Intrinsic Variability of GM Density Maps and its Implications to VBM Studies

Poliakov, Andrew V and McLaughlin, Judy and Valentine, Geoffry and Pitkanen, Ilona and Osterhout, Lee (2007) Intrinsic Variability of GM Density Maps and its Implications to VBM Studies. In: Human Brain Mapping.

[thumbnail of OHBM_2007_Poster_Intrinsic_Variability.pdf]

Download (3MB) | Preview
[thumbnail of OHBM_2007_Poster_Intrinsic_Variability.ppt] Slideshow

Download (1MB)


Voxel Based Morphometry (VBM) has been gaining popularity as an unbiased objective neuroimaging technique for identifying structural changes in the brain. VBM involves a voxel-wise comparison of the local concentration of gray matter (GM) in whole brain MRI scans. Although it was originally devised to examine structural abnormalities in patients, the technique has also been used successfully with healthy subjects. Recent VBM studies have investigated the impact of learning and practice on brain structure. Unlike certain medical conditions that may cause dramatic structural changes, effects observed in healthy subjects are expected to be small, therefore imposing stringent requirements on the sensitivity of the technique. The success of such studies depends on high quality imaging and subsequent accurate segmentation of GM. Segmentation results are inevitably affected by the presence of other tissues with similar intensity (dura matter, large blood vessels etc.), imaging artifacts (blood flow and eye movement, susceptibility artifacts etc.). Since these factors are non-homogeneous throughout the brain, segmentation is highly reproducible in some areas of cortex while it is less reliable in other areas. This non-homogeneity makes VBM sensitivity selective to areas where segmentation happens to be more robust.

We studied the intrinsic variability of GM density maps derived from scans obtained under identical conditions, i.e. the same subject, scanner and protocol. The data was acquired on GE Signa 1.5, (SPGR) and Philips Achieva 3T (MPRAGE) scanners. A distinction should be made between variability observed among scans acquired within the same session and that observed for different sessions, since the latter will also be affected by such factors as different head positioning and the somewhat altered state of both the subject and the scanner. The figure summarizes within-session variability of GM density maps observed using the GE Signa. Six SPGR scans were obtained in each of four subjects in one session, and the scan sessions were repeated nine weeks later as a part of longitudinal VBM study. Variability for one subject/session was estimated by computing the standard deviation of six GM density maps obtained using SPM5 unified segmentation/normalization framework and VBM5 toolkit. These were normalized by applying a transformation estimated as follows: all six scans were coregistered and averaged to obtain a low noise structural image volume and a single normalization transformation was estimated from it. Eight variability maps in standard (MNI) space corresponding to session/subject pairs were averaged to produce a map shown in the Figure. The color coded variability map is superimposed onto the GM probability density map (only the right hemisphere is shown in the figure).

We will present the findings of within and between session variability analyses derived from our data and from data obtained in other laboratories, and discuss implications and methodological considerations for planning and interpreting VBM studies of GM density. Preliminary results indicate that although different scanners and protocols produce varying patterns of GM variability maps, certain areas (e.g. tip of the temporal lobe) may consistently show increased variability.

Item Type: Book Section
Uncontrolled Keywords: Voxel Based Morphometry VBM Brain imaging mapping GM density maps SPM
Subjects: All Projects > Brain Mapping
Depositing User: Jim Brinkley
Date Deposited: 13 Sep 2007
Last Modified: 24 Jul 2017 23:33

Actions (login required)

View Item View Item