We propose Granger causality mapping (GCM) as an approach to explore directed influences between neuronal populations (effective connectivity) in fMRI data. The method does not rely on a priori specification of a model that contains pre-selected regions and connections between them. This distinguishes it from other fMRI effective connectivity approaches that aim at testing or contrasting specific hypotheses about neuronal interactions. Instead, GCM relies on the concept of Granger causality to define the existence and direction of influence from information in the data. Temporal precedence information is exploited to compute Granger causality maps that identify voxels that are sources or targets of directed influence for any selected region-of-interest. We investigated the method by simulations and by application to fMRI data of a complex visuomotor task. The presented exploratory approach of mapping influences between a region of interest and the rest of the brain can form a useful complement to existing models of effective connectivity. D 2004 Elsevier Inc. All rights reserved.

No abstract

This is the final paper in a Comments and Controversies series dedicated to “The identification of interacting networks in the brain using fMRI: Model selection, causality and deconvolution”. We argue that discovering effective connectivity depends critically on state-space models with biophysically informed observation and state equations. These models have to be endowed with priors on unknown parameters and afford checks for model Identifiability. We consider the similarities and differences among Dynamic Causal Modeling, Granger Causal Modeling and other approaches. We establish links between past and current statistical causal modeling, in terms of Bayesian dependency graphs and Wiener–Akaike–Granger–Schweder influence measures. We show that some of the challenges faced in this field have promising solutions and speculate on future developments.

Research indicates that dysfunctional food reward processing may contribute to pathological eating behaviour. It is widely recognized that both the amygdala and the orbitofrontal cortex (OFC) are essential parts of the brain's reward circuitry. The aims of this fMRI study were (1) to examine the effects of food deprivation and calorie content on reward processing in the amygdala and the OFC, and (2) to examine whether an explicit evaluation of foods is necessary for OFC, but not amygdalar activity. Addressing the first aim, healthy females were presented with high and low calorie food pictures while being either hungry or satiated. For the second aim, attention focus was manipulated by directing participants' attention either to the food or to a neutral aspect. This study shows that hunger interacts with the energy content of foods, modulating activity in the posterior cingulate cortex, medial OFC, insula, caudate putamen and fusiform gyrus. Results show that satiated healthy females show an increased reward processing in response to low calorie foods. Confirming our hypothesis, food deprivation increased activity following the presentation of high calorie foods, which may explain why treatments of obesity energy restricting diets often are unsuccessful. Interestingly, activity in both the amygdala and mOFC was only evident when participants explicitly evaluated foods. However, attention independent activity was found in the mPFC following the high calorie foods cues when participants where hungry. Current findings indicate that research on how attention modulates food reward processing might prove especially insightful in the study of the neural substrates of healthy and pathological eating behaviour.

Both the putative mirror neuron system (pMNS) and the ventral medial prefrontal cortex (vmPFC) are deemed important for social interaction: the pMNS because it supposedly "resonates" with the actions of others, the vmPFC because it is involved in mentalizing. Strictly speaking, the resonance property of the pMNS has never been investigated. Classical functional MRI experiments have only investigated whether pMNS regions augment their activity when an action is seen or executed. Resonance, however, entails more than only "going on and off together". Activity in the pMNS of an observer should continuously follow the more subtle changes over time in activity of the pMNS of the actor. Here we directly explore whether such resonance indeed occurs during continuous streams of actions. We let participants play the game of charades while we measured brain activity of both gesturer and guesser. We then applied a method to localize directed influences between the brains of the participants: between-brain Granger-causality mapping. Results show that a guesser's brain activity in regions involved in mentalizing and mirroring echoes the temporal structure of a gesturer's brain activity. This provides evidence for resonance theories and indicates a fine-grained temporal interplay between regions involved in motor planning and regions involved in thinking about the mental states of others. Furthermore, this method enables experiments to be more ecologically valid by providing the opportunity to leave social interaction unconstrained. This, in turn, would allow us to tap into the neural substrates of social deficits such as autism spectrum disorder.Granger causality | mirror system | social interaction | theory of mind | functional MRI

The functional MRI (fMRI) signal is an indirect measure of neuronal activity. In order to deconvolve the neuronal activity from the experimental fMRI data, biophysical generative models have been proposed describing the link between neuronal activity and the cerebral blood flow (the neurovascular coupling), and further the hemodynamic response and the BOLD signal equation. These generative models have been employed both for single brain area deconvolution and to infer effective connectivity in networks of multiple brain areas. In the current paper, we introduce a new fMRI model inspired by experimental observations about the physiological underpinnings of the BOLD signal and compare it with the generative models currently used in dynamic causal modeling (DCM), a widely used framework to study effective connectivity in the brain. We consider three fundamental aspects of such generative models for fMRI: (i) an adaptive two-state neuronal model that accounts for a wide repertoire of neuronal responses during and after stimulation; (ii) feedforward neurovascular coupling that links neuronal activity to blood flow; and (iii) a balloon model that can account for vascular uncoupling between the blood flow and the blood volume. Finally, we adjust the parameterization of the BOLD signal equation for different magnetic field strengths. This paper focuses on the form, motivation and phenomenology of DCMs for fMRI and the characteristics of the various models are demonstrated using simulations. These simulations emphasize a more accurate modeling of the transient BOLD responses - such as adaptive decreases to sustained inputs during stimulation and the post-stimulus undershoot. In addition, we demonstrate using experimental data that it is necessary to take into account both neuronal and vascular transients to accurately model the signal dynamics of fMRI data. By refining the models of the transient responses, we provide a more informed perspective on the underlying neuronal process and offer new ways of inferring changes in local neuronal activity and effective connectivity from fMRI.

The idea of an organized mode of brain function that is present as default state and suspended during goal-directed behaviors has recently gained much interest in the study of human brain function. The default mode hypothesis is based on the repeated observation that certain brain areas show task-induced deactivations across a wide range of cognitive tasks. In this event-related functional resonance imaging study we tested the default mode hypothesis by comparing common and selective patterns of BOLD deactivation in response to the demands on visual attention and working memory (WM) that were independently modulated within one task. The results revealed task-induced deactivations within regions of the default mode network (DMN) with a segregation of areas that were additively deactivated by an increase in the demands on both attention and WM, and areas that were selectively deactivated by either high attentional demand or WM load. Attention-selective deactivations appeared in the left ventrolateral and medial prefrontal cortex and the left lateral temporal cortex. Conversely, WM-selective deactivations were found predominantly in the right hemisphere including the medial-parietal, the lateral temporo-parietal, and the medial prefrontal cortex. Moreover, during WM encoding deactivated regions showed task-specific functional connectivity. These findings demonstrate that task-induced deactivations within parts of the DMN depend on the specific characteristics of the attention and WM components of the task. The DMN can thus be subdivided into a set of brain regions that deactivate indiscriminately in response to cognitive demand ("the core DMN") and a part whose deactivation depends on the specific task.

Deep brain stimulation of the subthalamic nucleus (STN) is a widely performed surgical treatment for patients with Parkinson’s disease. The goal of the surgery is to place an electrode centered in the motor region of the STN while lowering the effects of electrical stimulation on the non-motor regions. However, distinguishing the motor region from the neighboring associative and limbic areas in individual patients using imaging modalities was until recently difficult to obtain in vivo. Here, using ultra-high field MR imaging, we have performed a dissection of the subdivisions of the STN of individual Parkinson’s disease patients. We have acquired 7 T diffusion-weighted images of seventeen patients with Parkinson’s disease scheduled for deep brain stimulation surgery. Using a structural connectivity-based parcellation protocol, the STN’s connections to the motor, limbic, and associative cortical areas were used to map the individual subdivisions of the nucleus. A reproducible patient-specific parcellation of the STN into a posterolateral motor and gradually overlapping central associative area was found in all STNs, taking up on average 55.3% and 55.6% of the total nucleus volume. The limbic area was found in the anteromedial part of the nucleus. Our results suggest that 7T MR imaging may facilitate individualized and highly specific planning of deep brain stimulation surgery of the STN.