Advances in Brain Imaging: A Shift Towards Functional Connectivity

Numerous imaging techniques have been utilised to study the structure and function of the brain, including functional magnetic resonance imaging (fMRI), positron-emission tomography (PET) and X-ray computed tomography (CT). fMRI is a form of blood-oxygen-level dependent (BOLD) contrast imaging which measures the relative proportions of oxyhemoglobin and deoxyhemoglobin – based on the assumption that oxygenated blood flows to active neurons at a greater rate than inactive neurons (the hemodynamic response) – while PET detects gamma rays emitted by a positron-emitting radiotracer introduced into the body and CT combines many X-ray images to construct a 3D representation of the brain. fMRI has typically been used to map brain regions which become activated while the subject is engaged in a particular task, however more recently it has been realised that since brain regions often work in networks (e.g. the default mode network), understanding cognitive processes and behaviours at a neurobiological level requires analysis of the functional connectivity between brain regions constituting functional networks (Menon, 2011). The brain is a unique organ in that, despite its fixed anatomy, local interactions are dynamically modulated to allow a vast functional repertoire (Park & Friston, 2013). Thus, projects such as the Human Connectome Project have begun to map the functional connectivity of the brain using resting-state and task-based fMRI, while recent advances in neuroimaging technology have enabled mental processes and behaviours to be correlated with specific brain structures with greater accuracy and reliability. 

Tong & Pratte (2012) discuss the advances in brain imaging, presenting a wealth of information gathered on a range of mental processes using novel fMRI techniques such as multivoxel pattern analysis. The authors discuss how new techniques allow (to some extent) the ability to “mind read” which of two previously viewed images a subject was imagining, and go as far as to say: “As these methodologies continue to advance, it will become increasingly important to consider the ethical implications of this technology”. This is a bold statement emphasising the fast-evolving nature of brain imaging technology. The authors also review how numerous studies over the past decade have succeeded in decoding various top-down mental processes, such as feature-based attention (Kamitani & Tong, 2005), imagination (Reddy et al., 2010), episodic memory (Rissman & Wagner, 2012) and numerical processing (Knops et al., 2009). For example, Kamitano & Tong (2005) found, using statistical algorithms on fMRI data, that specific fMRI signals in the visual cortex (V1) could reliably predict which of eight stimulus orientations the subject was attending to, indicating that the visual cortex encodes detailed orientation information which can reliably predict subjective perception and suggesting direct orientation mapping at a neurobiological level. Later neuroimaging research by Tong et al. (2012) found that the correlation between such orientation-selective activity patterns and the quality of the sensory input could be directly predicted by the average BOLD amplitude in the brain region of interest using multivariate pattern analysis, offering a reliable model of fMRI pattern classification. This is a prime example of how both brain imaging technology and the way in which researchers use it has advanced over recent years, allowing unprecedented revelations in the neurobiology of specific brain functions.

Tong & Pratte (2012) also highlight methodological issues encountered in brain imaging research. For example, when analysing fMRI data collected while subjects watched humorous events in a video, it transpired that the ventricles were the most statistically informative brain region in predicting a subjects’ urge to laugh – even though it is highly unlikely that the ventricles themselves play a functional role in the cognitive processing of humorous events (Tong & Pratte, 2012). This error is dubbed the “fallacy of reverse inference” (Poldrack, 2006), and is eloquently demonstrated by neuroscientist James Fallon’s conclusion that he must be a psychopath based on PET images of his own brain (Fallon, 2013). Another problem faced is that brain regions are often associated with multiple mental processes, and regions implicated in particular functions tend to vary with experimental variables such as task demands and the specific baseline condition used to identify them (Rabinovich et al., 2012), and even between individual subjects (Kelly et al., 2012). For instance, the hippocampus is known to be involved in the recall of episodic memories, but has also been implicated in the imagining of future events; however, patients with hippocampal damage in some conditions retain unimpaired imagination, suggesting that the hippocampus may not be necessary for it (Martin et al., 2011). Furthermore, there is the issue of causation versus correlation. For example, a famous MRI study by Maguire et al. (2006) found that London taxi drivers had greater grey matter volume in the mid-posterior (but not anterior) hippocampus compared to London bus drivers, who are not required to learn the colossal amount of information which constitutes “the knowledge”. However, the results do not indicate whether these differences are a direct result of the learning of “the knowledge”, or whether subjects showing this particular neuroanatomy are predisposed to becoming London taxi drivers.

Bennett et al. (2009) highlighted some of the common issues faced when analysing fMRI data (e.g. the multiple comparisons problem) by providing evidence that fMRI scans of a dead salmon apparently showed statistically significant brain activity according to commonly-used statistical tests. Subsequent fMRI studies swiftly adopted the corrected comparison methods proposed (Bennett et al., 2009). Additionally, while identifying functional connectivity has recently been recognised as more important than identifying individual brain regions (Menon, 2011), there are limits to how much fMRI can reveal about connectivity. For example, fMRI studies demonstrate that activation of the prefrontal cortex during cognitive evaluation of threatening facial expressions is associated with an attenuated response of the amygdala, apparently indicating a functional network for emotional regulation (Hariri et al., 2003). However, fMRI data does not necessarily show that the amygdala was inhibited by the prefrontal cortex; this is merely inferred. In fact, such negative correlations are often incorrectly interpreted as “inhibitory interactions” (Kelly et al., 2012). Thus, novel techniques / computational models have been developed attempting to overcome these issues.

One such model is the state-space multivariate dynamical systems (MDS) model (Ryali et al., 2011), which takes into account inherent regional differences in the hemodynamic response and focuses on changes in latent signals rather than BOLD signals – which themselves do not necessarily measure the underlying neural activity. More recently optogenetic fMRI (ofMRI) – which combines optogenetic control of neural circuits with fMRI, enabling more direct investigation of connectivity in vivo (Lee, 2012) – when combined with the MDS model was found to be reliable method for identifying functional interactions between brain regions (Ryali et al., 2016).

Figure 1: ofMRI: optically-driven local excitation in defined rodent neocortical cells drives positive BOLD. a. Experimental schematic: transduced cells (triangles) and blue light delivery shown in M1 motor cortex. b. ChR2-EYFP expression in M1. c. ofMRI hemodynamic response during 6 consecutive epochs of optical stimulation. d. BOLD activation is observed at the site of optical stimulation. Adapted from Lee (2012).

Another advancement in statistical analysis being increasingly employed in fMRI studies is the shift from the identification of regions of interest (ROI) to the “parcellation” of whole brain resting-state fMRI data into spatially coherent regions of homogenous functional connectivity (e.g. by cluster analysis) (Craddock et al., 2012). This type of analysis offers several advantages, including providing a measure of the stability of resting-state functional networks between individual subject data and across grouped subjects data (Bellec et al., 2010). Variations between subjects in these intrinsic functional connectivity based “parcellations” does not appear to be correlated with structural variations; rather, they follow intrinsic variations in functional connectivity evoked by specific tasks (Mennes et al., 2010). Thus, studies are beginning to examine links between these task-evoked variations in resting-state functional connectivity and specific behaviours/mental processes (Kelly et al., 2012; Adelstein et al., 2011). Such novel methods of statistical analysis hold promise in elucidating stable functional networks across individuals, with the aim of explaining cognitive functions at a neurobiological level.

Furthermore, recent advances in MRI imaging techniques allow full brain scans to be completed up to 2-3 times faster. This is a significant advantage since it allows for greater statistical definition of neuronal networks (i.e clearer identification of functionally relevant networks), as well as enhanced visualisations of structural connections in the brain such as white matter tracts (Feinberg & Setsompop, 2013). Similarly, the development of high-field strength MRI scanners is expected to dramatically advance our understanding of the pathology of multiple sclerosis (MS) (Filippi, 2014). Advances in brain imaging techniques and how they are used is also benefitting the study of many other pathological behaviours such as those of autism or schizophrenia. Rather than focusing on pathology of individual brain regions, studies are increasingly focusing on aberrant interactions between specific distributed neural networks as well as more general disturbances in functional connectivity (Menon, 2011). For example, one study found that schizophrenic subjects show less integrated but more diverse functional connectivity during task-based behavioural measures compared to controls, even suggesting a possible advantage of the “schizophrenia connectome” (Lynall et al., 2010). Recent neuroimaging studies of autism spectrum disorder (ASD) have also shifted the focus to identifying abnormal connectivity (Vissers et al., 2012), with some studies reporting “underconnectivity” in frontal regions and others reporting “overconnectivity” – likely as a result of the specific methodological variables/analysis type in each study (Nair et al., 2014). Studies are beginning to examine the development of normal functional connectivity in the brain throughout adolescence and the effect of genes and the environment in the development of abnormal connectivity (Blakemore, 2012). However, while these advances are promising, there remain some limitations; for example, different mental diseases often affect the same resting-state networks, bringing into question the specificity of the findings of neuroimaging studies (Barkhof et al., 2014).

Figure 2: EEG waveform representations to spatial/map representations (and analyses).
From Michel & Murray (2012)

Recent advances in signal analysis have also allowed electroencephalography (EEG) to be used as an efficient brain imaging technique, particularly useful in the investigation of abnormal temporal dynamics in functional networks at the millisecond range. Michel & Murray (2012) argue that the full potential of EEG has been underestimated, proposing that proper analysis of the electric field potentials from each electrode can provide spatio-temporal information which may be useful as an adjunct to fMRI studies. Diffusion tensor imaging (DTI) – a form of MRI which tracks the diffusion of water molecules along white matter tracts – has also been used effectively in a recent connectome analysis study of major depressive disorder (MDD), finding reduced structural connectivity in regions constituting the default mode network as well as the frontal cortex, thalamus and caudate regions, thought to be involved in emotional and cognitive processing (Korgaonkar et al., 2014). However, structural connectivity does not necessarily predict functional connectivity.

With regard to connectome projects, Ohno et al. (2016) review recent advances in the acquisition and analysis of large “connectomic” data sets, while concluding that the combination of different imaging modalities and further advances in the (automated) analysis of data may revolutionise our understanding of the structural and functional connectome, elucidating neural mechanisms underlying not only basic sensory/motor functions and disease but also higher mental processes such as consciousness and intelligence.

Recent brain imaging studies have shown a marked paradigm shift from attempting to localise mental processes/behaviours to individual brain regions to identifying whole-brain structural and functional connectivity patterns which may become aberrant in various neurological/psychiatric disorders. As brain imaging technologies continue to advance, researchers continue to develop novel methods to exploit them to their full potential, allowing unprecedented advances in the neurobiology of specific cognitive functions and behaviours.

References: 

Adelstein, J. S., Shehzad, Z., Mennes, M., DeYoung, C. G., Zuo, X.-N., Kelly, C., Margulies, D. S., Bloomfield, A., Gray, J. R., Castellanos, F. X. and Milham, M. P. (2011) ‘Personality Is Reflected in the Brain’s Intrinsic Functional Architecture’, Valdes-Sosa, M. (ed.), PLoS ONE, 6(11), p. e27633.

Barkhof, F., Haller, S. and Rombouts, S. A. R. B. (2014) ‘Resting-State Functional MR Imaging: A New Window to the Brain’, Radiology, 272(1), pp. 29–49.

Bellec, P., Rosa-Neto, P., Lyttelton, O. C., Benali, H. and Evans, A. C. (2010) ‘Multi-level bootstrap analysis of stable clusters in resting-state fMRI’, NeuroImage, 51(3), pp. 1126–1139.

Bennett, C., Miller, M. and Wolford, G. (2009) ‘Neural correlates of interspecies perspective taking in the post-mortem Atlantic Salmon: an argument for multiple comparisons correction’, NeuroImage (Organization for Human Brain Mapping 2009 Annual Meeting), 47, Supplement 1, p. S125.

Blakemore, S.-J. (2012) ‘Imaging brain development: The adolescent brain’, NeuroImage, 61(2), pp. 397–406.

Craddock, R. C., James, G. A., Holtzheimer, P. E., Hu, X. P. and Mayberg, H. S. (2012) ‘A whole brain fMRI atlas generated via spatially constrained spectral clustering’, Human brain mapping, 33(8), p. 10.1002/hbm.21333.

Fallon, J. (2013) The Psychopath Inside: A Neuroscientist’s Personal Journey into the Dark Side of the Brain, New York, Penguin Publishing Group.

Feinberg, D. A. and Setsompop, K. (2013) ‘Ultra-fast MRI of the human brain with simultaneous multi-slice imaging’, Frontiers of In Vivo and Materials MRI Research, 229, pp. 90–100.

Filippi, M. (2014) ‘Recent advances in MS neuroimaging’, Multiple Sclerosis and Related Disorders, 3(6), p. 767.

Hariri, A. R., Mattay, V. S., Tessitore, A., Fera, F. and Weinberger, D. R. (2003) ‘Neocortical modulation of the amygdala response to fearful stimuli’, Biological Psychiatry, 53(6), pp. 494–501.

Kamitani, Y. and Tong, F. (2005) ‘Decoding the visual and subjective contents of the human brain’, Nat Neurosci, 8(5), pp. 679–685.

Kelly, C., Biswal, B., Craddock, R. C., Castellanos, F. X. and Milham, M. P. (2012) ‘Characterizing variation in the functional connectome: promise and pitfalls’, Trends in cognitive sciences, 16(3), p. 10.1016/j.tics.2012.02.001.

Knops, A., Thirion, B., Hubbard, E. M., Michel, V. and Dehaene, S. (2009) ‘Recruitment of an Area Involved in Eye Movements During Mental Arithmetic’, Sciencexpress, pp. 1–4.

Korgaonkar, M. S., Fornito, A., Williams, L. M. and Grieve, S. M. (2014) ‘Abnormal Structural Networks Characterize Major Depressive Disorder: A Connectome Analysis’, Molecular and Neural Systems In Depression, 76(7), pp. 567–574.

Lee, J. H. (2012) ‘Informing brain connectivity with optogenetic functional magnetic resonance imaging’, NeuroImage, 62(4), pp. 2244–2249.

Lynall, M.-E., Bassett, D. S., Kerwin, R., McKenna, P. J., Kitzbichler, M., Müller, U. and Bullmore, E. (2010) ‘Functional connectivity and brain networks in schizophrenia’, The Journal of neuroscience : the official journal of the Society for Neuroscience, 30(28), pp. 9477–9487.

Maguire, E. A., Woollett, K. and Spiers, H. J. (2006) ‘London taxi drivers and bus drivers: A structural MRI and neuropsychological analysis’, Hippocampus, 16(12), pp. 1091–1101.

Martin, V. C., Schacter, D. L., Corballis, M. C. and Addis, D. R. (2011) ‘A role for the hippocampus in encoding simulations of future events’, Proceedings of the National Academy of Sciences of the United States of America, 108(33), pp. 13858–13863.

Mennes, M., Kelly, C., Zuo, X.-N., Di Martino, A., Biswal, B. B., Castellanos, F. X. and Milham, M. P. (2010) ‘Inter-individual differences in resting-state functional connectivity predict task-induced BOLD activity’, NeuroImage, 50(4), pp. 1690–1701.

Menon, V. (2011) ‘Large-scale brain networks and psychopathology: a unifying triple network model’, Trends in Cognitive Sciences, 15(10), pp. 483–506.

Michel, C. M. and Murray, M. M. (2012) ‘Towards the utilization of EEG as a brain imaging tool’, NeuroImage, 61(2), pp. 371–385.

Nair, A., Keown, C. L., Datko, M., Shih, P., Keehn, B. and Müller, R.-A. (2014) ‘Impact of methodological variables on functional connectivity findings in autism spectrum disorders’, Human Brain Mapping, 35(8), pp. 4035–4048.

Ohno, N., Katoh, M., Saitoh, Y. and Saitoh, S. (2016) ‘Recent advancement in the challenges to connectomics’, Microscopy, 65(2), pp. 97–107.

Park, H.-J. and Friston, K. (2013) ‘Structural and Functional Brain Networks: From Connections to Cognition’, Science, 342(6158), [online] Available from: http://science.sciencemag.org/content/342/6158/1238411.abstract.

Poldrack, R. A. (2006) ‘Can cognitive processes be inferred from neuroimaging data?’, Trends in Cognitive Sciences, 10(2), pp. 59–63.

Rabinovich, M. I., Friston, K. J. and Varona, P. (2012) Principles of Brain Dynamics: Global State Interactions, Cambridge, Massachusetts, MIT Press.

Reddy, L., Tsuchiya, N. and Serre, T. (2010) ‘Reading the mind’s eye: Decoding category information during mental imagery’, NeuroImage, 50(2), pp. 818–825.

Rissman, J. and Wagner, A. D. (2012) ‘Distributed Representations in Memory: Insights from Functional Brain Imaging’, Annual Review of Psychology, 63, pp. 101–28.

Ryali, S., Shih, Y.-Y. I., Chen, T., Kochalka, J., Albaugh, D., Fang, Z., Supekar, K., Lee, J. H. and Menon, V. (2016) ‘Combining optogenetic stimulation and fMRI to validate a multivariate dynamical systems model for estimating causal brain interactions’, NeuroImage, 132, pp. 398–405.

Ryali, S., Supekar, K., Chen, T. and Menon, V. (2011) ‘Multivariate dynamical systems models for estimating causal interactions in fMRI’, NeuroImage, 54(2), pp. 807–823.

Tong, F., Harrison, S. A., Dewey, J. A. and Kamitani, Y. (2012) ‘Relationship between BOLD amplitude and pattern classification of orientation-selective activity in the human visual cortex’, NeuroImage, 63(3), pp. 1212–1222.

Tong, F. and Pratte, M. S. (2012) ‘Decoding Patterns of Human Brain Activity’, Annual Review of Psychology, 63(1), pp. 483–509.

Vissers, M. E., X Cohen, M. and Geurts, H. M. (2012) ‘Brain connectivity and high functioning autism: A promising path of research that needs refined models, methodological convergence, and stronger behavioral links’, Neuroscience & Biobehavioral Reviews, 36(1), pp. 604–625.