Maria Joao Rosa

Comparing families of dynamic causal models

Mathematical models of scientific data can be formally compared using Bayesian model evidence. Previous applications in the biological sciences have mainly focussed on model selection in which one first selects the model with the highest evidence and then makes inferences based on the parameters of that model. This “best model” approach is very useful but can become brittle if there are a large number of models to compare, and if different subjects use different models. To overcome this shortcoming we propose the combination of two further approaches: (i) family level inference and (ii) Bayesian model averaging within families. Family level inference removes uncertainty about aspects of model structure other than the characteristic of interest. For example: What are the inputs to the system? Is processing serial or parallel? Is it linear or nonlinear? Is it mediated by a single, crucial connection? We apply Bayesian model averaging within families to provide inferences about parameters that are independent of further assumptions about model structure. We illustrate the methods using Dynamic Causal Models of brain imaging data.

PRoNTo: Pattern Recognition for Neuroimaging Toolbox.

In the past years, mass univariate statistical analyses of neuroimaging data have been complemented by the use of multivariate pattern analyses, especially based on machine learning models. While these allow an increased sensitivity for the detection of spatially distributed effects compared to univariate techniques, they lack an established and accessible software framework. The goal of this work was to build a toolbox comprising all the necessary functionalities for multivariate analyses of neuroimaging data, based on machine learning models. The “Pattern Recognition for Neuroimaging Toolbox” (PRoNTo) is open-source, cross-platform, MATLAB-based and SPM compatible, therefore being suitable for both cognitive and clinical neuroscience research. In addition, it is designed to facilitate novel contributions from developers, aiming to improve the interaction between the neuroimaging and machine learning communities. Here, we introduce PRoNTo by presenting examples of possible research questions that can be addressed with the machine learning framework implemented in PRoNTo, and cannot be easily investigated with mass univariate statistical analysis.

Estimating the transfer function from neuronal activity to BOLD using simultaneous EEG-fMRI

Previous studies using combined electrical and hemodynamic measurements of brain activity, such as EEG and (BOLD) fMRI, have yielded discrepant results regarding the relationship between neuronal activity and the associated BOLD response. In particular, some studies suggest that this link, or transfer function, depends on the frequency content of neuronal activity, while others suggest that total neuronal power accounts for the changes in BOLD. Here we explored this dependency by comparing different frequency-dependent and -independent transfer functions, using simultaneous EEG-fMRI. Our results suggest that changes in BOLD are indeed associated with changes in the spectral profile of neuronal activity and that these changes do not arise from one specific spectral band. Instead they result from the dynamics of the various frequency components together, in particular, from the relative power between high and low frequencies. Understanding the nature of the link between neuronal activity and BOLD plays a crucial role in improving the interpretability of BOLD images as well as on the design of more robust and realistic models for the integration of EEG and fMRI.

Connectivity-based neurofeedback: Dynamic causal modeling for real-time fMRI

Neurofeedback based on real-time fMRI is an emerging technique that can be used to train voluntary control of brain activity. Such brain training has been shown to lead to behavioral effects that are specific to the functional role of the targeted brain area. However, real-time fMRI-based neurofeedback so far was limited to mainly training localized brain activity within a region of interest. Here, we overcome this limitation by presenting near real-time dynamic causal modeling in order to provide feedback information based on connectivity between brain areas rather than activity within a single brain area. Using a visual–spatial attention paradigm, we show that participants can voluntarily control a feedback signal that is based on the Bayesian model comparison between two predefined model alternatives, i.e. the connectivity between left visual cortex and left parietal cortex vs. the connectivity between right visual cortex and right parietal cortex. Our new approach thus allows for training voluntary control over specific functional brain networks. Because most mental functions and most neurological disorders are associated with network activity rather than with activity in a single brain region, this novel approach is an important methodological innovation in order to more directly target functionally relevant brain networks.

Sparse network-based models for patient classification using fMRI

Pattern recognition applied to whole-brain neuroimaging data, such as functional Magnetic Resonance Imaging (fMRI), has been successful at discriminating psychiatric patients from healthy subjects. However, predictive patterns obtained from whole-brain voxel-based features are difficult to interpret in terms of the underlying neurobiology. As is generally accepted, many psychiatric disorders, such as depression and schizophrenia, are brain connectivity disorders. Therefore, pattern recognition based on network models should provide more scientific insight and potentially more powerful predictions than voxel-based approaches. Here, we build a sparse network-based discriminative modelling framework, based on Gaussian graphical models and L1-norm regularised linear Support Vector Machines (SVM). The proposed framework provides easier pattern interpretation in terms of underlying network changes between groups, and we illustrate our technique by classifying patients with depression and controls, using fMRI data from a sad facial processing task.

Bayesian Model Selection Maps for Group Studies

This technical note describes the construction of posterior probability maps (PPMs) for Bayesian model selection (BMS) at the group level. This technique allows neuroimagers to make inferences about regionally specific effects using imaging data from a group of subjects. These effects are characterised using Bayesian model comparisons that are analogous to the F-tests used in statistical parametric mapping, with the advantage that the models to be compared do not need to be nested. Additionally, an arbitrary number of models can be compared together. This note describes the integration of the Bayesian mapping approach with a random effects analysis model for BMS using group data. We illustrate the method using fMRI data from a group of subjects performing a target detection task.

Connectivity changes underlying neurofeedback training of visual cortex activity

Neurofeedback based on real-time functional magnetic resonance imaging (fMRI) is a new approach that allows training of voluntary control over regionally specific brain activity. However, the neural basis of successful neurofeedback learning remains poorly understood. Here, we assessed changes in effective brain connectivity associated with neurofeedback training of visual cortex activity. Using dynamic causal modeling (DCM), we found that training participants to increase visual cortex activity was associated with increased effective connectivity between the visual cortex and the superior parietal lobe. Specifically, participants who learned to control activity in their visual cortex showed increased top-down control of the superior parietal lobe over the visual cortex, and at the same time reduced bottom-up processing. These results are consistent with efficient employment of top-down visual attention and imagery, which were the cognitive strategies used by participants to increase their visual cortex activity.

Can Emotional and Behavioral Dysregulation in Youth Be Decoded from Functional Neuroimaging

Introduction High comorbidity among pediatric disorders characterized by behavioral and emotional dysregulation poses problems for diagnosis and treatment, and suggests that these disorders may be better conceptualized as dimensions of abnormal behaviors. Furthermore, identifying neuroimaging biomarkers related to dimensional measures of behavior may provide targets to guide individualized treatment. We aimed to use functional neuroimaging and pattern regression techniques to determine whether patterns of brain activity could accurately decode individual-level severity on a dimensional scale measuring behavioural and emotional dysregulation at two different time points. Methods A sample of fifty-seven youth (mean age: 14.5 years; 32 males) was selected from a multi-site study of youth with parent-reported behavioral and emotional dysregulation. Participants performed a block-design reward paradigm during functional Magnetic Resonance Imaging (fMRI). Pattern regression analyses consisted of Relevance Vector Regression (RVR) and two cross-validation strategies implemented in the Pattern Recognition for Neuroimaging toolbox (PRoNTo). Medication was treated as a binary confounding variable. Decoded and actual clinical scores were compared using Pearson’s correlation coefficient (r) and mean squared error (MSE) to evaluate the models. Permutation test was applied to estimate significance levels. Results Relevance Vector Regression identified patterns of neural activity associated with symptoms of behavioral and emotional dysregulation at the initial study screen and close to the fMRI scanning session. The correlation and the mean squared error between actual and decoded symptoms were significant at the initial study screen and close to the fMRI scanning session. However, after controlling for potential medication effects, results remained significant only for decoding symptoms at the initial study screen. Neural regions with the highest contribution to the pattern regression model included cerebellum, sensory-motor and fronto-limbic areas. Conclusions The combination of pattern regression models and neuroimaging can help to determine the severity of behavioral and emotional dysregulation in youth at different time points.

Full Bayesian multi-task learning for multi-output brain decoding and accommodating missing data

Multi-task learning (MTL) has recently been demonstrated to be highly promising for decoding multiple target variables from neuroimaging data. Its primary advantage is that it makes more efficient use of the data than existing decoding models, leading to improved accuracy. In this work, we propose a novel Bayesian MTL approach, motivated by problems such as clinical applications where accurate quantification of uncertainty is crucial. We present a Markov chain Monte Carlo approach to perform inference in the model and demonstrate the approach using a publicly available neuroimaging dataset. We study the conditions where MTL is likely to improve performance: we first evaluate MTL as an approach for accommodating missing data, which is an important problem that has received little attention from the neuroimaging community. We then examine whether it is beneficial to include classification and regression tasks in the same model. We relate our conclusions to results from geostatistics, where MTL methods were pioneered, and make recommendations for neuroimaging practitioners using MTL.

Posterior Probability Maps

Posterior probability maps (PPMs) correspond to images of the probability or confidence that an activation at a particular voxel exceeds some specified threshold, given the data. PPMs are an alternative to voxel-wise classical inference, based on statistical parametric maps (SPMs), by allowing imaging neuroscientists to make Bayesian inferences about regionally specific effects in the brain. This article describes the construction of PPMs, including an empirical Bayes approach to estimate the maps. We also present PPMs and SPMs for the same effect in a single-subject functional magnetic resonance imaging dataset.