Computational protocol: Dissociable effects of methylphenidate, atomoxetine and placebo on regional cerebral blood flow in healthy volunteers at rest: A multi-class pattern recognition approach

Similar protocols

Protocol publication

[…] Scanning was performed on a General Electric Signa HDx 3T scanner and was timed to coincide with the peak plasma concentration for MPH and ATX (). Between 90 and 135 minutes post-dose, subjects rested quietly in the scanner while six whole-brain rCBF maps were acquired using a pulsed-continuous ASL sequence (pCASL; ). In this method, blood from the neck and base of the brain is labeled using a train of Hanning-shaped radio frequency (RF) pulses of 500 μs duration, and a time gap of 1000 μs between each Hanning pulse. The total duration of the pulse train is 1.5 s (s). A sequence of gradient pulses of similar duration and repetition rate was employed to obtain flow-driven adiabatic inversion. The highest gradient amplitude under the Hanning pulses and the average gradient intensity over the RF train duration, were 9 mT/m and 1 mT/m, respectively. These values were originally chosen to ensure that the adiabatic condition for inversion and the exclusion of the first aliased labeling plane away from the excitation bandwidth of the Hanning pulse, were both met (). In the control phase, the sign of alternate Hanning pulses was reversed, and the amplitudes of the gradient pulses were adjusted so that the net RF and gradient amplitudes over the 1.5 s irradiation were both zero. Thus, the magnetization transfer effect is compensated while achieving no inversion of arterial spins.Image acquisition was performed using a 3D interleaved spiral fast spin echo (FSE) readout () with parameters: TR = 4 s, TE = 32 ms, ETL = 64, 8 interleaves, spatial resolution = 1 × 1 × 3 mm. Three ‘control-labeled’ pairs were collected to produce the ‘perfusion weighted’ difference image.To quantify rCBF using this difference image, the sensitivity of the acquisition was calibrated to water at each voxel (). This is complicated by the spatial non-uniform sensitivity of the 8-channel coil employed for this work. The underlying tissue signal is used as an indicator of water sensitivity, and a water density in each voxel, or partition coefficient, is assumed. In the original methodology (), it was observed that the signal intensity in an inversion-prepared fluid-suppressed image was relatively constant for different tissues. This is likely because more complete recovery occurs for shorter T1 tissues, which tend to have lower water density. Using a neighborhood maximum algorithm to avoid regions with partial volume of suppressed fluid, a low resolution sensitivity map was created. This map was calibrated for water sensitivity by assuming the tissue was white matter with a water concentration of 0.735 g/ml () and a T1 of 900 ms, and using the equations for inversion recovery signal attenuation. By assuming that gray matter has a water concentration of 0.88 g/ml and a T1 of 1150 there was only a 5% calibration difference. This calibration produced a sensitivity map, C, equal to the fully relaxed MRI signal intensity produced by 1 g of water per milliliter of brain. With this co-registered sensitivity map C, we calculated cerebral blood flow (CBF) using the equation:CBF=ρb(Sc−Sl)2αCωaT1aexp−δT1a1−exp−tlT1awhere ρb is 1.05 g/ml (the density of brain tissue; ), α is the labeling efficiency (assumed to be 95% for labeling times 75% for background suppression; (), δ is 1.5 s (the post labeling delay; ) tl is 500 ms (the labeling duration), T1a is 1.4 ms (the T1 of arterial blood which was slightly lower than the value of ), ωa is 0.85 g/ml (the density of water in blood; ), Sl and Sc are the signal intensities in the labeled and control images, respectively. As is common in the ASL literature, this equation assumes that the labeled blood remains in the arterioles and capillaries and does not reach the tissue. The CBF quantification process does not alter the qualitative appearance of the images obtained by subtracting the label from the control image. The whole ASL pulse sequence, including the acquisition of calibration images, was performed in 6:08 min. After the acquisition of the pCASL scans, subjects performed a rewarded working memory task, which has been reported separately (). For each subject, a high-resolution T2-weighted FSE structural image was also acquired to assist registration of the pCASL scans to a common reference space with parameters: TR = 4.4 s, TE = 65 ms, FA = 90°, 36 × 4 mm thick oblique axial slices, in-plane resolution = 0.46 × 0.46 mm.Images were preprocessed using tools from the Statistical Parametric Mapping 5 (SPM5; and Functional Software Library (FSL; software packages. A three step procedure was employed to ensure maximally accurate registration of the pCASL image to a common reference image. First, extra-cerebral signal from the T2 structural scan was removed using the brain extraction tool included in FSL (BET; ) and the skull-stripped T2 image and its corresponding binary mask were co-registered to each pCASL image using SPM5. Second, the brain mask derived from the T2 image was applied to each pCASL image and the resulting skull stripped images were then co-registered back to the original T2 image (again with SPM5). Finally, the high resolution T2 image was used to compute SPM5 normalization parameters necessary to warp the image to the T2 MNI template provided with SPM5 and the resulting parameters were applied to the co-registered pCASL images in addition to the T2 image. Following normalization, each whole-brain pCASL image was spatially smoothed with an 8 mm isotropic Gaussian kernel and an average image was estimated for each subject and drug condition based on all scans. Since basal rCBF values are potentially different between participants, each image was then mean-centered within participants. In other words, a mean image was computed for each participant based on all images for that participant included in the classification problem and the mean was subtracted voxel-wise from each of the smoothed and averaged pCASL images. These mean-centered images were then reshaped into vectors and used as input to the classifiers. […]

Pipeline specifications

Software tools SPM, BET
Application Magnetic resonance imaging
Organisms Homo sapiens
Diseases Brain Stem Neoplasms
Chemicals Methylphenidate, Oxygen