Longitudinal stability of MRI for mapping brain change using tensor-based morphometry

Abstract

Measures of brain change can be computed from sequential MRI scans, providing valuable information on disease progression, e.g., for patient monitoring and drug trials. Tensor-based morphometry (TBM) creates maps of these brain changes, visualizing the 3D profile and rates of tissue growth or atrophy, but its sensitivity depends on the contrast and geometric stability of the images. As part of the Alzheimer's Disease Neuroimaging Initiative (ADNI), 17 normal elderly subjects were scanned twice (at a 2-week interval) with several 3D 1.5 T MRI pulse sequences: high and low flip angle SPGR/FLASH (from which Synthetic T1 images were generated), MP-RAGE, IR-SPGR (N = 10) and MEDIC (N = 7) scans. For each subject and scan type, a 3D deformation map aligned baseline and follow-up scans, computed with a nonlinear, inverse-consistent elastic registration algorithm. Voxelwise statistics, in ICBM stereotaxic space, visualized the profile of mean absolute change and its cross-subject variance; these maps were then compared using permutation testing. Image stability depended on: (1) the pulse sequence; (2) the transmit/receive coil type (birdcage versus phased array); (3) spatial distortion corrections (using MEDIC sequence information); (4) B1-field intensity inhomogeneity correction (using N3). SPGR/FLASH images acquired using a birdcage coil had least overall deviation. N3 correction reduced coil type and pulse sequence differences and improved scan reproducibility, except for Synthetic T1 images (which were intrinsically corrected for B1-inhomogeneity). No strong evidence favored B0 correction. Although SPGR/FLASH images showed least deviation here, pulse sequence selection for the ADNI project was based on multiple additional image analyses, to be reported elsewhere.

Introduction

Serial scanning of the human brain with MRI offers tremendous power to detect the earliest signs of illness, monitor disease progression and resolve drug effects in clinical trials that aim to prevent or slow the rate of brain degeneration. Structural MRI provides high-contrast 3D scans, offering excellent ability to differentiate gray and white matter, CSF and other tissues (including disease-related abnormalities). Moreover, with recent advances in mathematical and computational techniques for nonlinear image registration, researchers can now track local tissue change in the human brain based on serial MRI scans. One such approach is called tensor-based morphometry (TBM), which applies a nonlinear deformation field to the baseline scan to align it with the follow-up scan. Based on local analysis of the applied compression and expansion, rates of brain change can be inferred for specific regions of interest or presented in the form of a map. Tensor-based morphometry has been used to map growth patterns in the developing human brain (Thompson et al., 2000, Chung et al., 2001), degenerative rates in Alzheimer's disease and other dementias (Fox et al., 1997, Fox et al., 1999, Fox et al., 2000, Fox et al., 2001, O'Brien et al., 2001, Freeborough et al., 1996, Freeborough and Fox, 1997, Studholme et al., 2001) as well as tumor growth and multiple sclerosis lesions (Lemieux et al., 1998, Ge et al., 1999, Rey et al., 2002). In addition, there has been intensive work on the statistical analysis of deformation fields for detecting whether significant changes have occurred (Worsley, 1994, Worsley et al., 1999, Thompson et al., 1997, Ashburner et al., 1998, Cao and Worsley, 1999, Gaser et al., 1999, Woods, 2003, Fillard et al., 2005) as well as on the elastic and fluid registration algorithms to compute these deformations (Thompson and Toga, 1996a, Thompson and Toga, 1996b, Thompson and Toga, 2002, Fox and Freeborough, 1997, Studholme et al., 2001, Janke et al., 2001, Crum et al., 2001, Miller et al., 2002, Leow et al., 2005a, Leow et al., 2005b).

The Alzheimer's Disease Neuroimaging Initiative (ADNI; Mueller et al., submitted for publication(a), Mueller et al., submitted for publication(b), Mueller et al., in press; see http://www.loni.ucla.edu/ADNI and http://ADNI-info.org) is a large multi-site longitudinal MRI and FDG-PET study of 200 elderly controls, 400 mildly cognitively impaired subjects and 200 Alzheimer's disease subjects. One goal of this project is to develop improved imaging methods to measure longitudinal changes of the brain in normal aging, during the transition to early Alzheimer's disease, and in Alzheimer's disease patients. One of our specific aims was to develop a high-resolution 3D T1-weighted MRI scanning protocol that provided both between-vendor and between-site comparability, as well as longitudinal stability. Despite its usefulness for tracking brain change, there is little information regarding the stability and variability of various MR imaging techniques. Most evidence that MRI has good reproducibility comes from studies that have used rigid registration to identify systematic changes in overall brain volume in serial scans (Hajnal et al., 1995a, Hajnal et al., 1995b, Oatridge et al., 2001, Smith et al., 2002).

Therefore, we performed a series of pilot studies to compare different 3D T1-weighted MRI sequences. Once acquired, these scans were evaluated with a number of different image analysis techniques including: atlas-based measurements of hippocampal volume (Haller et al., 1997, Hsu et al., 2002), the boundary shift integral (Fox and Freeborough, 1997, Fox et al., 2000), voxel-based morphometry using Statistical Parametric Mapping (VBM; Ashburner and Friston, 2000), cortical thickness measures (Fischl and Dale, 2000) and tensor-based morphometry (TBM; Studholme et al., 2001, Leow et al., 2005a, Leow et al., 2005b).

The results provided in this paper concern 3D maps of the stability of different MRI imaging protocols and pre/post-processing techniques, in the context of mapping brain change using nonlinear image registration and TBM. Specifically, our goal was to determine which MR imaging sequences combined with which data correction methods were the most reproducible, and most reliable, resulting in least measurement variability. The foundation of our calibrations was based on the assumption that any serial MRI scan pair in this study should show minimal structural change related to aging or disease, and there should be no consistent change detected in a group of subjects scanned. This is plausible given that the elderly normal subjects were scanned twice using the same protocol, scanner and RF head coil over a very short interval (2 weeks). In individuals, there may still be minor changes due to subject-specific mechanical, circadian or tissue hydration effects on anatomy. There are also (non-pathological) sources of variability due to the interaction of the patient and the sequence/scanner. For example, subject movement is more likely with a longer sequence. Patient positioning is inevitably variable relative to the coil and scanner, which may have different impacts on the images depending on the scanner, sequence and coils. Differences among MRI scanning techniques were assessed by scanning the same subjects with four or five (depending on the MR system) different MRI pulse sequences in the same scanning session (IR-SPGR, MEDIC, high and low flip angle SPGR/FLASH and MP-RAGE). The low flip angle SPGR/FLASH images were not evaluated as an independent image type; they were used along with the high flip SPGR/FLASH scans to generate a Synthetic T1 image. Therefore, any regional structural difference picked up using TBM can be assumed to be random error or artifactual drift, related to geometric distortion of the scanner, uncorrected spatial distortions and variations in imaging signal or contrast-to-noise. Statistical analysis was therefore applied to maps of changes computed using TBM, providing baseline information on MRI imaging reliability, reproducibility and variability.