NEUROLOGY/2013/561365- Appendix e-1 -Page 1 out of 2
BrainFDG-PET imaging and data analysis.
All patients underwentFDG-PET at rest, after intravenous injection of 170 MBq of 18FDG. They were positioned comfortably in a quiet, dimly lit room several minutes before 18FDG administration and during the uptake phase of FDG for at least 30 min, and were instructed not to speak, read or be otherwise active. Each acquisition included a Computed Tomography (CT) transmission scan of the head (50mAs lasting 16 seconds) followed by a three-dimensional (3D) static emission of 15 minutes using a Biograph Truepoint 64 PET/CT scanner (Siemens, Erlagen, Germany) with resolution of PET system of 4-5 mm full-width half-maximum (FWHM). PET sections were reconstructed using iterative ordered-subset expectation maximization (OSEM) algorithm, corrected for scatter and attenuation using density coefficients derived from the low dose CT scan of the head obtained with the same scanner, with the proprietary software.
Group analysis. A general linear model was used to perform the voxel-by-voxel statisticsby means ofStatistical Parametric Mapping (SPM) software, version 8 (SPM8, Wellcome Department of Cognitive Neurology, London, UK). The images have been spatially normalized into the PET template available in SPM8, and then smoothed using a FWHM12-mm Gaussian kernel to increase the signal-to-noise ratio and to account for subtle variations in anatomic structures. Normalized datasets from the 12 patients with X-ALD and from 27 age-matched healthy subjects (mean age: 39 years; range 23-63 years) were compared on a voxel-by-voxel basis using a one-way analysis of variance (ANOVA): the two groups of X-ALD patients and healthy controls were modeled as conditions, with age at PET scan as nuisance covariate. The effect of global differences in metabolism was removed by using the proportional scaling global mean to 50, with threshold masking relative to the default value of 0.8. Voxel-based statistical analysis was performed by applying a threshold of p0.01, false discovery rate (FDR)-corrected for significance, and considering clusters of at least 250 voxels. In addition, we performed voxel-wise correlation analysis between glucose metabolism and Millon Clinical Multiaxial Inventory – III scores byapplying a threshold of p0.05 for significance and considering cluster of at least 25 voxels. The coordinates of the most significant voxel in a cluster are reported as coordinates in the Montreal Neurological Institute (MNI) space.
Single-subject analysis. Single-subject analysis was performed by using the software Scenium, which was conceived to compare the metabolic changes of a single subject with a predefined reference database of 30 normal subjects. Details of the subject selection and preparation can be found in the work of Reiman and colleagues.e1Statistics is computed and displayed on a voxelbyvoxel basis. Each normal scan is transformed using a deformable registration and then smoothed using a FWHM 12-mm isotropic Gaussian kernel. The patient scan is first registered into the space of the standard geometry from the Montreal Neurological Institute (MNI) and smoothed using the FWHM 12-mm Gaussian kernel, as the normal brains. Subsequently, Scenium computed the number of standard deviations from the mean of each voxel, where the mean and standard deviation are taken from the corresponding voxel in the normal brain. According to the model, this statistics follows a t-distribution with 30 degrees of freedom. The Lateral Cortical Views of Scenium visualized the entire cortex in a single image. It is formed by projecting the cortex to a depth of 20 mm onto a single flat plane. The brain has been split at the front and the cortex, then opened out and flattened in such a way that the posterior parts are shown in the center and the anterior parts are at the outer edges of the image; the right and left sides of the cortex are displayed side by side with anatomical markers displayed for orientation. The color map gives a visual overview of uptake statistics. The statistic color spectrum represents the number of standard deviations (i.e., the Z-score) from normal for the pixel value, calculated relatively to a t-distribution with 30 degrees of freedom. Z-score >2 or <-2 is considered significant.