29 October 2019 in Molecular Psychiatry
11 June 2019 in PLOS Computational Biology
Popular computational models of decision-making make specific assumptions about learning processes that may cause them to underfit observed behaviours. Here we suggest an alternative method using recurrent neural networks (RNNs) to generate a flexible family of models that have sufficient capacity to represent the complex learning and decision- making strategies used by humans. In this approach, an RNN is trained to predict the next action that a subject will take in a decision-making task and, in this way, learns to imitate the processes underlying subjects’ choices and their learning abilities. We demonstrate the benefits of this approach using a new dataset drawn from patients with either unipolar (n = 34) or bipolar (n = 33) depression and matched healthy controls (n = 34) making decisions on a two-armed bandit task. The results indicate that this new approach is better than baseline reinforcement-learning methods in terms of overall performance and its capacity to predict subjects’ choices. We show that the model can be interpreted using off-policy simulations and thereby provides a novel clustering of subjects’ learning processes—something that often eludes traditional approaches to modelling and behavioural analysis. Computational models of decision-making provide a quantitative characterisation of the learning and choice processes behind human actions. Designing a computational model is often based on manual engineering which an iterative process to examine the consistency between different aspects of the model and the empirical data. In practice, however, inconsistencies between the model and observed behaviours can remain hidden behind examined summary statistics. To address this limitation, we developed a recurrent neural network (RNNs) as a flexible type of model that can automatically characterize human decision-making processes without requiring tweaking and engineering. To show the benefits of this new approach, we collected data on a decision-making task conducted on subjects with either bipolar or unipolar depression, as well as healthy controls. The results showed that, indeed, important aspects of decision-making remained uncaptured by typical computational models and even their enhanced variants, but were captured by RNNs automatically. Further, we were able to show that the nature of such processes can be unveiled by simulating the model under various conditions. This new approach can be used, therefore, as a standalone model of decision-making or as a baseline model to evaluate how well other candidate models fit observed data.
1 March 2017 in Biological Psychology
23 October 2021 in Translational Psychiatry
Bipolar disorder (BD) is commonly misdiagnosed as major depressive disorder (MDD). This is understandable, as depression often precedes mania and is otherwise indistinguishable in both. It is therefore imperative to identify neural mechanisms that can differentiate the two disorders. Interrogating resting brain neural activity may reveal core distinguishing abnormalities. We adopted an a priori approach, examining three key networks documented in previous mood disorder literature subserving executive function, salience and rumination that may differentiate euthymic BD and MDD patients. Thirty-eight patients with BD, 39 patients with MDD matched for depression severity, and 39 age-gender matched healthy controls, completed resting-state fMRI scans. Seed-based and data-driven Independent Component analyses (ICA) were implemented to examine group differences in resting-state connectivity (pFDR < 0.05). Seed analysis masks were target regions identified from the fronto-parietal (FPN), salience (SN) and default-mode (DMN) networks. Seed-based analyses identified significantly greater connectivity between the subgenual cingulate cortex (DMN) and right dorsolateral prefrontal cortex (FPN) in BD relative to MDD and controls. The ICA analyses also found greater connectivity between the DMN and inferior frontal gyrus, an FPN region in BD relative to MDD. There were also significant group differences across the three networks in both clinical groups relative to controls. Altered DMN–FPN functional connectivity is thought to underlie deficits in the processing, management and regulation of affective stimuli. Our results suggest that connectivity between these networks could potentially distinguish the two disorders and could be a possible trait mechanism in BD persisting even in the absence of symptoms.
1 January 2017 in NeuroImage: Clinical
Functional neurological symptom disorder refers to the presence of neurological symptoms not explained by neurological disease. Although this disorder is presumed to reflect abnormal function of the brain, recent studies in adults show neuroanatomical abnormalities in brain structure. These structural brain abnormalities have been presumed to reflect long-term adaptations to the disorder, and it is unknown whether child and adolescent patients, with illness that is typically of shorter duration, show similar deficits or have normal brain structure. High-resolution, three-dimensional T1-weighted magnetic resonance images (MRIs) were acquired in 25 patients (aged 10-18 years) and 24 healthy controls. Structure was quantified in terms of grey matter volume using voxel-based morphometry. Post hoc, we examined whether regions of structural difference related to a measure of motor readiness to emotional signals and to clinical measures of illness duration, illness severity, and anxiety/depression. Patients showed greater volumes in the left supplementary motor area (SMA) and right superior temporal gyrus (STG) and dorsomedial prefrontal cortex (DMPFC) (corrected p < 0.05). Previous studies of adult patients have also reported alterations of the SMA. Greater SMA volumes correlated with faster reaction times in identifying emotions but not with clinical measures. The SMA, STG, and DMPFC are known to be involved in the perception of emotion and the modulation of motor responses. These larger volumes may reflect the early expression of an experience-dependent plasticity process associated with increased vigilance to others' emotional states and enhanced motor readiness to organize self-protectively in the context of the long-standing relational stress that is characteristic of this disorder.
2 March 2021 in Translational Psychiatry
Behavioural disturbances in attention deficit hyperactivity disorder (ADHD) are thought to be due to dysfunction of spatially distributed, interconnected neural systems. While there is a fast-growing literature on functional dysconnectivity in ADHD, far less is known about the structural architecture underpinning these disturbances and how it may contribute to ADHD symptomology and treatment prognosis. We applied graph theoretical analyses on diffusion MRI tractography data to produce quantitative measures of global network organisation and local efficiency of network nodes. Support vector machines (SVMs) were used for comparison of multivariate graph measures of 37 children and adolescents with ADHD relative to 26 age and gender matched typically developing children (TDC). We also explored associations between graph measures and functionally-relevant outcomes such as symptom severity and prediction of methylphenidate (MPH) treatment response. We found that multivariate patterns of reduced local efficiency, predominantly in subcortical regions (SC), were able to distinguish between ADHD and TDC groups with 76% accuracy. For treatment prognosis, higher global efficiency, higher local efficiency of the right supramarginal gyrus and multivariate patterns of increased local efficiency across multiple networks at baseline also predicted greater symptom reduction after 6 weeks of MPH treatment. Our findings demonstrate that graph measures of structural topology provide valuable diagnostic and prognostic markers of ADHD, which may aid in mechanistic understanding of this complex disorder.