Spatial relationship between the interhemispheric fissure plane and the head symmetry plane Qingmao Hu * , Wieslaw L. Nowinski Biomedical Imaging Laboratory, Institute for Infocomm Research, 21 Heng Mui Keng Terrace, Singapore 119613, Singapore Received 13 March 2003; received in revised form 13 March 2003; accepted 17 March 2003 Abstract The spatial relationship between the plane passing through the interhemispheric fissure (IFP) and the symmetry plane (SP) that maximizes brain or head symmetry remains weakly explored. A new method based on local symmetry of the brain/head is proposed to extract the symmetry plane robustly within 6 s. The SPs for 98 diversified MRI data sets (80 normal, 18 pathological) were extracted for the brain (BSP) and head (HSP) and compared with the ground-truth IFPs. The results show that the chance that the IFP and HSP are far is 0.53 (52/98), which increases to 0.78 for pathological data. This suggests that these two planes are generally not the same and care shall be taken to choose the IFP or HSP depending on a specific application. D 2003 Published by Elsevier Science B.V. Keywords: Midsagittal plane; Symmetry; Brain; Head; Outlier removal; Neuroinformatics 1. Introduction The plane separating the two human brain hemispheres is called the midsagittal plane (MSP). Extraction of the MSP is important for spatial normalization; intrasubject, intersubject, and multimodal registration; and automatic detection of the anterior and posterior commissures. Prima et al. [1] grouped the existing methods into two 0531-5131/03 D 2003 Published by Elsevier Science B.V. doi:10.1016/S0531-5131(03)00311-X Abbreviations: MSP, midsagittal plane; IFP, interhemispheric fissure plane; SP, symmetry plane; HSP, head symmetry plane; BSP, brain (cerebrum) symmetry plane. * Corresponding author. Tel.: +65-6874-8281; fax: +65-6774-8056. E-mail addresses: [email protected] (Q. Hu), [email protected] (W.L. Nowinski). International Congress Series 1256 (2003) 220 – 225
6
Embed
Spatial relationship between the interhemispheric fissure plane and the head symmetry plane
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
International Congress Series 1256 (2003) 220–225
Spatial relationship between the interhemispheric
fissure plane and the head symmetry plane
Qingmao Hu*, Wieslaw L. Nowinski
Biomedical Imaging Laboratory, Institute for Infocomm Research, 21 Heng Mui Keng Terrace,
Singapore 119613, Singapore
Received 13 March 2003; received in revised form 13 March 2003; accepted 17 March 2003
Abstract
The spatial relationship between the plane passing through the interhemispheric fissure (IFP) and
the symmetry plane (SP) that maximizes brain or head symmetry remains weakly explored. A new
method based on local symmetry of the brain/head is proposed to extract the symmetry plane
robustly within 6 s. The SPs for 98 diversified MRI data sets (80 normal, 18 pathological) were
extracted for the brain (BSP) and head (HSP) and compared with the ground-truth IFPs. The results
show that the chance that the IFP and HSP are far is 0.53 (52/98), which increases to 0.78 for
pathological data. This suggests that these two planes are generally not the same and care shall be
taken to choose the IFP or HSP depending on a specific application.
Q. Hu, W.L. Nowinski / International Congress Series 1256 (2003) 220–225 221
categories based on either detection of the interhemispheric fissure (IFP) or brain/head
symmetry. The relationship between the MSPs of these two categories remains weakly
explored.
This paper proposes a new method to extract the MSP based on local symmetry. By
employing this method, a spatial relationship between these two categories of MSPs is
explored. To differentiate them, the MSP of the first category is denoted as interhemi-
spheric fissure plane (IFP), while the MSP of the second category is denoted as symmetry
plane (SP). In addition, the relationship between the head symmetry plane (HSP) and the
brain (cerebrum) symmetry plane (BSP) is studied.
2. Materials and methods
2.1. Image data
Altogether, 98 MRI data sets from the USA (four hospitals), France (one hospital),
Japan (one hospital), and Singapore (four hospitals) were tested. Two criteria to choose
the data were applied: (1) each data set shall contain a complete head so that both IFP
and HSP can be extracted, and (2) all data sets shall be representative to have various
imaging parameters as well as be with and without pathology. For variations in noise
level and inhomogeneity, the 20 normal MRI data sets from the Internet Brain
Segmentation Repository (http://neuro-www.mgh.harvard.edu/cma/isbr) were chosen.
Among 98 MRI data sets, 57 are T1-weighted, 36 are SPGR, 3 are T2-weighted, 1
is PD-weighted, and 1 is FLAIR. Eighteen data sets are pathological while 80 are
normal.
2.2. Extraction of the symmetry plane
Existing methods for extracting the symmetry plane (SP) are based on global
symmetry in the volume either through 2D slice processing [2] or direct 3D processing
[1]. As the human brain is just roughly symmetric, the assumption of global symmetry
does not generally hold. A different approach is to assume local symmetry instead [3,4].
Here, ‘‘local’’ stands for a neighborhood instead of the entire volume or the whole slice
image.
The coordinate axes are as follows: X along lateral direction, Y along anterior–posterior
direction, and Z along dorsal–ventral direction. A search line is represented by its middle
point and its angle with respect to the Y axis of the axial image. The symmetry line of an
axial slice is the search line along which the local asymmetry reaches its minimum. The
local asymmetry of a search line is calculated as the sum of gray level difference of all
pixels on five pairs of parallel lines 2, 4, 6, 8, and 10 mm away from the search line.
Sixteen axial slices evenly distributed along the whole head for the HSP or the entire
cerebrum for the BSP are picked out to calculate 16 symmetry lines. Suppose (xc,yc,si) is
the mass center of the ith axial slice, the middle point of all the search lines is denoted as
(x,y,si), and the angle between the search line and Y axis of the axial image is h. Coarse andfine searching are carried out to gain both speed and accuracy for localization of the