensor
Tracking
Fuzzy c-means
Segmentation of
voxel space
Classification
of subcortical
tracks.
CONCLUSION
So by surveying and reviewing different literatures I have understood the problem in depth
and also studied possible approaches to solve problem and look for the improvement
suggested by different authors to design new algorithm which solve the problem in effective
way. Visualization of brain neural maps using stream tube model gives more accurate results
in the detection of brains tumor surgery. Clustering is an important tool for creating
abstraction and filtration. In general abstraction and filtration can help to overcome the
difficulties of visual complexity.
REFERENCE
1.
2.
3.
Basak Alper, Benjamin Bach, Nathalie Henry Riche, Tobias Isenberg and Jean-Daniel
Fekete, Weighted Graph Comparison Techniques for Brain Connectivity Analysis
Community Health & Institute-CHI 2013, pp.483-492, April 27May 2, 2013, Paris,
France ,ACM 2013.
Maria Giulia Preti, Nikos Makris, Maria Marcella Lagana, George Papadiamitiou,
Francesca Bagilo, Ludovica Griffanti, Raffaello Nemni, Pietro Cecconi, Carl-Fredrik
Westin and Giuseppe Baselli, A novel approach of fMRI- guided tractography analysis
within a group: construction of an fMRI-guided tractography atlas, 34th Annual
International Conference of the IEEE EMBS ,San Diego, California USA, pp. 2283-2288
28 August - 1 September, 2012
Radu Jianu, Cagatay Demiralp and David H.Laidlaw, Exploring Brain Connectivity with
Two-Dimensional Neural Maps, IEEE Transaction on Visualization and Computer
Graphics, vol. 18, no. 6, pp. 978-987, June 2012.
_______________________________
Page 192

ISSN: 2250-1797
_______________________
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Archana Venkataraman, Koene R.A. Van Dijk, Randy L.Bucker and Polina Golland,
Exploring Functional Connectivity in FMRI via Clustering, Proceedings of the IEEE
International Conference on Acoustics, Speech, and Signal Processing-ICASSP, pp. 441444, 19-24 June 2009, Taipei, Taiwan.
Mahnaz Maddah, W. Eric L. Grimson, Simon K. Warfield and William M. Wells,A
Unified Framework for Clustering and Quantitative Analysis of White Matter Fiber
Tracts,Elsevier Medical Image Analysis, vol. 12, no. 2, pp. 191-202, 2008.
David Akers, Wizard of Oz for Participatory Design: Inventing an Interface for 3D
Selection of Neural Pathway Estimates, Proceedings, Computer Human Interaction (CHI
06) Extended Abstracts, pp. 454-459, 2006.
Anthony Sherbondy, David Akers, Rachel Mackenzie, Robert Dougherty and Brian
Wandell, Exploring Connectivity of the Brains White Matter with Dynamic Queries,
IEEE Trans. Visualization and Computer Graphics, vol. 11, no. 4, pp. 419-430, 2005.
Bart Moberts, Anna Vilanova and Jarke J. van Wijk, Evaluation of Fiber Clustering
Methods for Diffusion Tensor Imaging, Proc. IEEE Visualization (VIS 05), pp. 65-72,
2005.
Anders Brun, Hans Knutsson and Hae-Jeong Park, Clustering Fiber Traces Using
Normalized Cuts, Christian Barillot, David R. Haynor, Pierre Hellier (Eds.): Medical
Image Computing and Computer-Assisted Intervention - MICCAI 2004, 7th International
Conference Saint-Malo, France, pp.368-375, September 26-29, 2004, Proceedings, Part
II. Springer 2004.
Song Zhang, Cagatay Demiralp, and David Laidlaw, Visualizing Diffusion Tensor MR
Images Using Streamtubes and Streamsurfaces, IEEE Trans. Visualization and
Computer Graphics, vol. 9, no. 4, pp. 454-462, Oct.-Dec. 2003.
Susumu Mori and Peter C.M. van Zijl, Fiber Tracking: Principles and Strategies
review, NMR in Biomedicine, vol. 15, nos. 7/8, pp. 468-480, 2002.
Joshua S. Shimony, Avi Z. Snyder, Nicholas Lori and Thomas E. Contum, Automated
Fuzzy Clustering of Neuronal Pathways in Diffusion Tensor Tracking, Proceedings.
International Society. For Magnetic Resonance in Medicine, 2002.
I.Corouge, S. Gouttard, and G. Gerig, Towards a shape model of white matter fiber
bundles using diffusion tensor MRI,In International Symposium on Biomedical Imaging,
Conference Proceedings, pp. 344347, 2004.
Z. Ding, J.C. Gore, and A.W. Anderson, Visualization and quantification of neuronal
fiber pathways, In IEEE Visualization01, Conference Proceedings, pp. 453456. IEEE
Computer Society, 2001.
Anders Brun, Hae-Jeong Park, Hans Knutsson, and Carl-Fredrik Westin,Coloring of DTMRI fiber traces using laplacian eigenmaps, In Computer Aided System TheoryEUROCAST Conference Proceedings, pp.564572. Springer Verlag, February 2428
2003.
_______________________________
Page 193

A Comparative Study of Brain Connectivity Visualization and Clustering Techniques

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Comparative Study of Brain Connectivity Visualization and Clustering Techniques

Uploaded by

Copyright:

Available Formats

International Journal of Computer Application

Issue 4, Volume 1 (January-February 2014)

International Journal of Computer Application

Visualize using Graph Method

Visualize using fMRI guided Tractography

International Journal of Computer Application

Visualize using Neural Maps

Visualize of Brain in Three dimensional view

Visualize using Diffusion tensor imaging and Magneto resonance

Visualize using Stream tube and Stream surface

Visualize using Diffusion tensor imaging tractography

International Journal of Computer Application

Clustering via K-Means and Spectral Clustering

The utilization of information driven bunching techniques for utilitarian connectivity

Clustering using Expectation Maximization Algorithm

International Journal of Computer Application

Clustering using Normalized Cut Criterion

Clustering using Fuzzy C-Means Algorithm

International Journal of Computer Application

Comparison of Visualization Techniques

Table-1. Comparison of Visualization Techniques

International Journal of Computer Application

Comparison of Clustering Techniques.

Table-2. Comparison of Clustering Techniques.

Exploring Functional Connectivity in

A unified framework for clustering and

Clustering Fiber Traces Using

Automated Fuzzy Clustering of

International Journal of Computer Application

You might also like