Depression recognition using resting-state and event-related fMRI signals

Purpose

This paper aimed to develop a method for depression detection using blood-oxygen-level-dependent (BOLD) response estimated from event-related signals and resting-state functional magnetic resonance imaging (fMRI) signals together.

Materials and Methods

Thirteen patients with unipolar depression and matched healthy subjects were recruited. Resting state data of each subject were collected. Thereafter, event-related paradigm was undertaken using sad facial stimuli. The resting-state fMRI signal was deemed as the baseline of each subject's activity. Coefficient marks were designed to sort and select temporal independent components of event-related signals. Thereafter, stimulus-evoked BOLD response components inside event-related signal were extracted and taken as features to discriminate depressive patients from healthy controls.

Results

Accuracy rate for depression recognition was 77.27% with P value of .017 for whole-brain analysis and 81.82% with P value of .009 for region-of-interest analysis. The effectiveness and the superiority of the proposed method for disease recognition were demonstrated via the performance comparison with three other typical methods.

Conclusions

The proposed model was effective in depression recognition.     

Origins of intersubject variability of blood oxygenation level dependent and arterial spin labeling fMRI: implications for quantification of brain activity

Accurate localization of brain activity using blood oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fMRI) has been challenged because of the large BOLD signal within distal veins. Arterial spin labeling (ASL) techniques offer greater sensitivity to the microvasculature but possess low temporal resolution and limited brain coverage. In this study, we show that the physiological origins of BOLD and ASL depend on whether percent change or statistical significance is being considered. For BOLD and ASL fMRI data collected during a simple unilateral hand movement task, we found that in the area of the contralateral motor cortex the centre of gravity (CoG) of the intersubject coefficient of variation (CV) of BOLD fMRI was near the brain surface for percent change in signal, whereas the CoG of the intersubject CV for Z-score was in close proximity of sites of brain activity for both BOLD and ASL. These findings suggest that intersubject variability of BOLD percent change is vascular in origin, whereas the origin of inter-subject variability of Z-score is neuronal for both BOLD and ASL. For longer duration tasks (12 s or greater), however, there was a significant correlation between BOLD and ASL percent change, which was not evident for short duration tasks (6 s). These findings suggest that analyses directly comparing percent change in BOLD signal between pre-defined regions of interest using short duration stimuli, as for example in event-related designs, may be heavily weighted by large-vessel responses rather than neuronal responses.     

Applying functional MRI to the spinal cord and brainstem

Functional magnetic resonance imaging of the spinal cord (spinal fMRI) has facilitated the noninvasive visualization of neural activity in the spinal cord (SC) and brainstem of both animals and humans. This technique has yet to gain the widespread usage of brain fMRI, due in part to the intrinsic technical challenges spinal fMRI presents and to the narrower scope of applications it fulfills. Nonetheless, methodological progress has been considerable and rapid. To date, spinal fMRI studies have investigated SC function during sensory or motor task paradigms in spinal cord injury (SCI), multiple sclerosis (MS) and neuropathic pain (NP) patient populations, all of which have yielded consistent and sensitive results. The most recent study in our laboratory has successfully used spinal fMRI to examine cervical SC activity in a SCI patient with a metallic fixation device spanning the C₄ to C₆ vertebrae, a critical step in realizing the clinical utility of the technique. The literature reviewed in this article suggests that spinal fMRI is poised for usage in a wide range of patient populations, as multiple groups have observed intriguing, yet consistent, results using standard, readily available MR systems and hardware. The next step is the implementation of this technique in the clinic to supplement standard qualitative behavioral assessments of SCI. Spinal fMRI may offer insight into the subtleties of function in the injured and diseased SC, and support the development of new methods for treatment and monitoring.     

Classifier ensembles for fMRI data analysis: an experiment

Functional magnetic resonance imaging (fMRI) is becoming a forefront brain–computer interface tool. To decipher brain patterns, fast, accurate and reliable classifier methods are needed. The support vector machine (SVM) classifier has been traditionally used. Here we argue that state-of-the-art methods from pattern recognition and machine learning, such as classifier ensembles, offer more accurate classification. This study compares 18 classification methods on a publicly available real data set due to Haxby et al. [Science 293 (2001) 2425–2430]. The data comes from a single-subject experiment, organized in 10 runs where eight classes of stimuli were presented in each run. The comparisons were carried out on voxel subsets of different sizes, selected through seven popular voxel selection methods. We found that, while SVM was robust, accurate and scalable, some classifier ensemble methods demonstrated significantly better performance. The best classifiers were found to be the random subspace ensemble of SVM classifiers, rotation forest and ensembles with random linear and random spherical oracle.     

Testing methodologies for the nonlinear analysis of causal relationships in neurovascular coupling

We investigated the use and implementation of a nonlinear methodology for establishing which changes in neurophysiological signals cause changes in the blood oxygenation level-dependent (BOLD) contrast measured in functional magnetic resonance imaging. Unlike previous analytical approaches, which used linear correlation to establish covariations between neural activity and BOLD, we propose a directed information-theoretic measure, the transfer entropy, which can elucidate even highly nonlinear causal relationships between neural activity and BOLD signal. In this study we investigated the practicality of such an analysis given the limited data samples that can be collected experimentally due to the low temporal resolution of BOLD signals. We implemented several algorithms for the estimation of transfer entropy and we tested their effectiveness using simulated local field potentials (LFPs) and BOLD data constructed to match the main statistical properties of real LFP and BOLD signals measured simultaneously in monkey primary visual cortex. We found that using the advanced methods of entropy estimation implemented and described here, a transfer entropy analysis of neurovascular coupling based on experimentally attainable data sets is feasible.     

Assessment of spatial BOLD sensitivity variations in fMRI using gradient-echo field maps

Clinical blood oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fMRI) is becoming increasingly valuable in, e.g., presurgical planning, but the commonly used gradient-echo echo-planar imaging (GE-EPI) technique is sometimes hampered by macroscopic field inhomogeneities. This can affect the degree of signal change that will occur in the GE-EPI images as a response to neural activation and the subsequent blood oxygenation changes, i.e., the BOLD sensitivity (BS). In this study, quantitative BS maps were calculated directly from gradient-echo field maps obtainable on most clinical scanners. In order to validate the accuracy of the calculated BS-maps, known shim gradients were applied and field maps and GE-EPI images of a phantom were acquired. Measured GE-EPI image intensity was then compared with the calculated (predicted) image intensity (pII) which was obtained from the field maps using theoretical expressions for image-intensity loss. The validated expressions for pII were used to calculate the corresponding predicted BOLD sensitivity (pBS) maps in healthy volunteers. Since the field map is assumed to be valid throughout an entire fMRI experiment, the influence of subject motion on the pBS maps was also assessed. To demonstrate the usefulness of such maps, pBS was investigated for clinically important functional areas including hippocampus, Broca's area and primary motor cortex. A systematic left/right pBS difference was observed in Broca's area and in the hippocampus, most likely due to magnetic field inhomogeneity of the particular MRI-system used in this study. For all subjects, the hippocampus showed pBS values above unity with a clear anterior–posterior gradient and with an abrupt drop to zero pBS in the anterior parts of hippocampus. It is concluded that GE field maps can be used to accurately predict BOLD sensitivity and that this parameter is useful to assess spatial variations which will influence fMRI experiments.     

Neural specificity of acupuncture stimulation from support vector machine classification analysis

Acupoint specificity, as a crucial issue in acupuncture neuroimaging studies, is still a controversial topic. Previous studies have generally adopted a block-based general linear model (GLM) approach, which predicts the temporal changes in the blood oxygenation level-dependent signal conforming to the “on-off” specifications. However, this method might become impractical since the precise timing and duration of acupuncture actions cannot be specified a priori. In the current study, we applied a data-driven multivariate classification approach, namely, support vector machine (SVM), to explore the neural specificity of acupuncture at gall bladder 40 (GB40) using kidney 3 (KI3) as a control condition (belonging to different meridians but the same nerve segment). In addition, to verify whether the typical GLM approach is sensitive enough in exploring the neural response patterns evoked by acupuncture, we also employed the GLM method to the same data sets. The SVM analysis detected distinct neural response patterns between GB40 and KI3 — positive predominantly for the GB40, while negative following the KI3. By contrast, group analysis from the GLM showed that acupuncture at these different acupoints can both evoke similar widespread signal decreases in multiple brain regions, and most of these regions were spatially overlapped, mainly distributing in the limbic and subcortical structures. Our findings may provide additional evidence to support the specificity of acupuncture, relevant to its clinical efficacy. Moreover, we also proved that GLM analysis is prone to be susceptible to errors and is not appropriate for detecting neural response patterns evoked by acupuncture stimulation.     

A conditional Granger causality model approach for group analysis in functional magnetic resonance imaging

Granger causality model (GCM) derived from multivariate vector autoregressive models of data has been employed to identify effective connectivity in the human brain with functional magnetic resonance imaging (fMRI) and to reveal complex temporal and spatial dynamics underlying a variety of cognitive processes. In the most recent fMRI effective connectivity measures, pair-wise GCM has commonly been applied based on single-voxel values or average values from special brain areas at the group level. Although a few novel conditional GCM methods have been proposed to quantify the connections between brain areas, our study is the first to propose a viable standardized approach for group analysis of fMRI data with GCM. To compare the effectiveness of our approach with traditional pair-wise GCM models, we applied a well-established conditional GCM to preselected time series of brain regions resulting from general linear model (GLM) and group spatial kernel independent component analysis of an fMRI data set in the temporal domain. Data sets consisting of one task-related and one resting-state fMRI were used to investigate connections among brain areas with the conditional GCM method. With the GLM-detected brain activation regions in the emotion-related cortex during the block design paradigm, the conditional GCM method was proposed to study the causality of the habituation between the left amygdala and pregenual cingulate cortex during emotion processing. For the resting-state data set, it is possible to calculate not only the effective connectivity between networks but also the heterogeneity within a single network. Our results have further shown a particular interacting pattern of default mode network that can be characterized as both afferent and efferent influences on the medial prefrontal cortex and posterior cingulate cortex. These results suggest that the conditional GCM approach based on a linear multivariate vector autoregressive model can achieve greater accuracy in detecting network connectivity than the widely used pair-wise GCM, and this group analysis methodology can be quite useful to extend the information obtainable in fMRI.     

Balanced steady-state free precession with parallel imaging gives distortion-free fMRI with high temporal resolution

Research on the functions of the human brain requires that functional magnetic resonance imaging (MRI) moves towards producing images with less distortion and higher temporal and spatial resolution. This study compares passband balanced steady-state free precession (bSSFP) acquisitions with and without parallel imaging (PI) to investigate whether combining PI with this pulse sequence is a viable option for functional MRI. Such a novel combination has the potential to offer the distortion-free advantages of bSSFP with the reduced acquisition time of PI. Scans were done on a Philips 3T Intera, using the installed bSSFP pulse sequence, both with and without the sensitivity encoding (SENSE) PI option. The task was a visual flashing checkerboard, and the viewing window covered the visual cortex. Sensitivity comparisons with and without PI were done using the same manually drawn region of interest for each time course of the subject, and comparing the z-score summary statistics: number of voxels with z>2.3, the mean of those voxels, their 90th percentile and their maximum value. We show that PI greatly improves the temporal resolution in bSSFP, reducing the volume acquisition time by more than half in this study to 0.67 s with 3-mm isotropic voxels. At the same time, a statistically significant increase was found for the maximum z-score using bSSFP with PI as compared to without it (P=.02). This improvement can be understood in terms of physiological noise, as demonstrated by noise measurements. This produces observed increases in the overall temporal signal to noise of the functional time series, giving greater sensitivity to functional activations with PI. This study demonstrates for the first time the possibility of combining PI with bSSFP to achieve distortion-free functional images without loss of sensitivity and with high temporal resolution.