See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/285989246

FMRI brain artifact due to normalization: A study

Conference Paper  in  Communications in Computer and Information Science · November 2011

DOI: 10.1007/978-3-642-25462-8_27

CITATION READS

1 1,203

5 authors, including:

J. SatheeshKumar Reghunadhan Rajesh

Bharathiar University Central University of Kerala
66 PUBLICATIONS   555 CITATIONS    98 PUBLICATIONS   1,161 CITATIONS   

SEE PROFILE SEE PROFILE

Perumal Arumuga Chandrasekharan Kesavadas

S.T. Hindu College Sree Chitra Tirunal Institute for Medical Sciences and Technology
65 PUBLICATIONS   466 CITATIONS    357 PUBLICATIONS   4,363 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

VIDEO COMPRESSION View project

Virtual Reality for Neuroanatomy teaching View project

All content following this page was uploaded by Reghunadhan Rajesh on 10 September 2016.

The user has requested enhancement of the downloaded file.

 FMRI Brain Artifact Due to Normalization:
A Study

J. SatheeshKumar1, , R. Rajesh1, , S. Arumugaperumal2, ,

C. Kesavdass3,   , and R. Rajeswari1,
1
Bharathiar University, Coimbatore, India
2
ST Hindu College, Nagerkoil, India
3
Sree Chitra Tirunal Institute for Medical Science and Technology,
Trivandrum, India
{jsathee,kollamrajeshr}@ieee.org,
{arumugam.visvenk,chandrakesav,rrajeswari}@gmail.com

Abstract. Medical Imaging is an application of image processing in

which normalization is one of the important process involved in most
of the medical image analysis. Normalization is the process of mapping
source image with same stereotaxic space. This can be done by register-
ing each image to the same template, where template can be constructed
by considering average of large number of high resolution MRImages.
Normalizing source image with common existing template will help in
analyzing inter subject relationships based on various factors, such as,
age, sex etc. But for analyzing single patient data, normalization step can
be skipped by registering source image with subject’s/patient’s anatom-
ical data. Since there may be a variation between template and subject
data, normalization step may either stretch or shrink the source image,
where there are high chances of shift in motor activation area. This paper
proves with experimental results of a trivial example of a subject, where
normalization step have to be ignored for single subject analysis.

Keywords: Normalization, Realignment, MRI, Registration.

1 Introduction

Image processing has broad spectrum of applications in which Medical imaging

is an interesting area for scientists and medical researcher. Medical imaging is
a process of acquiring, analyzing and inferring known and unknown informa-
tion from an image of a human or living organism. Latest developments and
innovations in medical history show the role and importance of medical imag-
ing applications and its significant influence on increasing average human life

Dr. J. Satheesh Kumar, Dr. R. Rajesh and Ms. R. Rajeswari are with Department
of Computer Applications, School of Computer Science and Engineering.

Dr.S. Arumugaperumal is with Department of Computer Science.

Dr. C. Kesavadas is with Department of Imaging Sciences and Interventional
Radiology.

A. Abd Manaf et al. (Eds.): ICIEIS 2011, Part III, CCIS 253, pp. 306–319, 2011.

c Springer-Verlag Berlin Heidelberg 2011
 FMRI Brain Artifact Due to Normalization 307

span time[4],[5],[6], [26], [42], [49]. Significant applications of medical imaging

are Detection of cancerous region from the skin[16], Identifying tumor from
human brain[18], [19],[29], [38], [39], Understanding the functionality and be-
havior of human brain structure[44], [46], [47], Identifying cancer from mam-
mogram images[25], [32], [37], [48], analyzing the functionality of adrenal gland
images[10],[17], [31], [51], etc.
Brain is an important part of human body which is having complex structure
of neurons. Understanding structure and functionality of brain is still a chal-
lenging task for medical researchers due to significant increase in brain related
diseases in the past decades. Significant studies which have been carried out on
brain are Positron emission tomography study on emotional responses[22], Emo-
tional analysis on human brain [2], [3], Neuroanatomical correlates of pleasant
and unpleasant conditions on brain[27], Analysis on corpus callosum[14], Quan-
titative analysis on connectivity of brain[33], Behavioral responses of brain[20],
[24], [30], [34], [43], etc.
Generally different modalities [PET, MRI, fMRI, MEG, SPECT] are available,
which brings various information on brain activity based on some characteristics.
The fMRI is one of the effective modality of acquiring brain images for analyzing
motor activation area. Image Normalization is one of the important preprocess-
ing step, where, the difference between source object and reference image are
reduced by using different normalization methods. This paper discusses about
the influence of image normalization during image pre processing among inter
subject comparison and role of normalization in intra subject analysis. This pa-
per proves that normalization step can be ignored for single subject analysis.
Section 2 elucidates various image preprocessing steps needed for medical image
analysis of brain and section 3 deals with the importance of normalization as well
as the place where the normalization process seems to be difficult during image
analysis. Section 4 deals with results and discussions and section 5 concludes
the paper.

2 Image Preprocessing Steps for Brain Image Analysis

2.1 Image Acquisition
The first step involves reading T1 weighted MR images of patients by using
single scanner with high resolution, in which, each image refers to sequence of
tissue with some specific thickness [of 1.6mm]. Even slight variation in resolution
between two different scanner may have higher influence during analysis phase.
Two types of data can be resulted by most of the scanners namely structured
and functional data, where structured data have more resolution compare with
functional data. Functional images of patient’s data can be overlaid with struc-
tural data of the same subject for identifying specific motor activation area [1],
[8], [9], [11], [15], [28],[35],[36], [50].
308 J. SatheeshKumar et al.

2.2 Spatial Normalization and Transformation

Spatial normalization is the process of mapping source image with some reference
image shown in Figure 1 for reducing residual errors between source and target
images. An advantage using spatially normalized images shown in Figure 2 is that
the motor activation area for different functionalities can be analyzed accurately
based on set of meaningful coordinates with in standard space [7]. The first
step of normalization is spatial transformation shown in Figure 3, which can
be broadly classified as label-based techniques and non-label based techniques
[7], [12]. Similar features (labels) in the image and template can be identified
by label-based approach, where as, spatial transformation that minimizes some
index of the difference between an object and a template image can be analyzed
by non-label based approaches.

Fig. 1. Reference Images used to map with source image: MNI Template image (left),
Single subject T1 mean image (right)

Fig. 2. (1).Results before normalization in a study, where, multiple subjects were an-
alyzed by playing auditory words. (2). Result after normalization.
 FMRI Brain Artifact Due to Normalization 309

Fig. 3. Reorientation of images: Source Image(left), translation of source image through

zooms and shears(right)

2.3 Segmentation and Extraction

By using segmentation techniques, the images after normalization can be de-

composed into various factor like gray matter [GM], white matter[WM] and
cerebrospinal fluid [CSF] based on intensities of voxels or pixels. Image extrac-
tion is also one of the process of removing noise. Some non-brain voxels may
have the similar intensities of tissue like gray matter that can be removed by
taking some effective brain extraction step[37].

2.4 Smoothing

The process of smoothing takes place after effective image extraction technique.
Smoothing is a kind of enhancing an image for accurately analyzing target. Gen-
erally Isotropic Gaussian kernel are used for smoothing. The images to analyze
the differences between groups and local volume of various tissues (such as, gray
matter and white matter) can be calculated from the smoothed images based
on intensity value of pixel or voxel value of an image. Finally, results can be
compared by applying various statistical approaches[45].

2.5 Statistical Analysis

Many techniques have been proposed for statistically analyzing fMRI data like,
multi variate analysis of co-variance (MANCOV), canonical correlation analysis
(CCA) can be used to analyze the difference between groups of images, and a
variety of these are in general use. The aim of such analysis is to produce an
image identifying the regions, which show significant signal change in response
to the task. Each pixel is assigned a value dependent on the likelihood of the null
hypothesis, that the observed signal changes can be explained purely by random
variation in the data consistent with its variance, being false. Such an image is
called a statistical parametric map [13], [21], [23].
310 J. SatheeshKumar et al.

Fig. 4. Steps involved in medical image processing: (1). Source image (2).High res-
olution mean MR image(template, used to map with source image) (3). Image after
normalization (4) Images after reducing noise by taking effective segmentation tech-
niques (5) result after smoothing segmented images.

3 Normalization - As a Dragon Creating Disadvantages

As mentioned earlier, normalization is the process of adjusting images by super-

imposing source image with the reference image or template. The template is a
mean image, which can be constructed by considering average of large number
of high resolution MR images. When a study like, identifying relational differ-
ences with same modality, the concept of normalization is needed so as to reduce
errors or differences (like sum of squared differences) between source image and
template image shown in Figure 2. After successful image acquisition and co
registration of multiple subjects with similar modality based on several factors
(like, age, sex, diseases etc.,), there are higher chances of variations in every pa-
tient’s brain shapes[45], [40], [41]. In order to get correct and best-known result,
all the images can be mapped with same stereotaxic space, where by noise and
unnecessary portions can be removed so that relationship between subjects can
be accurately identified. Figure 4 clearly shows sequence of image pre processing
steps during image analysis. Source images are generally mapped with prede-
fined templates (MNI) or it can be mapped with user defined reference image
that can be constructed by using structural data of the patient or subject.
Due to the realignment by motion correction parameters during inter subject
analysis of normalization phase, some subjects might have changed to make best
fit with reference image as well as to reduce the residual error between object
and template image. In a study about single subjects brain, the mapping process
of subject image with pre defined template (MNI created based on considering
13 subjects high resolution MR images) is not suggestible during image analysis.
When the mean image is used for mapping, there may be higher possibilities of
variation between object and reference image. Because of stretching and shrink-
ing based on template, there may be higher chances of losing data Figure 5 or
significant shift in motor activation area.
 FMRI Brain Artifact Due to Normalization 311

Fig. 5. Complexities in images after normalization, where, single subject data is su-
perimposed with predefined template. Higher chances of data lose (top right figures)
because of variation between source image and template image.

Fig. 6. Normalized images of a subject during the study of different music

4 Results and Discussions

Functional magnetic resonance images used in this study was obtained from
Sree Chitra Tirunal Institute for Medical Science and Technology, Tiruvanan-
thapuram, using 1.5 T MRI system (Siemens Avanto). A structural image of
the subject’s brain with 176 slices of 1 mm thickness was imaged for overlaying
final results. The subject was asked to hear different types of music like karnatic,
instrumental and white noise during the experiment. Functional images of 336
volumes were obtained, where each volume consist of 36 slices of 3 mm thickness
at the scanning rate of 3.58 seconds per volume. The boxcar paradigm used in
this experimental analysis is shown in the Figure 7. The total experiment took
around 20 minutes to complete the scan.
Experiment have carried out with a trivial example of subject, where the
normalization seems to be difficult. The shape of the patient left and right brain
312 J. SatheeshKumar et al.

Fig. 7. Boxcar design paradigm for the experiment

vary significantly. Hence the normalization with template seems to be difficult

and the results obtained after normalization are shown in the Figure 6. It is
clear from the figure that data loses due to normalization. Fields significantly
activated by music 1 and music 2 for normalized and unnormalized images are
shown in the tables 1, 2, 3, 4.

Fig. 8. Maximum Intensity projection of normalized images activated by music1

The results shows maximum intensity projection(MIP) for music 1 and music
2 on so called glass brain. Figure 9 shows MIP for music 1 at position [-58 -10 -29]
and Figure 11 shows MIP for music 2 at position [8 55 -40] with out performing
normalization during preprocessing of images. The MIP for normalized images
are shown in the Figure 8 and Figure 10 at position [-58 0 -10] for music 1
and at position [18 -90 -38] for music 2. The results clearly shows that the MIP
position for music has changed from actual location to some other position due to
normalization. MNI coordinate position for these MIP can be identified by using
a meta analysis matlab tool box(AMAT). Table 5 shows brain region for MIP
coordinates on MNI space. These micro level changes will lead the radiologists
and medical researchers for miss-understanding about hidden information in
 FMRI Brain Artifact Due to Normalization 313

Fig. 9. Maximum Intensity projection for music1 without performing normalization

Fig. 10. Maximum Intensity projection of normalized images activated by music2

314 J. SatheeshKumar et al.

Fig. 11. Maximum Intensity projection for music2 without performing normalization

Table 1. Fields significantly activated, patients image without performing normaliza-

tion, by music 1 obtained after t-test (p-values are adjusted for search volume)

Cluster-level voxel-level x, y, z {mm}

pcorr kE puncorr pF W E−corr pF DR−corr T ZE puncorr

0.000 203 0.000 0.000 0.000 6.01 5.84 0.000 -59 -10 -29
0.044 0.001 4.72 4.63 0.000 -51 -2 -37
0.249 0.004 4.22 4.16 0.000 -66 -22 -11
0.000 141 0.000 0.000 0.000 5.72 5.58 0.000 58 8 -41
0.003 0.000 5.31 5.20 0.000 66 -4 -33
0.794 6 0.207 0.915 0.026 3.55 3.51 0.000 52 -36 -11
0.972 2 0.469 0.999 0.070 3.17 3.15 0.001 27 6 -41
0.991 1 0.621 0.999 0.073 3.14 3.12 0.001 10 -54 -59
0.991 1 0.621 0.999 0.077 3.13 3.10 0.001 23 -6 -37
 FMRI Brain Artifact Due to Normalization 315

Table 2. Fields significantly activated, patients image without performing normaliza-

tion, by music 2 obtained after t-test (p-values are adjusted for search volume)

Cluster-level voxel-level x, y, z {mm}

pcorr kE puncorr pF W E−corr pF DR−corr T ZE puncorr

0.850 5 0.248 0.813 0.530 3.68 3.64 0.000 8 56 -41

0.183 21 0.027 0.882 0.530 3.60 3.56 0.000 24 -92 -44
0.936 0.530 3.51 3.47 0.000 16 -96 -44
0.794 6 0.207 0.910 0.530 3.56 3.52 0.000 42 18 -59
0.972 2 0.469 0.974 0.530 3.41 3.37 0.000 -60 -57 -7
0.566 10 0.109 0.980 0.530 3.38 3.35 0.000 -48 14 -59
0.991 1 0.621 0.998 0.557 3.18 3.16 0.001 1 -27 -48

Table 3. Fields significantly activated normalized images by music 1 obtained after

t-test (p-values are adjusted for search volume)

Cluster-level voxel-level x, y, z {mm}

pcorr kE puncorr pF W E−corr pF DR−corr T ZE puncorr

0.000 1683 0.000 0.000 0.000 6.17 5.99 0.000 -58 0 -10
0.000 0.000 5.86 5.70 0.000 -64 -10 -8
0.032 0.001 4.83 4.74 0.000 -48 14 -20
0.000 1072 0.000 0.001 0.000 5.74 5.60 0.000 58 20 -18
0.003 0.000 5.41 5.29 0.000 66 4 -10
0.417 0.006 4.08 4.02 0.000 52 8 -12
0.714 61 0.144 0.836 0.017 3.70 3.66 0.000 52 -36 12
0.994 9 0.582 0.979 0.035 3.43 3.40 0.000 10 -42 -50
1.000 1 0.882 0.999 0.068 3.18 3.15 0.001 24 4- 14
0.999 3 0.771 1.000 0.070 3.17 3.14 0.001 -42 -6 -20
1.000 1 0.882 1.000 0.073 3.15 3.12 0.001 26 6 -18
0.999 2 0.820 1.000 0.076 3.14 3.11 0.001 28 12 -18
316 J. SatheeshKumar et al.

Table 4. Fields significantly activated normalized images by music 2 obtained after

t-test (p-values are adjusted for search volume)

Cluster-level voxel-level x, y, z {mm}

pcorr kE puncorr pF W E−corr pF DR−corr T ZE puncorr

0.280 134 0.038 0.823 1.000 3.71 3.67 0.000 18 -90 -38
0.838 1.000 3.70 3.65 0.000 28 -88 -36
0.991 11 0.539 0.919 1.000 3.59 3.55 0.000 10 70 -10
0.983 15 0.467 0.920 1.000 3.59 3.55 0.000 44 34 -38
0.999 3 0.771 0.988 1.000 3.39 3.35 0.000 -44 32 -38
0.995 8 0.606 0.993 1.000 3.34 3.31 0.000 -62 -56 14
1.000 1 0.882 0.998 1.000 3.24 3.21 0.001 42 34 -44

Table 5. Maximum Intensity Projection on MNI space

Experiment with/without M IP V alues Brainregionbased

normalization onM N Itemplate

Music 1 Un Normalized images -59 -10 -29 inferior temporal cortex

Normalized Images -58 00 -10 anterior middle temporal gyrus

Music 2 Un Normalized images 08 56 -41 medial orbitofrontal cortex

Normalized Images 18 -90 -38 Right cerebellum

complex human brain structure and hence normalization phase can be avoided
for single subject data.

5 Conclusion
This paper explains various image processing steps and the role of normalization
in medical image analysis. This paper proves by contradiction that due to higher
chances of variation between source and predefined template, normalization can
be ignored for single patient data to get best-known result so that data lost can
be avoided.
Acknowledgement. The first two authors are thankful to the Department of
Imaging Sciences and Interventional Radiology, Sree Chitra Tirunal Institute
for Medical Sciences and Technology for supporting them to do their research
training in the institution. They are also thankful to all staff of the Depart-
ment of Computer Applications, School of Computer Science and Engineering,
Bharathiar University, India for their support. The first, second and fifth authors
are thankful for the partial funding support received from University Grants
Commission(UGC), India.
 FMRI Brain Artifact Due to Normalization 317

References
1. Klautau, A.: Multiplicative Homomorphic Processing and its Application to Image
Enhancement (2000)
2. Ahren, G.L., Schwartz, G.E.: Differential lateralization for positive and negative
emotion in the human brain: EEG spectral analysis. Neuropsychologia 23, 745–755
(1985)
3. Angrilli, A., Palomba, D., Cantagallo, A., Maietti, A., Stegagno, L.: Emotional
impairment after right orbitofrontal lesion in a patient without cognitive deficits.
NeuroReport 10, 1741–1746 (1999)
4. Sherbondy, A., Akers, D., Mackenzie, R., Dougherty, R., Wandell, B.: Exploring
Connectivity of the Brains White Matter with Dynamic Queries. IEEE Transac-
tions on Visualization and Computer Graphics 11(4), 419–430 (2005)
5. May, A., Gaser, C.: Magnetic resonance-based morphometry: A window into struc-
tural plasticity of the brain. Current Opinion in Neurology 19, 407–411 (2006)
6. Ashburner, Friston, K.: Morphometry, PhD Thesis, Chapter 6 (2000)
7. Ashburner, J.: Friston K, Nonlinear Spatial Normalization using Basis Functions,
Welcome department of cognitive neurology. Human Brain Mapping 7(4), 254–266
(1999)
8. Ashburner, J., Friston, K.J.: Voxel-based morphometry-The Methods. Neuro. Im-
age 11, 805–821 (2000)
9. Bogorodzki, P., Rogowska, J., Yurgelun-Todd, D.A.: Structural group classification
technique based on regional fMRI BOLD responses. IEEE Transactions Medical
Imaging 24(3), 389–398 (2005)
10. Chang, A., Glazer, H.S., Lee, J.K.T., Ling, D., Heiken, J.: Adrenal gland: MR
imaging. Radiology 163, 123–128 (1987)
11. Daniel, N.R., Dennis Jr., M.H.: Modern Signal Processing, vol. 46. MSRI Publica-
tions
12. Davatzikos, C.: Computational neuroanatomy using shape transformations - Hand-
book of medical imaging, vol. 16, pp. 249–260. Academic Press (2000)
13. Cahn, D.A., Sullivan, E.V., Shear, P.K., Marsh, L., Fama, R., Lim, K.O., Yesavage,
J.A., Tinklenberg, J.R.: Adolf Pfefferbaum: Structural MRI correlates of recogni-
tion memory in Alzheimer’s disease. Journal of the International Neuropsycholog-
ical Society 4, 106–114 (1998)
14. Lee, D.J., Chen, Y., Schlaug, G.: Corpus Callosum: Musician and Gender
Ejects 14(2), 205–209 (2003)
15. Selle, D., Spindler, W., Preim, B., Peitgen, H.-O.: Mathematical Methods in Med-
ical Imaging: Analysis of Vascular Structures for Liver Surgery Planning (2000)
16. Ercal, F., Moganti, M., Stoecker, W.V., Moss, R.H.: Detection Of Skin Tumor
Boundaries In Color Images. IEEE Transactions on Medical Imaging 12(3) (1993)
17. Krestin, G.P., Steinbrich, W., Friedmann, G.: Adrenal masses: Evaluation with
fast dynamic gradient echo MR imaging and Gd-DTPA-enhanced dynamic studies.
Radiology 171, 675–680 (1989)
18. Gibbs, P., Buckley, D., Blackb, S., Horsman, A.: Tumour volume determination
from MR images by morphological segmentation. Physics in Medicine and Biol-
ogy 41, 2437–2446 (1996)
19. Görlitz, L., Menze, B.H., Weber, M.-A., Kelm, B.M., Hamprecht, F.A.: Semi-
supervised Tumor Detection in Magnetic Resonance Spectroscopic Images using
Discriminative Random Fields. In: Hamprecht, F.A., Schnörr, C., Jähne, B. (eds.)
DAGM 2007. LNCS, vol. 4713, pp. 224–233. Springer, Heidelberg (2007)
318 J. SatheeshKumar et al.

20. Guimaraes, A.R., Melcher, J.R., Talavage, T.M., Baker, J.R., Ledden, P., Rosen,
B.R., Kiang, N.K.S., Fullerton, B.C., Weisskoff, R.M.: Imaging Subcortical Audi-
tory Activity in Humans. Human Brain Mapping 6, 33–41 (1998)
21. Friedl, H., Kauermann, G.: Standard Errors for EM Estimates in Generalized Lin-
ear Models with Random Effects. Biometrics 56(3), 761–767
22. Royet, J.P., Zald, D., Versace, R., Costes, N., Lavenne, F., Koenig, O., Gervais,
R.: Emotional Responses to Pleasant and Unpleasant factory, Visual, and Au-
ditory Stimuli: a Positron Emission Tomography Study. The Journal of Neuro-
science 20(20), 7752–7759 (2000)
23. Keller, S.S., Wieshmann, U.C., Mackay, C.E., Denby, C.E., Webb, J., Roberts, N.:
Voxel based morphometry of grey matter abnormalities in patients with medically
intractable temporal lobe epilepsy: effects of side of seizure onset and epilepsy
duration. Journal of Neurology Neurosurgery and Psychiatry 73, 648–655 (2002)
24. Kling, A., Steklis, H.D.: A neural basis for affiliative behavior in non-human pri-
mates. Brain, Behavior, and Evolution 13, 216–238 (1976)
25. Kobatake, H., Yoshinaga, Y., Murakami, M.: Automated detection of malignant
tumors on mammogram. In: Proceedings of the IEEE International Conference on
Image Processing, vol. 1, pp. 407–410 (1994)
26. Kubota, J., et al.: Neurol Neurosurg Psychiatry, Alcohol consumption and frontal
lobe shrinkage: study of 1432 non-alcoholic subjects. JNNP 71, 104–106 (2001)
27. Lane, R.D., Reiman, E., Bradley, M.M., Lang, P.J., Ahern, G.L., Davidson, R.J.:
Neuroanatomical correlates of pleasant and unpleasant emotion. Neuropsycholo-
gia 35, 1437–1444 (1997)
28. Lawrence, A.A., Ritter, G.X.: Cellular topology and its applications in image pro-
cessing. International Journal of Parallel Programming 12 (1983)
29. Lefohn, A.E., Cates, J.E., Whitaker, R.T.: Interactive, GPU-Based Level Sets for
3D Brain Tumor Segmentation. In: Ellis, R.E., Peters, T.M. (eds.) MICCAI 2003.
LNCS, vol. 2878, pp. 564–572. Springer, Heidelberg (2003)
30. McEwen, B.S.: Physiology and neurobiology of stress and adaptation: Central role
of the brain. Physiological Reviews 87, 873–904 (2007)
31. Mitchell, D.G., Crovello, M., Matteucci, T., Petersen, R.O.: Miettinen MM Benign
adrenocortical masses: diagnosis with chemical shift MR imaging. Radiology 185,
345–351 (1992)
32. Wirth, M., Lyan, J., Nikitenko, D., Stapinski, A.: Removing radiopaque artifacts
from mammograms using area morphology. In: Proceedings of SPIE Medical Imag-
ing: Image processing, vol. 5370, pp. 1054–1065 (2004)
33. Murre, J., Sturdy, D.: The connectivity of the brain: multi-level quantitative anal-
ysis. Biological cybernetics, Neuroreport 73(6), 529–545 (1995)
34. Noriuchi, M., Kikuchi, Y., Senoo, A.: The functional neuroanatomy of maternal
love: Mothers response to infants attachment behaviors. Biological Psychiatry 63,
415–423 (2008)
35. Ohser, J.,Schladitz, K., Koch, K., Nothe, M.: Diffraction by image processing and
its application in materials science. ITWM, Nr.67 (2004)
36. Patel, J., Lee, K.F., Goldberg, B.: The role of ultra sonography in the diagnosis of
certain neurologic disorders. Neuroradiology, 1432–1920 (Online)
37. Petrick, N., Chan, H.P., Sahiner, B., Helvie, M.A.: Combined adaptive enhance-
ment and region-growing segmentation of breast masses on digitized mammograms.
Medical Physics 26(8), 1642–1654 (1999)
38. Prastawa, M., Bullitt, E., Ho, S., Gerig, G.: A brain tumor segmentation framework
based on outlier detection. Medical Image Analysis 8(3), 275–283 (2004)
 FMRI Brain Artifact Due to Normalization 319

39. Prastawa, M., Bullitt, E., Moon, N., Leemput, K.V., Gerig, G.: Automatic brain tu-
mor segmentation by subject specific modification of atlas priors. Acad. Radiol. 10,
1341–1348 (2003)
40. Rajesh, R., SatheeshKumar, J., Arumugaperumal, S., Kesavdas, C.: Have a look at
the 3 dimensional view of t-statistics? - Isn’t it cute ginger. The Neuroradiology 21,
31–34 (2008)
41. Rajesh, R., SatheeshKumar, J., Arumugaperumal, S., Kesavdas, C.: On identify-
ing micro level error in realignment phase of statistical parametric mapping. The
Neuroradiology Journal 20, 491–493 (2007)
42. Rowland, Clinical, legal, and research issues in dementia. Am. J. Alzheimers Dis-
orders Other Demen 21, NP (2006)
43. Rusch, N., van Elst, L.T., Ludaescher, P., Wilke, M., Huppertz, H.J., Thiel, T.,
Ebert, D.: A voxel-based morphometric MRI study in female patients with bor-
derline personality disorder. NeuroImage 20, 385–392 (2003)
44. SatheeshKumar, J., Arumugaperumal, S., Rajesh, R., Kesavdas, C.: A Note on
Visualization of Information from Three Dimensional Time Series of Brain. Inter-
national Journal of Recent Trends in Engineering 1(2), 173–175 (2009)
45. SatheeshKumar, J., Rajesh, R., Arumugaperumal, S., Kesavdas, C.: A Novel Algo-
rithm for an Efficient Realigning of fMRI Data Series of Brain. ICGST International
Journal on Graphics, Vision and Image Processing 9(I), 35–40 (2009)
46. SatheeshKumar, J., Arumugaperumal, S., Rajesh, R., Kesavdas, C.: On experi-
menting with functional magnetic resonance imaging on lip movement. The Neu-
roradiology Journal 21, 23–30 (2008)
47. SatheeshKumar, J., Arumugaperumal, S., Kesavdas, C., Rajesh, R.: Does Brain
react on Indian music? - An functional Magnetic Resonance Imaging study. In:
IEEE International Joint Conference on Neural Networks (IJCNN 2008), pp. 2696–
2703 (2008)
48. Sahiner, B., Chan, H.P., Wei, D., Petrick, N., Hlvie, M.A., Adler, D.D., Goodsit,
M.M.: Image feature selection by a genetic algorithm: Application to classifications
of mass and normal breast tissue. Medical Physics 23, 1671–1684 (1996)
49. Dehaene, S., Le Clec’H, G., Cohen, L., Poline, J.-B., van de Moortele, P.-F., Le
Bihan, D.: Inferring behavior from functional brain images. Nature Neuroscience 1,
549 (1998)
50. Perry, S.W.: Applications of Image processing to mine warfare sonar, DSTO-GD-
0237
51. Tsushima, Y., Ishizaka, H., Matsumoto, M.: Adrenal masses: differentiation with
chemical shift, fast low-angle shot MR imaging. Radiology 186, 705–709 (1993)

View publication stats