Reliability of MRI-derived measurements of human cerebral cortical thickness: The effects of field strength, scanner upgrade and manufacturer
Introduction
Techniques that enable the in vivo MRI-derived quantitative measurement of properties of the human cerebral cortex, such as thickness, are beginning to demonstrate important potential applications in basic and clinical neuroscience. Changes in the gray matter that makes up the cortical sheet are manifested in normal aging (Jack et al., 1997, Salat et al., 1999, Salat et al., 2004, Sowell et al., 2003, Sowell et al., 2004), Alzheimer's disease (Dickerson et al., 2001, Thompson et al., 2003, Lerch et al., 2005), Huntington's disease (Rosas et al., 2002), corticobasal degeneration (Boeve et al., 1999), amyotrophic lateral sclerosis (Kiernan and Hudson, 1994), multiple sclerosis (Sailer et al., in press) and schizophrenia (Thompson et al., 2001, Kuperberg et al., 2003, Narr et al., 2005). Progressive thinning of the cortex follows a disease-specific regional pattern in certain diseases, such as Alzheimer's disease (Thompson et al., 2003); thus, in vivo cortical thickness measures could be useful as a biomarker of the evolution of the disease. Longitudinal imaging-based biomarkers of disease progression will likely be of great utility in evaluating the efficacy of disease-modifying therapies (Dickerson and Sperling, 2005).
Measurement of cortical thickness from MRI data is a non-trivial task. Manual thickness measurements are difficult to obtain due to the highly convoluted nature of the cortex. It can take a trained anatomist several days to manually label a high-resolution set of MR brain images, and even this labor-intensive procedure allows only the measurement of cortical volume, not cortical thickness, because the cortical thickness is a property that can only be properly measured if the location and orientation of both the gray/white and pial surfaces are known. To facilitate automatic thickness measurement, many computerized methods have been proposed in the literature for segmenting the cortex and finding the cortical surfaces from MRI data (Dale et al., 1999, Joshi et al., 1999, MacDonald et al., 1999, Xu et al., 1999, Zeng et al., 1999, Van Essen et al., 2001, Shattuck and Leahy, 2002, Sowell et al., 2003, Barta et al., 2005, Han et al., 2005a).
Although the validation of MRI-derived cortical thickness measurements has been performed against regional manual measurements derived from both in vivo and post-mortem brain scans (Rosas et al., 2002, Kuperberg et al., 2003, Salat et al., 2004), the reliability of measures of this fundamental morphometric property of the brain has received relatively little systematic investigation (Fischl and Dale, 2000, Rosas et al., 2002, Kuperberg et al., 2003, Sowell et al., 2004, Lerch and Evans, 2005). Most of these studies approach reliability by comparing thickness measurements across different subjects or by performing repeated scans on a few subjects acquired within the same scan session or within very short scan intervals (for example, the subjects were removed from the scanner and then scanned again in 5 min (Sowell et al., 2004)). This approach may greatly underestimate the sources of variability within and between studies.
Variability in MRI-derived morphometric measures may result from subject-related factors, such as hydration status (Walters et al., 2001), instrument-related factors, such as field strength, scanner manufacturer or pulse sequence, or data-processing-related factors, including not only software package but also the parameters chosen for analysis. All of these factors may contribute to differences between typical cross-sectional studies (e.g., when interpreting differences between two studies of patients with Alzheimer's disease vs. controls scanned at a single time point on one scanner). Longitudinal studies of normal development, aging or disease progression face additional challenges associated with both subject-related factors as well as instrument-related factors (e.g., major scanner upgrades). For multi-site studies, it is critical to understand and adjust for instrument-related differences between sites, such as scanner manufacturer, field strength and other hardware components. Finally, longitudinal multi-center studies, such as the Alzheimer's Disease Neuroimaging Initiative, must contend with all of these factors while attempting to detect subtle effects. Thus, detailed quantitative data regarding the degree to which each of the factors outlined above contribute to variability in cortical thickness (and other measures) could be very helpful for both study design and interpretation. Unfortunately, little work in this area has been performed.
Specifically, it is not yet clear how cortical thickness measures vary as a function of MRI instrument-related factors, such as field strength, scanner manufacturer and scanner software and hardware upgrades. Knowledge of the degree to which different MRI instrument-related factors affect the reliability of cortical thickness measures is essential for the interpretation of these measures in basic and clinical neuroscientific studies. Furthermore, this knowledge is critical if cortical thickness measures are to find applications as biomarkers in clinical trials of putative treatments for neurodegenerative or other neuropsychiatric diseases.
We undertook this study to evaluate the reliability of a cortical thickness measurement method both within and across different scanner platforms and field strengths, with the goal of quantitatively identifying the factors that are the greatest contributors to cortical thickness variability. Two groups of test–retest data sets were acquired and analyzed. In the first data set, 15 healthy older subjects were scanned four times at 2-week intervals on three different scanner platforms (test scan on Siemens Sonata 1.5 T, re-test scan on Siemens Sonata 1.5 T, cross-site scan on GE Signa 1.5 T, cross-field-strength scan on Siemens Trio 3 T). Older participants were studied so that anatomical variability related to atrophy and age-related signal changes was represented. The 2-week interval was chosen so that elements of variability related to subject hydration status and minor instrument drift would be included, which may be artificially minimized when the test–retest interval is several min to ∼1 day. First, the test–retest reliability of cortical thickness measurements was investigated from the two Siemens Sonata sessions. Next, analyses were performed on the effects of various instrument-related factors, including: a) different MR manufacturer (Siemens vs. GE); b) different field strength (1.5 T vs. 3 T); c) different pulse sequences (MPRAGE versus multiple flip angle, multi-echo FLASH); d) different number of data acquisitions (one MPRAGE vs. two averaged MPRAGE acquisitions). Finally, effects of several data processing-related factors were analyzed, including: a) different levels of spatial smoothing of the raw thickness maps; and b) different processing schemes (cross-sectional versus longitudinal).
The second data set consisted of 5 healthy younger subjects scanned repeatedly before and after a major scanner upgrade, with the goal of evaluating the reliability of thickness measurements in longitudinal studies that contend with scanner upgrades.
In this study, thickness measurements were performed using the FreeSurfer software package, which is an automated method for cortical surface reconstruction and thickness computation. Although comparison with other thickness measurement methods is beyond the scope of this paper, the effects of several aspects of the processing system within FreeSurfer were studied as noted above.
Section snippets
Data acquisition
Two groups of test–retest data sets were acquired and analyzed to characterize the reliability of cortical thickness estimation.
The first group of test–retest data consists of MRI scans acquired from 15 healthy older subjects (age between 66 and 81 years; mean: 69.5 years; SD: 4.8 years. 8 males, 7 females). All participants provided informed consent in accordance with the Human Research Committee of Massachusetts General Hospital. Each subject underwent 4 scan sessions at 2-week intervals (two
Reliability of global cortical thickness measure
Global thickness measures for each subject are presented in Table 1, where the mean and standard deviation were computed over the whole cortical surface combining left and right hemispheres. Overall, the range of cortical thickness values is similar for each individual subject across different scan sessions. The reproducibility of global mean cortical thickness is further demonstrated in Fig. 1 for the four test–retest comparisons. As can be seen, the global mean thickness is an exceedingly
Conclusion
The purpose of this study is to evaluate the reliability (precision) of an automated thickness measurement method both within- and across-scanner platforms and field strength. We also evaluated the effects on thickness measurement reliability of different imaging acquisition protocols (including number of acquisitions and imaging sequences) and different data processing or post-processing (smoothing of thickness map) schemes. Finally, we investigated the impact of a scanner upgrade on thickness
Acknowledgments
This research was supported by the following grants: a) NCRR Morphometry Biomedical Informatics Research Network (U24 RR021382), b) NCRR P41-RR14075 and RO1-RR16594-01A1, c) Pfizer Inc., d) the NIA (K23-AG22509 and P01-AG04953) and e) the MIND Institute.
References (48)
- et al.
Cortical thickness analysis in autism with heat kernel smoothing
NeuroImage
(2005) - et al.
Cortical surface-based analysis: I. Segmentation and surface reconstruction
NeuroImage
(1999) - et al.
Neuroimaging biomarkers for clinical trials of disease-modifying therapies in Alzheimer's disease
NeuroRx
(2005) - et al.
MRI-derived entorhinal and hippocampal atrophy in incipient and very mild Alzheimer's disease
Neurobiol. Aging
(2001) - et al.
Cortical surface-based analysis: II. Inflation, flattening, and a surface-based coordinate system
NeuroImage
(1999) - et al.
Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain
Neuron
(2002) - et al.
Improved optimization for the robust and accurate linear registration and motion correction of brain images
NeuroImage
(2002) - et al.
Brain segmentation and the generation of cortical surfaces
NeuroImage
(1999) - et al.
Cortical thickness analysis examined through power analysis and a population simulation
NeuroImage
(2005) - et al.
A hybrid approach to the skull stripping problem in MRI
NeuroImage
(2004)
BrainSuite: an automated cortical surface identification tool
Med. Image Anal.
More than meets the eye: significant regional heterogeneity in human cortical T1
Magn. Reson. Imaging
On-line automatic slice positioning for brain MR imaging
NeuroImage
A stochastic model for studying the laminar structure of cortex from MRI
IEEE Trans. Med. Imag.
Pathologic heterogeneity in clinically diagnosed corticobasal degeneration
Neurology
Anatomy of the Cortex
Statistical Power Analysis for the Behavioral Sciences
Pattern Classification
Measuring the thickness of the human cerebral cortex from magnetic resonance images
Proc. Natl. Acad. Sci. U. S. A.
High-resolution intersubject averaging and a coordinate system for the cortical surface
Hum. Brain Mapp.
Automated manifold surgery: constructing geometrically accurate and topologically correct models of the human cerebral cortex
IEEE Trans. Med. Imag.
Sequence-independent segmentation of magnetic resonance images
NeuroImage
Automatically parcellating the human cerebral cortex
Cereb. Cortex
Cited by (1233)
Growing importance of brain morphometry analysis in the clinical routine: The hidden impact of MR sequence parameters
2024, Journal of NeuroradiologyA preliminary investigation into cortical structural alterations in adolescents with nonsuicidal self-injury
2023, Psychiatry Research - NeuroimagingThe effect of processing pipelines, input images and age on automatic cortical morphology estimates
2023, Computer Methods and Programs in Biomedicine
- 1
These authors contributed equally to this work.