Influence of MR system, field strength and automatic planning routines on Voxel Based Morphometry (VBM)

Purpose: Since the introduction of automated methods for analyzing structural brain imaging data in 2000 (Ashburner J, Friston K: Voxel based morphometry – the methods; NeuroImage, 2000), the number of published papers using this methodology has grown exponentially. Considering the fact that almost every fMRI-study is accompanied with the acquisition of an high resolution T1-weigthed data set, a huge amount of data sets has been acquired over the years, across studies/sites and scanners. In principle, these data could be used for additional morphometric analysis. However, the influence of factors like field strength, scanner systems, sequence or even scan planning is not known. The goal of the study was investigate the influence of the factors MR-system, field strength as well as the use of an automated versus manual planning procedure in the context of VBM studies.

Materials and Methods: Using an intraindividual approach, 8 subjects were examined four times on two different 3T MR systems (3.0 T Achieva, Philips Medical Systems), twice with and twice without an automated planning routine (SmartScan). To assess the influence of field strength, scanner system and sequence, we additionally examined 13 subjects 4 times on the same day on four different MR systems: on two different 1.5T scanners (Gyroscan NT, Philips Medical Systems; Gyroscan Intera, Philips Medical Systems) and on two 3T scanners (3.0 T Achieva, Philips Medical Systems, equipped with different gradient systems). The sequences chosen were defined by the protocols established as “standard“ anatomical reference at the time the different scanner were set up for fMRI at our institution.

The sequence parameters were as follows:

1.5T Gyroscan NT:	3D-T1-TFE, (TE/TR/Flip)=(3.6/15.36/30°), 2 x averaging
1.5T Gyroscan Intera:	3D-T1-TFE, (TE/TR/Flip)=(3.6/15.36/30°), 2 x averaging
3.0T Achieva 1:	3D-T1-TFE, (TE/TR/Flip)=(3.7/8.114/8°), SENSE factor: 2.3
3.0T Achieva 2:	3D-T1-TFE, (TE/TR/Flip)=(3.7/8.114/8°), SENSE factor: 2.3

The preprocessing was performed using SPM5 as well as the VBM5 toolbox developed by Christian Gaser. SPSS 14.0 was used for multivariate ANOVA for dependent samples.

Results : The comparison of the total gray matter (GM) for the different planning techniques did not reveal any significant differences. There was no significant difference between the test – retest reliability between the results of the manual and automated planning procedure. There were no significant differences within or across the different planning techniques. The comparison across field strength as well as across scanners-systems/sequences revealed significant differences between as well as within field strengths. Even though the difference within the same field strength were significantly smaller compared to the differences across the different field strengths, these were significant for both, total gray matter volume and total white matter volume.

Conclusion: Our data show clearly that pooling of structural data sets across scanners is a critical issue. Even within the same field strength, significant differences between scanner systems were identified. Further analysis will have to determine the spatial distribution of these differences and whether these phanomena are global effects due to slight differences in overall SNR.