Rutgers-Princeton Center for Computational Cognitive Neuropsychiatry

Nicolas Schuck
Princeton Neuroscience Institute
Princeton University

Prefrontal representations of environmental states are commonly assumed to reflect a mental model of the world that guides learning, memory and decision-making. The nature of such state representations in the human brain is not well studied, however. In particular little is know about how the current state of the environment is reflected in frontal activity and how an established mental model is changed, a process that requires internal simulation and is important for the ability to make strategy improvements. I will present two studies in which we investigated such prefrontal state representations during stable task performance or sudden strategy changes. Results from fMRI pattern classification analyses showed that such representations can be decoded from orbitofrontal cortex and signals from medial PFC reflected internally simulated strategy changes.

Megan deBettencourt
Princeton Neuroscience Institute
Princeton University

Remaining focused is important for most everyday tasks, such as being productive at work or driving safely in traffic. Despite this importance, sustaining attention is difficult and prone to frequent lapses. We hypothesized that some of these behavioral errors result from an inability to accurately monitor one’s own attentional state, and that enhancing this metacognitive awareness could lead to lasting improvements in sustained attention. I’ll first describe the development of a real-time fMRI system that involves measuring attentional state and providing closed-loop neurofeedback by altering the stimuli. Then, I’ll present work from a study that used this feedback to demonstrate a training effect: participants who received accurate neurofeedback improved on a sustained attention task. Finally, I’ll extend these results to a pilot study that applied similar techniques to a group of depressed individuals, to train away a bias towards negatively valenced information. Together, these studies suggest that real-time fMRI may enable powerful, customized, and rapid cognitive training.

Elizabeth Torres
Psychology, Cognitive Science, Neuroscience, Biomedical Engineering
Computational Biomedical Imaging and Modeling
Rutgers University

There is a critical need for new analytics to personalize behavioral data analysis across different fields, including kinesiology, sports science and behavioral neuroscience. Specifically, to better translate and integrate basic research into patient care we need to radically transform the methods by which we describe and interpret movement data. We introduce a new statistical platform that enables the analyses of fluctuations in motor performance from natural movements and access various data sets to illustrate their use. These include data from newborns; data from large open-databases where we examine the effects of psychotropic medications on the volitional control of bodily rhythms and data from a large cohort involving various disorders of the nervous system. We show that hidden in the ‘noise’, smoothed out by averaging movement kinematics data, lies a wealth of information that can detect very early risk for neurodevelopmental derail/stagnation; and selectively differentiates neurological and mental disorders such as Parkinson’s disease (PD), deafferentation, Autism Spectrum Disorders (ASD), and Schizophrenia (SZ) from typically developing and typically aging controls. Further we show how to use these methods to assess medication effects on motor control. We empirically estimate the statistical parameters of the probability distributions for each individual in the various groups and report the parameter ranges for each clinical group after characterization of healthy developing and aging groups. We coin this newly proposed platform for individualized behavioral analyses ‘precision phenotyping’ to distinguish it from the type of observational-behavioral phenotyping prevalent in clinical studies or from the ‘one-size-fits-all’ model in basic movement science. We further propose the use of this platform as a unifying statistical framework to characterize brain disorders of known etiology in relation to idiopathic neurological disorders with similar phenotypic manifestations.

For background, please see: http://journal.frontiersin.org/article/10.3389/fneur.2016.00008/full

Nathaniel Daw
Princeton Neuroscience Institute and Psychology Department
Princeton University

I will present a very informal overview of a line of work I’ve been involved with for some years. This concerns a particular variable — the long-run average reward per timestep — which we believe the brain tracks and uses to guide behavior in a variety of circumstances. These include choices involving energetic expenditure, vigor and motivation; aspiration levels and whether to settle or seek alternative options in situations like foraging or mate selection; patience and time discounting; and self control and automaticity vs. deliberation. Computationally, this is because at least in some circumstances, the average reward measures the opportunity cost of time spent, which gives it a key role assessing tradeoffs and prioritization among different possible options. I’ll discuss some of our attempts to measure peoples’ use of this quantity experimentally, and to relate it to underlying biological systems including dopamine and stress hormones. The main goal of this presentation, however, is to brainstorm with the CCNP community as to whether this mechanism is relevant to disorders of mood, particularly depression, and about experimental strategies we might pursue to test this.

Brian Keane
University Behavioral Health Care
Rutgers University

Schizophrenia has been extensively studied for over a century but there are not yet any objective tests that reliably tell us about the presence, stage, or state of the illness.  Here, I make the case that brief, inexpensive, and noninvasive behavioral assessments of visual functioning yield surprising insights.   Three processes are specifically considered: contour integration, in which a collinear group of edges pops out of a cluttered array; collinear facilitation, in which a low contrast oriented element becomes easier to see when flanked by collinear elements; and visual shape completion, in which quartets of sectored circles (pac-men) become easier to discriminate when forming a single globally completed shape.  I provide psychophysical evidence that each of the foregoing processes is impaired in schizophrenia.  The deficits are large in magnitude, reliant upon specific visual cortical structures, and unexplained by group differences in medication, motivation, or attention.  The deficits at times are specific to schizophrenia, arise by the first psychotic episode, and worsen with prolonged illness duration or elevated conceptual disorganization.  Taken together, behaviorally established visual deficits may furnish a sort of biomarker, objectively flagging the presence, stage, or state of schizophrenia.

 

Yael Niv
Princeton Neuroscience Institute and Psychology Department
Princeton University

In the past couple of decades reinforcement learning has emerged as a central framework for thinking of trial and error learning in the basal ganglia. However, it has been difficult to link synaptic modification to overt behavioral changes. Rodent models of DYT1 dystonia, a single-mutation motor disorder, demonstrate increased long-term potentiation and decreased long-term depression in corticostriatal synapses. Computationally, such asymmetric learning predicts risk taking in probabilistic tasks. Here we test DYT1 dystonia patients on a simple reinforcement learning task, and demonstrate abnormal risk taking correlated with disease severity, thereby implicating striatal plasticity in shaping choice behavior in humans. Our results are also relevant to the CCNP community as they suggest (and demonstrate) that behavioral tasks married to precise computational models may provide a non-invasive window to diagnosing and characterizing underlying neurological and mental disorders.

Susan Murphy
HE Robbins Distinguished University Professor of Statistics,
Research Professor, Institute for Social Research & Professor of Psychiatry
University of Michigan

Mobile Health concerns the use of mobile devices for both collecting real-time data, for processing these data and for delivering real-time treatment. Some of these treatments are accessible via the device 24/7 whereas other treatments are intended to be pushed adaptively and just-in-time to the user. Might there be therapeutic effect provided by varying whether or not to push a treatment or to varying the type of treatment? A potential therapeutic role of variation is inconsistent with the usual model of a Markov Decision Process. How do we rectify this inconsistency?

Steve Silverstein
Professor of Psychiatry, Rutgers Robert Wood Johnson Medical School, and
Director of Research, Rutgers University Behavioral Health Care

Schizophrenia is a heterogeneous syndrome in terms of symptoms and course of illness.  While much work has focused on understanding the nature of psychotic symptoms, and to a lesser extent negative symptoms, the highly disabling disorganization syndrome has received little attention.  This presentation will begin by giving clinical examples of disorganization in thought, language, and perception.  Experimental data will then be used to demonstrate examples of disorganization in perception, their relationships to thought disorder, and the view that each of these is a manifestation of a canonical processing impairment involving coordination of cognitive activity based on context. The information theoretic concept of coherent infomax will then be described as a way to formally understand contextual modulation of perceptual and cognitive activity, and its failures in schizophrenia.
Implementation of the neural goal function of coherent infomax via apical amplification and dis-amplification will be described, as will the implications of this view for an understanding of schizophrenia.