Roland N. Boubela

Beyond Noise: Using Temporal ICA to Extract Meaningful Information from High-Frequency fMRI Signal Fluctuations during Rest

Analysis of resting-state networks using fMRI usually ignores high-frequency fluctuations in the BOLD signal – be it because of low TR prohibiting the analysis of fluctuations with frequencies higher than 0.25 Hz (for a typical TR of 2 s), or because of the application of a bandpass filter (commonly restricting the signal to frequencies lower than 0.1 Hz). While the standard model of convolving neuronal activity with a hemodynamic response function suggests that the signal of interest in fMRI is characterized by slow fluctuation, it is in fact unclear whether the high-frequency dynamics of the signal consists of noise only. In this study, 10 subjects were scanned at 3 T during 6 min of rest using a multiband EPI sequence with a TR of 354 ms to critically sample fluctuations of up to 1.4 Hz. Preprocessed data were high-pass filtered to include only frequencies above 0.25 Hz, and voxelwise whole-brain temporal ICA (tICA) was used to identify consistent high-frequency signals. The resulting components include physiological background signal sources, most notably pulsation and heart-beat components, that can be specifically identified and localized with the method presented here. Perhaps more surprisingly, common resting-state networks like the default-mode network also emerge as separate tICA components. This means that high-frequency oscillations sampled with a rather T1-weighted contrast still contain specific information on these resting-state networks to consistently identify them, not consistent with the commonly held view that these networks operate on low-frequency fluctuations alone. Consequently, the use of bandpass filters in resting-state data analysis should be reconsidered, since this step eliminates potentially relevant information. Instead, more specific methods for the elimination of physiological background signals, for example by regression of physiological noise components, might prove to be viable alternatives.

The spectral diversity of resting-state fluctuations in the human brain.

In order to assess whole-brain resting-state fluctuations at a wide range of frequencies, resting-state fMRI data of 20 healthy subjects were acquired using a multiband EPI sequence with a low TR (354 ms) and compared to 20 resting-state datasets from standard, high-TR (1800 ms) EPI scans. The spatial distribution of fluctuations in various frequency ranges are analyzed along with the spectra of the time-series in voxels from different regions of interest. Functional connectivity specific to different frequency ranges (<0.1 Hz; 0.1–0.25 Hz; 0.25–0.75 Hz; 0.75–1.4 Hz) was computed for both the low-TR and (for the two lower-frequency ranges) the high-TR datasets using bandpass filters. In the low-TR data, cortical regions exhibited highest contribution of low-frequency fluctuations and the most marked low-frequency peak in the spectrum, while the time courses in subcortical grey matter regions as well as the insula were strongly contaminated by high-frequency signals. White matter and CSF regions had highest contribution of high-frequency fluctuations and a mostly flat power spectrum. In the high-TR data, the basic patterns of the low-TR data can be recognized, but the high-frequency proportions of the signal fluctuations are folded into the low frequency range, thus obfuscating the low-frequency dynamics. Regions with higher proportion of high-frequency oscillations in the low-TR data showed flatter power spectra in the high-TR data due to aliasing of the high-frequency signal components, leading to loss of specificity in the signal from these regions in high-TR data. Functional connectivity analyses showed that there are correlations between resting-state signal fluctuations of distant brain regions even at high frequencies, which can be measured using low-TR fMRI. On the other hand, in the high-TR data, loss of specificity of measured fluctuations leads to lower sensitivity in detecting functional connectivity. This underlines the advantages of low-TR EPI sequences for resting-state and potentially also task-related fMRI experiments.

fMRI measurements of amygdala activation are confounded by stimulus correlated signal fluctuation in nearby veins draining distant brain regions

Imaging the amygdala with functional MRI is confounded by multiple averse factors, notably signal dropouts due to magnetic inhomogeneity and low signal-to-noise ratio, making it difficult to obtain consistent activation patterns in this region. However, even when consistent signal changes are identified, they are likely to be due to nearby vessels, most notably the basal vein of rosenthal (BVR). Using an accelerated fMRI sequence with a high temporal resolution (TR = 333 ms) combined with susceptibility-weighted imaging, we show how signal changes in the amygdala region can be related to a venous origin. This finding is confirmed here in both a conventional fMRI dataset (TR = 2000 ms) as well as in information of meta-analyses, implying that “amygdala activations” reported in typical fMRI studies are likely confounded by signals originating in the BVR rather than in the amygdala itself, thus raising concerns about many conclusions on the functioning of the amygdala that rely on fMRI evidence alone.

Fully exploratory network independent component analysis of the 1000 functional connectomes database

The 1000 Functional Connectomes Project is a collection of resting-state fMRI datasets from more than 1000 subjects acquired in more than 30 independent studies from around the globe. This large, heterogeneous sample of resting-state data offers the unique opportunity to study the consistencies of resting-state networks at both subject and study level. In extension to the seminal paper by Biswal et al. (2010), where a repeated temporal concatenation group independent component analysis (ICA) approach on reduced subsets (using 20 as a pre-specified number of components) was used due to computational resource limitations, we herein apply Fully Exploratory Network ICA (FENICA) to 1000 single-subject independent component analyses. This, along with the possibility of using datasets of different lengths without truncation, enabled us to benefit from the full dataset available, thereby obtaining 16 networks consistent over the whole group of 1000 subjects. Furthermore, we demonstrated that the most consistent among these networks at both subject and study level matched networks most often reported in the literature, and found additional components emerging in prefrontal and parietal areas. Finally, we identified the influence of scan duration on the number of components as a source of heterogeneity between studies.

Reduced default mode network suppression during a working memory task in remitted major depression

Insufficient default mode network (DMN) suppression was linked to increased rumination in symptomatic Major Depressive Disorder (MDD). Since rumination is known to predict relapse and a more severe course of MDD, we hypothesized that similar DMN alterations might also exist during full remission of MDD (rMDD), a condition known to be associated with increased relapse rates specifically in patients with adolescent onset. Within a cross-sectional functional magnetic resonance imaging study activation and functional connectivity (FC) were investigated in 120 adults comprising 78 drug-free rMDD patients with adolescent- (n = 42) and adult-onset (n = 36) as well as 42 healthy controls (HC), while performing the n-back task. Compared to HC, rMDD patients showed diminished DMN deactivation with strongest differences in the anterior-medial prefrontal cortex (amPFC), which was further linked to increased rumination response style. On a brain systems level, rMDD patients showed an increased FC between the amPFC and the dorsolateral prefrontal cortex, which constitutes a key region of the antagonistic working-memory network. Both whole-brain analyses revealed significant differences between adolescent-onset rMDD patients and HC, while adult-onset rMDD patients showed no significant effects. Results of this study demonstrate that reduced DMN suppression exists even after full recovery of depressive symptoms, which appears to be specifically pronounced in adolescent-onset MDD patients. Our results encourage the investigation of DMN suppression as a putative predictor of relapse in clinical trials, which might eventually lead to important implications for antidepressant maintenance treatment.

Dynamic contrast-enhanced magnetic resonance imaging of breast tumors at 3 and 7 T: a comparison.

ObjectivesThe objective of this study was to compare the image quality, contrast enhancement behavior, and diagnostic value of bilateral 3-dimensional dynamic contrast-enhanced breast magnetic resonance imaging (MRI), with high spatial and temporal resolution, at 3 and 7 T, in the same patient group. Materials and MethodsTwenty-four consecutive patients (mean [SD] age, 57 [17] years) were included in this prospective institutional review board–approved study. Written informed consent was obtained from all patients. T1-weighted 3-dimensional sequences (time-resolved angiography with stochastic trajectories) were optimized at 3 and 7 T, with high temporal (both 14 seconds) and spatial resolution (1.1 × 1.1 × 1.1 mm3 [3 T], 0.7 × 0.7 × 0.7 mm3 [7 T]): echo time/repetition time, 2.84/6.01 milliseconds (3 T) and 2.5/4.75 milliseconds (7 T); acquisition time, 9 minutes (3 T/7 T). Dotarem® (gadoterate meglumine, Guerbet, Roissy CdG, France) contrast agent was injected intravenously as a bolus (0.2 mL/kg of body weight) after 3 baseline images. The images were rated according to breast imaging-reporting and data system by 2 radiologists in consensus. Signal-to-noise ratio and average enhancement ratios were measured quantitatively by means of region of interest analysis. In addition, B1 mapping was done in the same 5 healthy subjects at both field strengths. ResultsTwenty-eight enhancing lesions were detected in the 24 patients at both field strengths (16 malignant, 12 benign). At 7 T, higher contrast than that at 3 T and good image quality were achieved. With the high spatial isotropic resolution of 0.7 mm at 7 T, images with more detailed information could be acquired when compared with those acquired at 3 T. Sensitivity was 93.75% and 100%, at 3 and 7 T, respectively. Specificity was 91.67% at both field strengths. The signal-to-noise ratio at both field strengths was comparable, but at 7 T, the spatial resolution was 3.2-times higher than that at 3 T. A signal-to-noise ratio decrease toward prepectoral breast regions due to B1 inhomogeneities was observed at both field strengths but was stronger at 7 T (51%) than at 3 T (19%)(P = 0.0002). At 7 T, B1+ dropped by 20.7% and 32.8% in the prepectoral and lateral region of the breast in healthy subjects. ConclusionsOur comparison study shows that 7-T DCE-MRI provides simultaneous high temporal and spatial resolution that is significantly improved compared with lower field strengths, but further technical improvements are necessary to overcome B1 inhomogeneity problems at 7 T to fully unfold the potential of breast MRI at 7 T.

Sex differences in the functional connectivity of the amygdalae in association with cortisol

Human amygdalae are involved in various behavioral functions such as affective and stress processing. For these behavioral functions, as well as for psychophysiological arousal including cortisol release, sex differences are reported. Here, we assessed cortisol levels and resting-state functional connectivity (rsFC) of left and right amygdalae in 81 healthy participants (42 women) to investigate potential modulation of amygdala rsFC by sex and cortisol concentration. Our analyses revealed that rsFC of the left amygdala significantly differed between women and men: Women showed stronger rsFC than men between the left amygdala and left middle temporal gyrus, inferior frontal gyrus, postcentral gyrus and hippocampus, regions involved in face processing, inner-speech, fear and pain processing. No stronger connections were detected for men and no sex difference emerged for right amygdala rsFC. Also, an interaction of sex and cortisol appeared: In women, cortisol was negatively associated with rsFC of the amygdalae with striatal regions, mid-orbital frontal gyrus, anterior cingulate gyrus, middle and superior frontal gyri, supplementary motor area and the parietal-occipital sulcus. Contrarily in men, positive associations of cortisol with rsFC of the left amygdala and these structures were observed. Functional decoding analyses revealed an association of the amygdalae and these regions with emotion, reward and memory processing, as well as action execution. Our results suggest that functional connectivity of the amygdalae as well as the regulatory effect of cortisol on brain networks differs between women and men. These sex-differences and the mediating and sex-dependent effect of cortisol on brain communication systems should be taken into account in affective and stress-related neuroimaging research. Thus, more studies including both sexes are required.

Ultra-high resolution brain metabolite mapping at 7 T by short-TR Hadamard-encoded FID-MRSI ☆

ABSTRACT MRSI in the brain at ≥7 T is a technique of great promise, but has been limited mainly by low B0/B1+‐homogeneity, specific absorption rate restrictions, long measurement times, and low spatial resolution. To overcome these limitations, we propose an ultra‐high resolution (UHR) MRSI sequence that provides a 128×128 matrix with a nominal voxel volume of 1.7×1.7×8 mm3 in a comparatively short measurement time. A clinically feasible scan time of 10–20 min is reached via a short TR of 200 ms due to an optimised free induction decay‐based acquisition with shortened water suppression as well as parallel imaging (PI) using Controlled Aliasing In Parallel Imaging Results IN Higher Acceleration (CAIPIRINHA). This approach is not limited to a rectangular region of interest in the centre of the brain, but also covers cortical brain regions. Transversal pulse‐cascaded Hadamard encoding was able to further extend the coverage to 3D‐UHR‐MRSI of four slices (100×100×4 matrix size), with a measurement time of 17 min. Lipid contamination was removed during post‐processing using L2‐regularisation. Simulations, phantom and volunteer measurements were performed. The obtained single‐slice and 3D‐metabolite maps show the brain in unprecedented detail (e.g., hemispheres, ventricles, gyri, and the contrast between grey and white matter). This facilitates the use of UHR‐MRSI for clinical applications, such as measurements of the small structures and metabolic pathologic deviations found in small Multiple Sclerosis lesions. HIGHLIGHTSUltra‐high resolution MRSI (128×128 in‐plane matrix) at 7 T.Parallel imaging and short TR of 200 ms make UHR‐MRSI clinically feasible (10–20 min).Pulse‐cascaded Hadamard encoding provides 3D‐MRSI coverage.

Big Data Approaches for the Analysis of Large-Scale fMRI Data Using Apache Spark and GPU Processing: A Demonstration on Resting-State fMRI Data from the Human Connectome Project

Technologies for scalable analysis of very large datasets have emerged in the domain of internet computing, but are still rarely used in neuroimaging despite the existence of data and research questions in need of efficient computation tools especially in fMRI. In this work, we present software tools for the application of Apache Spark and Graphics Processing Units (GPUs) to neuroimaging datasets, in particular providing distributed file input for 4D NIfTI fMRI datasets in Scala for use in an Apache Spark environment. Examples for using this Big Data platform in graph analysis of fMRI datasets are shown to illustrate how processing pipelines employing it can be developed. With more tools for the convenient integration of neuroimaging file formats and typical processing steps, big data technologies could find wider endorsement in the community, leading to a range of potentially useful applications especially in view of the current collaborative creation of a wealth of large data repositories including thousands of individual fMRI datasets.

A highly parallelized framework for computationally intensive MR data analysis.

ObjectThe goal of this study was to develop a comprehensive magnetic resonance (MR) data analysis framework for handling very large datasets with user-friendly tools for parallelization and to provide an example implementation.Materials and methodsCommonly used software packages (AFNI, FSL, SPM) were connected via a framework based on the free software environment R, with the possibility of using Nvidia CUDA GPU processing integrated for high-speed linear algebra operations in R. Three hundred single-subject datasets from the 1,000 Functional Connectomes project were used to demonstrate the capabilities of the framework.ResultsA framework for easy implementation of processing pipelines was developed and an R package for the example implementation of Fully Exploratory Network ICA was compiled. Test runs on data from 300 subjects demonstrated the computational advantages of a processing pipeline developed using the framework compared to non-parallelized processing, reducing computation time by a factor of 15.ConclusionThe feasibility of computationally intensive exploratory analyses allows broader access to the tools for discovery science.