Effects of spatial fMRI resolution on the classification of naturalistic movies.

Studies involving multivariate pattern analysis (MVPA) of BOLD fMRI data generally attribute the success of the information-theoretic approach to BOLD signal contrast on the fine spatial scale of millimeters facilitating the classification or decoding of perceptual stimuli. However, to date MVPA studies that have actually explored fMRI resolutions at less than 2 mm voxel size are rare and limited to small sets of unnatural stimuli (like visual gratings) as well as specific sub-regions of the brain, notably the primary somatosensory cortices. To investigate what spatial scale best supports high information extraction under more general conditions this study combined naturalistic movie stimuli with high-resolution fMRI at 7 T and linear discriminant analysis (LDA) of global and local BOLD signal patterns. Contrary to predictions, LDA and similar classifiers reached a maximum in classification accuracy (CA) at a smoothed resolution close to 3 mm, well above the 1.2 mm voxel size of the fMRI acquisition. Maximal CAs around 90% were contingent upon global fMRI signal patterns comprising 4 k-16 k of the most reactive voxels distributed sparsely throughout the occipital and ventro-temporal cortices. A Searchlight analysis of local fMRI patterns largely confirmed the global results, but also revealed a small subset of brain regions in early visual cortex showing limited increases in CA with higher resolution. Principal component analysis of the global and local fMRI signal patterns suggested that reproducible neuronal contributions were spatially auto-correlated and smooth, while other components of higher spatial frequency were likely related to physiological noise and responsible for the reduced CA at higher resolution. Systematic differences between experiments and subjects suggested that higher CA was significantly correlated with more consistent behavior revealed by eye tracking. Thus, the optimal resolution of fMRI data for MVPA was mainly limited by physiological noise of high spatial frequency as well as behavioral (in-)consistency.

Investigating lipids as a source of chemical exchange-induced MRI frequency shifts.

While magnetic susceptibility is a major contributor to NMR resonance frequency variations in the human brain, a substantial contribution may come from the chemical exchange of protons between water and other molecules. Exchange-induced frequency shifts fe have been measured in tissue and protein solutions, but relatively lipid-rich white matter (WM) has a larger fe than gray matter, suggesting that lipids could contribute. Galactocerebrosides (GC) are a prime candidate as they are abundant in WM and susceptible to exchange. To investigate this, fe was measured in a model of WM lipid membranes in the form of multilamellar vesicles (MLVs), consisting of a 1:2 molar ratio of GC and phospholipids (POPC), and in MLVs with POPC only. Chemical shift imaging with 15% volume fraction of dioxane, an internal reference whose protons are assumed not to undergo chemical exchange, was used to remove susceptibility-induced frequency shifts in an attempt to measure fe in MLVs at several lipid concentrations. Initial analysis of these measurements indicated a necessity to correct for small unexpected variations in dioxane concentration due to its effect on the water frequency shift. To achieve this, the actual dioxane concentration was inferred from spectral analysis and its additional contribution to fe was removed through separate experiments which showed that the water-dioxane frequency shift depended linearly on the dioxane concentration at low concentrations with a proportionality constant of -0.021 ± 0.002 ppb/mM in agreement with published experiments. Contrary to expectations and uncorrected results, for GC + POPC vesicles, the dependence of the corrected fe on GC concentration was insignificant (0.023 ± 0.037 ppb/mM; r2 = 0.085, p > 0.57), whereas for the POPC-only vesicles a small but significant linear increase with POPC concentration was found: 0.044 ± 0.008 ppb/mM (r2 = 0.877, p < 0.01). These findings suggest that the exchange-induced contribution of lipids to frequency contrast in WM may be small. Published 2016. This article is a U.S. Government work and is in the public domain in the USA.

The PRESTO technique for fMRI.

In the early days of BOLD fMRI, the acquisition of T(2)(*) weighted data was greatly facilitated by rapid scan techniques such as EPI. The latter, however, was only available on a few MRI systems that were equipped with specialized hardware that allowed rapid switching of the imaging gradients. For this reason, soon after the invention of fMRI, the scan technique PRESTO was developed to make rapid T(2)(*) weighted scanning available on standard clinical scanners. This method combined echo shifting, which allows for echo times longer than the sequence repetition time, with acquisition of multiple k-space lines per excitation. These two concepts were combined in order to achieve a method fast enough for fMRI, while maintaining a sufficiently long echo time for optimal contrast. PRESTO has been primarily used for 3D scanning, which minimized the contribution of large vessels due to inflow effects. Although PRESTO is still being used today, its appeal has lessened somewhat due to increased gradient performance of modern MRI scanners. Compared to 2D EPI, PRESTO may have somewhat reduced temporal stability, which is a disadvantage for fMRI that may not outweigh the advantage of reduced inflow effects provided by 3D scanning. In this overview, the history of the development of the PRESTO is presented, followed by a qualitative comparison with EPI.

Tissue-specific imaging is a robust methodology to differentiate in vivo T1 black holes with advanced multiple sclerosis-induced damage.

BACKGROUND AND PURPOSE: Brains of patients with multiple sclerosis (MS) characteristically have "black holes" (BHs), hypointense lesions on T1-weighted (T1W) spin-echo (SE) images. Although conventional MR imaging can disclose chronic BHs (CBHs), it cannot stage the degree of their pathologic condition. Tissue-specific imaging (TSI), a recently introduced MR imaging technique, allows selective visualization of white matter (WM), gray matter (GM), and CSF on the basis of T1 values of classes of tissue. We investigated the ability of TSI-CSF to separate CBHs with longer T1 values, which likely represent lesions containing higher levels of destruction and unbound water. MATERIALS AND METHODS: Eighteen patients with MS, who had already undergone MR imaging twice (24 months apart) on a 1.5T scanner, underwent a 3T MR imaging examination. Images acquired at 1.5T included sequences of precontrast and postcontrast T1W SE, T2-weighted (T2W) SE, and magnetization transfer (MT). Sequences obtained at 3T included precontrast and postcontrast T1W SE, T2W SE, T1 inversion recovery prepared fast spoiled gradient recalled-echo (IR-FSPGR) and TSI. A BH on the 3T-IR-FSPGR was defined as a CBH if seen as a hypointense, nonenhancing lesion with a corresponding T2 abnormality for at least 24 months. CBHs were separated into 2 groups: those visible as hyperintensities on TSI-CSF (group A), and those not appearing on the TSI-CSF (group B). RESULTS: Mean MT ratios of group-A lesions (0.22 +/- 0.06, 0.13-0.35) were lower (F(1,13) = 60.39; P < .0001) than those of group-B lesions (0.32 +/- 0.03, 0.27-0.36). CONCLUSIONS: Group-A lesions had more advanced tissue damage; thus, TSI is a potentially valuable method for qualitative and objective identification.

Pittfalls of MRI measurement of white matter perfusion based on arterial spin labeling.

Although arterial spin labeling (ASL) MRI has been successfully applied to measure gray matter (GM) perfusion in vivo, accurate detection of white matter (WM) perfusion has proven difficult. Reported literature values are not consistent with each other or with perfusion measured with other modalities. In this work, the cause of these inconsistencies is investigated. The results suggest that WM perfusion values are substantially affected by the limited image resolution and by signal losses caused by the long transit times in WM, which significantly affect the label. From gadolinium diethylenetriamine pentaacetic acid (Gd-DTPA) bolus-tracking experiments (N=6), it is estimated that the transit time can be several seconds long in deep WM. Furthermore, simulations show that even at a spatial resolution of 7 microl voxel size, contamination by the GM signals can exceed 40% of the actual WM signal. From 10-min long flow-sensitive alternating inversion recovery ASL (FAIR-ASL) measurements at 3T in normal subjects (N=7), using highly sensitive detectors, it is shown that single-voxel (7 mul) deep WM perfusion values have an signal-to-noise ratio (SNR) less than 1. The poor sensitivity and heterogeneous transit time limit the applicability of ASL for measurement of perfusion in WM.

Real-time shimming to compensate for respiration-induced B0 fluctuations.

In MRI of human brain, the respiratory cycle can induce B0-field fluctuations through motion of the chest and fluctuations in local oxygen concentration. The associated NMR frequency changes can affect the MRI data in various ways and lead to temporal signal fluctuations, and image artifacts such as ghosting and blurring. Since the size of the effect scales with magnetic field strength, artifacts become particularly problematic at fields above 3.0T. Furthermore, the spatial dependence of the B0-field fluctuations complicates their correction. In this work, a new method is presented that allows compensation of field fluctuations by modulating the B0 shims in real time. In this method, a reference scan is acquired to measure the spatial distribution of the B0 effect related to chest motion. During the actual scan, this information is then used, together with chest motion data, to apply compensating B0 shims in real time. The method can be combined with any type of scan without modifications to the pulse sequence. Real-time B0 shimming is demonstrated to substantially improve the phase stability of EPI data and the image quality of multishot gradient-echo (GRE) MRI at 7T.

fMRI study of effort and information processing in a working memory task.

It is unclear how effort translates into brain function. In this study we endeavored to identify the activity in a working memory task that is related to the allocation of mental resources. Such activity, if present, would be a likely candidate to explain how effort works in terms of brain function. Eleven healthy participants performed a Sternberg task with a memory-set of one, three, or five consonants in an fMRI study. Probe stimuli were either one consonant or one digit. We expected digits to be processed automatically and consonants to require working memory. Because the probe type was unpredictable and subjects had to respond as fast as possible, we expected subjects to allocate mental resources on the basis of the memory-set size, not the probe type. Accordingly, we anticipated that activity in regions involved in effort would be a function of the size of the memory-set, but independent of the type of probe. We found that the reaction-time for digits increased in line with our expectation of automatic processing and the reaction time for letters increased in line with our expectation of controlled processing. fMRI revealed that activity in the right ventral-prefrontal cortex changed as a function of effort. The ventral anterior cingulate cortex and hypothalamus showed reduced activity as a function of effort. Activity in regions regarded as pivotal for working memory (among others, the left dorsolateral prefrontal cortex, anterior cingulate cortex) appeared to be predominantly related to information processing and not involved in effort.

Proton MR spectroscopic imaging in multiple sclerosis.

We studied 24 patients with multiple sclerosis (MS) by proton magnetic resonance spectroscopic imaging (1H-MRSI) to assess the neurochemical pathology of the white-matter lesions (WML) and normal-appearing white matter (NAWM). Our 1H-MRSI technique allowed simultaneous measurement of N-acetylaspartate (NAA), choline-containing compounds (Cho), and creatine plus phosphocreatine (Cr) signal intensities from four 15-mm slices divided into 0.84 ml single-volume elements. In WML we found significantly lower NAA/Cr and NAA/Cho ratios and a significantly higher Cho/Cr ratio than in NAWM or control white matter. In NAWM, NAA/Cr and Cho/Cr were significantly lower than in control white matter. 1H-MRSI was compatible with damage to myelin in WML, and with axonal damage and/or dysfunction in WML and NAWM. These findings extend data on involvement of NAWM in MS beyond the abnormalities visible on MRI.

High-sensitivity single-shot perfusion-weighted fMRI.

A method is presented for measurement of perfusion changes during brain activation using a single-shot pulsed spin labeling technique. By employing a double-inversion labeling strategy, stationary tissue (background) signal was suppressed while minimally affecting perfusion sensitivity. This allowed omission of the otherwise required reference scan, resulting in twofold-improved temporal resolution. The method was applied to visual and motor cortex activation studies in humans, and compared to standard FAIR-type perfusion labeling techniques. Experiments performed at 1.5T and 3.0T indicate a close to 90% suppression of background signal, at a cost of an 11% and 9%, respectively, reduction in perfusion signal. Combined with the twofold increase in signal averaging, and a reduction in background signal fluctuations, this resulted in a 64% (1.5T, N = 3) and a 128% (3T, N = 4) overall improvement in sensitivity for the detection of activation-related perfusion changes. Magn Reson Med 46:88-94, 2001. Published 2001 Wiley-Liss, Inc.

Simultaneous BOLD/perfusion measurement using dual-echo FAIR and UNFAIR: sequence comparison at 1.5T and 3.0T.

Functional MRI (fMRI) studies designed for simultaneously measuring Blood Oxygenation Level Dependent (BOLD) and Cerebral Blood Flow (CBF) signal often employ the standard Flow Alternating Inversion Recovery (FAIR) technique. However, some sensitivity is lost in the BOLD data due to inherent T1 relaxation. We sought to minimize the preceding problem by employing a modified UN-inverted FAIR (UNFAIR) technique, which (in theory) should provide identical CBF signal as FAIR with minimal degradation of the BOLD signal. UNFAIR BOLD maps acquired from human subjects (n = 8) showed significantly higher mean z-score of approximately 17% (p < 0.001), and number of activated voxels at 1.5T. On the other hand, the corresponding FAIR perfusion maps were superior to the UNFAIR perfusion maps as reflected in a higher mean z-score of approximately 8% (p = 0.013), and number of activated voxels. The reduction in UNFAIR sensitivity for perfusion is attributed to increased motion sensitivity related to its higher background signal, and, T2 related losses from the use of an extra inversion pulse. Data acquired at 3.0T demonstrating similar trends are also presented.

Proton MR spectroscopic imaging without water suppression.

To improve reproducibility in proton magnetic resonance (MR) spectroscopic imaging in human brain, simultaneous acquisition of the internal water reference and metabolite signals was evaluated. Measurements in healthy volunteers showed that the increase in dynamic range from signal oversampling was sufficient to avoid digitization errors. In addition, use of singular value decomposition techniques and finite impulse response filters proved effective in separating water and metabolite signals and providing estimates of the metabolite concentrations.

A protocol for assessing subtraction errors of arterial spin-tagging perfusion techniques in human brain.

A protocol for assessing signal contributions from static tissue (subtraction errors) in perfusion images acquired with arterial spin-labeling (ASL) techniques in human brain is proposed. The method exploits the reduction of blood T(1) caused by the clinically available paramagnetic contrast agent, gadopentetate dimeglumine (Gd-DTPA). The protocol is demonstrated clinically with multislice FAIR images acquired before, during, and after Gd-DTPA administration using a range of selective inversion widths. Perfusion images acquired postcontrast for selective inversion widths large enough (threshold) to avoid interaction with the imaging slice had signal intensities reduced to noise level, as opposed to subtraction errors manifested on images acquired using inversion widths below the threshold. The need for these experiments to be performed in vivo is further illustrated by comparison with phantom results. The protocol allows a one-time calibration of relevant ASL parameters (e.g., selective inversion widths) in vivo, which may otherwise cause subtraction errors. Magn Reson Med 43:896-900, 2000. Published 2000 Wiley-Liss, Inc.

High-speed interlaced spin-echo magnetic resonance imaging.

A new method is introduced for increasing the efficiency in multislice single spin-echo MRI. The method interlaces the excitation and measurement of different slices, resulting in an effective use of the echo delay time between RF excitation and reception. Under certain conditions, the method allows for scan time reduction compared to standard single spin-echo MRI, in particular for long echo times. The technique is demonstrated in examples of brain scans, indicating that a substantial increase is scan speed can be achieved without loss in image signal-to-noise ratio or contrast. Potential applications include perfusion imaging using T(2)-contrast agents, as well as BOLD-based functional imaging. Magn Reson Med 43:905-908, 2000. Published 2000 Wiley-Liss, Inc.

Proton magnetic resonance spectroscopic imaging in children with recurrent primary brain tumors.

PURPOSE: Proton magnetic resonance spectroscopic imaging ((1)H-MRSI) is a noninvasive technique for spatial characterization of biochemical markers in tissues. We measured the relative tumor concentrations of these biochemical markers in children with recurrent brain tumors and evaluated their potential prognostic significance. PATIENTS AND METHODS: (1)H-MRSI was performed on 27 children with recurrent primary brain tumors referred to our institution for investigational drug trials. Diagnoses included high-grade glioma (n = 10), brainstem glioma (n = 7), medulloblastoma/peripheral neuroectodermal tumor (n = 6), ependymoma (n = 3), and pineal germinoma (n = 1). (1)H-MRSI was performed on 1. 5-T magnetic resonance imagers before treatment. The concentrations of choline (Cho) and N-acetyl-aspartate (NAA) in the tumor and normal brain were quantified using a multislice multivoxel method, and the maximum Cho:NAA ratio was determined for each patient's tumor. RESULTS: The maximum Cho:NAA ratio ranged from 1.1 to 13.2 (median, 4.5); the Cho:NAA ratio in areas of normal-appearing brain tissue was less than 1.0. The maximum Cho:NAA ratio for each histologic subtype varied considerably; approximately equal numbers of patients within each tumor type had maximum Cho:NAA ratios above and below the median. Patients with a maximum Cho:NAA ratio greater than 4.5 had a median survival of 22 weeks, and all 13 patients died by 63 weeks. Patients with a Cho:NAA ratio less than or equal to 4.5 had a projected survival of more than 50% at 63 weeks. The difference was statistically significant (P =.0067, log-rank test). CONCLUSION: The maximum tumor Cho:NAA ratio seems to be predictive of outcome in children with recurrent primary brain tumors and should be evaluated as a prognostic indicator in newly diagnosed childhood brain tumors.

Multislice perfusion imaging in human brain using the C-FOCI inversion pulse: comparison with hyperbolic secant.

Perfusion studies based on pulsed arterial spin labeling have primarily applied hyperbolic secant (HS) pulses for spin inversion. To optimize perfusion sensitivity, it is highly desirable to implement the HS pulse with the same slice width as the width of the imaging pulse. Unfortunately, this approach causes interactions between the slice profiles and manifests as residual signal from static tissue in the resultant perfusion image. This problem is currently overcome by increasing the selective HS width relative to the imaging slice width. However, this solution increases the time for the labeled blood to reach the imaging slice (transit time), causing loss of perfusion sensitivity as a result of T(1) relaxation effects. In this study, we demonstrate that the preceding problems can be largely overcome by use of the C-shaped frequency offset corrected inversion (FOCI) pulse [Ordidge et al., Magn Reson Med 1996;36:562]. The implementation of this pulse for multislice perfusion imaging on the cerebrum is presented, showing substantial improvement in slice definition in vivo compared with the HS pulse. The sharper FOCI profile is shown to reduce the physical gap (or "safety margin") between the inversion and imaging slabs, resulting in a significant increase in perfusion signal without residual contamination from static tissue. The mean +/- SE (n = 6) gray matter perfusion-weighted signal (DeltaM/M(o)) without the application of vascular signal suppression gradients were 1.19 +/- 0. 10% (HS-flow-sensitive alternating inversion recovery [FAIR]), and 1. 51 +/- 0.11% for the FOCI-FAIR sequence. The corresponding values with vascular signal suppression were 0.64 +/- 0.14%, and 0.91 +/- 0. 08% using the HS- and FOCI-FAIR sequences, respectively. Compared with the HS-based data, the FOCI-FAIR results correspond to an average increase in perfusion signal of up to between 26%-30%. Magn Reson Med 42:1098-1105, 1999.

Investigation of low frequency drift in fMRI signal.

Low frequency drift (0.0-0.015 Hz) has often been reported in time series fMRI data. This drift has often been attributed to physiological noise or subject motion, but no studies have been done to test this assumption. Time series T*2-weighted volumes were acquired on two clinical 1.5 T MRI systems using spiral and EPI readout gradients from cadavers, a normal volunteer, and nonhomogeneous and homogeneous phantoms. The data were tested for significant differences (P = 0.001) from Gaussian noise in the frequency range 0.0-0.015 Hz. The percentage of voxels that were significant in data from the cadaver, normal volunteer, nonhomogeneous and homogeneous phantoms were 13.7-49.0%, 22.1-61.9%, 46.4-68.0%, and 1.10%, respectively. Low frequency drift was more pronounced in regions with high spatial intensity gradients. Significant drifting was present in data acquired from cadavers and nonhomogeneous phantoms and all pulse sequences tested, implying that scanner instabilities and not motion or physiological noise may be the major cause of the drift.

Comparison of 3D BOLD functional MRI with spiral acquisition at 1.5 and 4.0 T.

In order to investigate the merit of high field strength for BOLD-contrast-based functional magnetic resonance imaging (fMRI) studies, multishot gradient-echo fMRI experiments during motor cortex activation were performed on 1.5- and 4.0-T scanners with equivalent hardware, on the same volunteers. In these studies, artifactual vascular enhancement related to inflow effects was minimized, and large brain areas were covered by using a 3D scan technique. Temporal signal stability was optimized by using spiral readout gradients. The sensitivity for detection of activated regions was assessed by measuring the number of "activated voxels" and their average t score in predefined regions of interest. When comparing fMRI experiments with the same total scan time, performed on six subjects, and with acquisition parameters optimized for each field strength separately, the 4.0-T scanner proved to give superior results, with a 70% greater number of activated voxels and a 20% higher average t score for the activated voxels.

Optimization of fast acquisition methods for whole-brain relative cerebral blood volume (rCBV) mapping with susceptibility contrast agents.

Fast gradient-echo magnetic resonance scan techniques with spiral and rectilinear (echoplanar) k-space trajectories were optimized to perform bolus-tracking studies of human brain. Cerebral hemodynamics were studied with full brain coverage, a spatial resolution of 4 mm, and a temporal resolution of 2 seconds. The sensitivity of the techniques to detect image signal-intensity changes during the first pass of the contrast agent was studied at a range of TEs using dedicated experiments. For single-shot versions of spiral scanning and echoplanar imaging techniques with a 0.1-mmol/kg injection of gadolinium diethylenetriamine pentaacetic acid using a mechanical injector at 10 mL/sec under 1.5 T, the maximum sensitivity was obtained at TEs between 35 and 45 msec. At TEs less than 35 msec, signal-intensity artifacts were observed in the images. Analysis of the point-spread function revealed that susceptibility changes induced by the contrast agent can result in signal shifts to neighboring voxels. These artifacts are attributed to susceptibility-related signal changes during the acquisition window.

Technical solution for an interactive functional MR imaging examination: application to a physiologic interview and the study of cerebral physiology.

Studies with functional magnetic resonance (MR) imaging produce large unprocessed raw data sets in minutes. The analysis usually requires transferring of the data to an off-line workstation, and this process frequently occurs after the subject has left the MR unit. The authors describe a hardware configuration and processing software that captures whole-brain raw data files as they are being produced from the MR unit. It then performs the reconstruction, registration, and statistical analysis, and displays the results in seconds after completion of the MR image acquisition.

Serial whole-brain magnetization transfer imaging in patients with relapsing-remitting multiple sclerosis at baseline and during treatment with interferon beta-1b.

BACKGROUND AND PURPOSE: To determine whether occult disease fluctuates with macroscopic lesions during the natural history of multiple sclerosis (MS) and whether therapeutic interventions affect occult disease, we performed serial monthly magnetization transfer (MT) imaging in patients with relapsing-remitting MS in a crossover trial with interferon beta-lb. METHODS: Serial whole-brain magnetization transfer ratios (MTRs) in eight patients with relapsing-remitting MS and in four control subjects were plotted as normalized histograms, and MTR parameters were compared with contrast-enhancing lesions and bulk white matter lesion load. RESULTS: In patients with relapsing-remitting MS, the histographic peak of 0.25+/-0.01 and the histographic mean of 0.21+/-0.01 were statistically lower than corresponding values in control subjects, in whom the histographic peak was 0.27+/-0.01 and the histographic mean was 0.23+/-0.01. When histograms (with MTRs ranging from 0.0 to 0.5) were analyzed by quartiles (quartile 1 to quartile 4) based on histographic area, voxels with low MTRs in quartile 1 (0 to 0.12) increased during the baseline period and corresponded to bulk white matter lesion load. Interferon beta-lb reduced enhancing lesions by 91% and mean bulk white matter lesion load by 15%, but had no effect on MTR in this patient cohort. CONCLUSION: Occult disease in normal-appearing white matter of patients with relapsing-remitting MS measured by MTR parallels the waxing and waning pattern of enhancing lesions and bulk white matter lesion load during the baseline period. MTR is not altered by interferon beta-lb, which raises the possibility of ongoing disease in normal-appearing white matter (not detected by conventional MR sequences).