Validated instruments such as questionnaires, patient-reported outcome measures and clinician-rated psychopathology scales, are indispensable for measuring symptom burden and mental state, and for defining outcomes in both psychiatric practice and clinical trials. Most often, the values on the instrument’s multiple items (dimensions) are added to derive a single, univariate (scalar) sum-score. Although this approach simplifies interpretation, there are always many possible combinations of individual items that can yield the same sum-score. Two patients can therefore obtain identical scores on a given instrument, despite having very different combinations of underlying item scores corresponding to different patterns of clinical symptoms. The same is also true when a single patient is measured at two different time points, where the resulting sum-scores can obscure changes that may be clinically meaningful. We present an alternative analytic framework, which leverages geometric concepts to represent measurements as points in a vector space. Using this framework, we show why sum-scores obscure information present in measurements of clinical state, and also provide a straightforward algorithm to mitigate against this problem. Clinically-relevant outcomes, such as remission or patient-centered treatment goals, can be represented intuitively, as reference points or ‘anchors’ within this space. Using real-world data, we then demonstrate how measuring the relative distance between points and anchors preserves more information, allowing outcomes such as proximity to remission, to be defined and measured.

Clinical trials in psychiatry inherit methods for design and statistical analysis from evidence-based medicine. However, trials in other clinical disciplines benefit from a more specific relationship between instruments that measure disease state (e.g. biomarkers, clinical signs), the underlying pathology and diagnosis such that primary outcomes can be readily defined. Trials in psychiatry use diagnosis (i.e. a categorical label for a syndrome) as a proxy for the underlying disorder, and outcomes are defined, for example, as a percentage change in a univariate total score on some clinical instrument. We label this approach to defining outcomes weak aggregation of disease state. Univariate measures are necessary, because statistical methodology is both tractable and well-developed for scalar outcomes, but we show that weak aggregate approaches do not capture disease state sufficiently, potentially leading to loss of information about response to intervention. We demonstrate how multivariate disease state can be captured using geometric concepts of spaces defined over routine clinical instruments, and show how clinically meaningful disease states (e.g. representing different profiles of symptoms, recovery or remission) can be defined as prototypes (geometric locations) in these spaces. Then, we show how to derive univariate (scalar) measures, which capture patient's relationships to these prototypes and argue these represent strong aggregates of disease state that may be a better basis for outcome measures. We demonstrate our proposal using a large publically available dataset. We conclude by discussing the impact of strong aggregates for analyses in traditional and novel trial designs.

The development of drugs to improve cognition in patients with schizophrenia is a major unmet clinical need. A number of promising compounds failed in recent clinical trials, a pattern linked to poor translation between preclinical and clinical stages of drug development. Seeking proof of efficacy in early Phase 1 studies in surrogate patient populations (for example, high schizotypy individuals where subtle cognitive impairment is present) has been suggested as a strategy to reduce attrition in the later stages of drug development. However, there is little agreement regarding the pattern of distribution of schizotypal features in the general population, creating uncertainty regarding the optimal control group that should be included in prospective trials. We aimed to address this question by comparing the performance of groups derived from the general population with low, average and high schizotypy scores over a range of cognitive and oculomotor tasks. We found that tasks dependent on frontal inhibitory mechanisms (N-Back working memory and anti-saccade oculomotor tasks), as well as a smooth-pursuit oculomotor task were sensitive to differences in the schizotypy phenotype. In these tasks the cognitive performance of 'low schizotypes' was significantly different from 'high schizotypes' with 'average schizotypes' having an intermediate performance. These results indicate that for evaluating putative cognition enhancers for treating schizophrenia in early-drug development studies the maximum schizotypy effect would be achieved using a design that compares low and high schizotypes.

Introduction Cognitive impairments are common in schizophrenia and impact disproportionately on real world functioning. Our current antipsychotic-medications do not offer any significant benefit for cognitive deficits. Psychological approaches have some positive-effects but require an integrated psychological and occupational focus to optimise cognitive performance, which has been difficult to implement in routine clinical-practice. Contemporary novel investigational drugs trialled in schizophrenia have failed to show any significant benefit for cognitive-symptoms;despite showing promise in earlier phase-2-studies(Goff-et-al.,2011;Choi-et-al.,2013). Recent data has suggested that there may be a subset of patients responding to interventions to improve cognitive performance(Vercammen-et-al,2011;Murthy-et-al,2012). Objectives We review the literature and use our own cognitive-training-data to examine how one might define this group, and propose a methodology for future clinical-studies of cognition in schizophrenia, predicated on the use of adaptive-designs incorporating subtyping into the fabric of the studies. Aims To categorise schizophrenia patients according to baseline performance and to investigate if this differentiation will predict their response to cognitive training(CT). Methods 47 schizophrenia-patients were recruited and classified in'learners'and'non-learners'based on learning-performance on day1-baseline-assessment measured by Mathew´s-correlation-coefficient and completed CT. We used multilevel-regressions to investigate differences between the defined groups in learning. Results According to MCC-performance at day-1(session two),24 participants were classified as'learners'and 23 as'non-learners'. We found significant-differences in response to CT-between the defined groups(p<0.0001). Download : Download full-size image Conclusions We were able to distinguish between responders/non-responders on baseline-assessment. Our results showed that CT-enhanced-performance in the 'learners-group'relative to'non-learners-group'. Trial-design needs to be adaptive to optimise outcomes in trials modifying cognitive-dysfunction-in-schizophrenia. One option would be to stratify the-sample on their early-baseline ability to respond to cognitive training-and to treat these cognitive-responders with medication and non-responders with an enhanced-programme of psychological-and-occupational-therapy.

BackgroundPeople with psychoses often report fixed, delusional beliefs that are sustained even in the presence of unequivocal contrary evidence. Such delusional beliefs are the result of integrating new and old evidence inappropriately in forming a cognitive model. We propose and test a cognitive model of belief formation using experimental data from an interactive ‘Rock Paper Scissors’ (RPS) game.MethodParticipants (33 controls and 27 people with schizophrenia) played a competitive, time-pressured interactive two-player game (RPS). Participants' behavior was modeled by a generative computational model using leaky integrator and temporal difference methods. This model describes how new and old evidence is integrated to form a playing strategy to beat the opponent and to provide a mechanism for reporting confidence in one's playing strategy to win against the opponent.ResultsPeople with schizophrenia fail to appropriately model their opponent's play despite consistent (rather than random) patterns that can be exploited in the simulated opponent's play. This is manifest as a failure to weigh existing evidence appropriately against new evidence. Furthermore, participants with schizophrenia show a ‘jumping to conclusions’ (JTC) bias, reporting successful discovery of a winning strategy with insufficient evidence.ConclusionsThe model presented suggests two tentative mechanisms in delusional belief formation: (i) one for modeling patterns in other's behavior, where people with schizophrenia fail to use old evidence appropriately, and (ii) a metacognitive mechanism for ‘confidence’ in such beliefs, where people with schizophrenia overweight recent reward history in deciding on the value of beliefs about the opponent.

Background: Auditory verbal hallucinations (AVH) are the most prevalent symptom in schizophrenia. They are associated with increased activation within the temporoparietal cortices and are refractory to pharmacological and psychological treatment in approximately 25% of patients. Low frequency repetitive transcranial magnetic stimulation (rTMS) over the temporoparietal cortex has been demonstrated to be effective in reducing AVH in some patients, although results have varied. The cortical mechanism by which rTMS exerts its effects remain unknown, although data from the motor system is suggestive of a local cortical inhibitory effect. We explored neuroimaging differences in healthy volunteers between application of a clinically utilized rTMS protocol and a sham rTMS equivalent when undertaking a prosodic auditory task.Method: Single-blind placebo controlled fMRI study of 24 healthy volunteers undertaking an auditory temporoparietal activation task, who received either right temporoparietal rTMS or sham RTMS.Results: The main effect of group was bilateral inferior parietal deactivation following real rTMS. An interaction of group and task type showed deactivation during real rTMS in the right superior temporal gyrus (STG), left thalamus, left postcentral gyrus and cerebellum. However, the left parietal lobe showed an increase in activation following right sided real rTMS, but this increase was specific to a non-linguistic, tone-sequence task.Conclusion: rTMS does cause local inhibitory effects. not only in the underlying region of application, but also in functionally connected cortical regions. However, there is also a related, task dependent, increase in activation within selected cortical areas in the contralateral hemisphere; these are likely to reflect compensatory mechanisms, and such cortical activation may in some cases contribute to, or retard, some of the therapeutic effects seen with rTMS. (C) 2009 Elsevier Ltd. All rights reserved.

This paper presents a new connectionist model of spatial language based on real psycholinguistic data. It puts together various constraints on object knowledge (“what”) and on object localisation (“where”) in order to influence the comprehension of a range of linguistic terms, mirroring what participants do in experiments. The computational model consists of a vision processing module for input scenes, an Elman network module for the representation of object dynamics, and a dual-route network for the production of object names and linguistic prepositions describing the scene. Preliminary simulations on the prediction of spatial term ratings are presented, and extensions of the model to vague quantifiers and other syntactic categories are considered.

When accepting a parcel from another person, we are able to use information about that person’s movement to estimate in advance the weight of the parcel, that is, to judge its weight from observed action. Perceptual weight judgment provides a powerful method to study our interpretation of other people’s actions, but it is not known what sources of information are used in judging weight. We have manipulated full form videos to obtain precise control of the perceived kinematics of a box lifting action, and use this technique to explore the kinematic cues that affect weight judgment. We find that observers rely most on the duration of the lifting movement to judge weight, and make less use of the durations of the grasp phase, when the box is first gripped, or the place phase, when the box is put down. These findings can be compared to the kinematics of natural box lifting behaviour, where we find that the duration of the grasp component is the best predictor of true box weight. The lack of accord between the optimal cues predicted by the natural behaviour and the cues actually used in the perceptual task has implications for our understanding of action observation in terms of a motor simulation. The differences between perceptual and motor behaviour are evidence against a strong version of the motor simulation hypothesis.

Embodied theories of cognition propose that symbol systems are analogue (e.g. Barsalou, 1999; Glenberg, 1997), as opposed to the classicist view that they are amodal e.g. Newell and Simon (1976), Fodor (1998). The fundamental problem of symbol grounding (Harnad, 1990) is resolved in embodied theories by admitting only theories of symbolic representation that are grounded in the perceptual system’s representation (rather than by reference or mapping of amodal symbols through the sensory systems of the agent). These are often called analogical representations (Mandler, 1998). Barsalou’s (1999) proposal for perceptual symbol systems (PSS) provides just such a framework for how analogue symbols might come into being, but remains agnostic on the implementation of these PSSs. In this paper, we advance an implementation of PSSs which might fill this explanatory gap. We provide descriptions, an implementation and results from a model and its consequences for Barsalou’s theory and embodied representations generally. We constrain our model to the visual modality, but without loss of generality.

One of the most active areas of multimedia research is into content based retrieval (CBR) of multimedia information. Using tools from the disparate disciplines of image processing, audio processing, video processing and others, approaches to CBR for the different media are being produced and integrated into multimedia systems, quite frequently in an ad hoc way. Also, traditional hypertext systems allow textual information to be arbitrarily linked so that users can navigate between related parts of the information in a system. Many of these systems can use multimedia information such as images, sounds and video clips that can also be linked. In such hypermedia systems the links are usually created using specific locations in particular documents. Some systems, such as Microcosm, also allow links to be created by specifying the text that forms one anchor of the link. These links are created once and can be followed from any location where the text occurs. To achieve this, the system has to examine documents currently being viewed by the user and look for matches between text in the documents and text that forms anchors of links within the system. Where matches occur the system can highlight the text as a source anchor for a link. This is a form of content based navigation(CBN). Navigational links are dynamically created by matching the content of currently viewed documents with the content of previously created link anchors. For textual documents this is relatively easy to implement since comparing text strings is often straightforward. Content based navigation for multimedia documents also involves comparing and matching selections of images, video and sound with each other. Our MAVIS 1 project addressed the problem of integrated content based retrieval and navigation from non-text media but there are many limitations associated with such systems. This paper will describe some of the reasons for these limitations and present an overview of MAVIS 2, a new architecture for multimedia content based retrieval and navigation. The architecture not only presents a consistent cross media approach to CBR and CBN but also includes the integration of a multimedia thesaurus which can substantially improve the flexibility and versatility of the multimedia information system, enhancing the capabilities of both content based retrieval and content based navigation. MAVIS stands for Multimedia Architecture for Video, Image and Sound and the MAVIS 2 architecture also includes integration of intelligent agents which support navigation by utilising both the media content and semantic concepts in the multimedia thesaurus to develop rapid paths from media based features to semantic concepts or to generate pseudo thesaurus groupings by feature clustering when no semantic relations are available. MAVIS 2 is currently being implemented and the paper will not only present details of the architecture but also show examples of retrieval and navigation with the prototype system.

The purpose of this paper is two-fold. We begin by exploring the emerging trend to view multimedia information in terms of low-level and high-level components; the former being feature-based and the latter the 'semantics' intrinsic to what is portrayed by the media object. Traditionally, this has been viewed by employing analogies with generative linguistics. Recently, a new perceptive based on the semiotic tradition has been alluded to in several papers. We believe this to be a more appropriate approach. From this, we propose an approach for tackling this problem which uses an associative data structure expressing authored information together with intelligent agents acting autonomously over this structure. We then show how neural networks can be used to implement such agents. The agents act as 'vehicles' for bridging the gap between multimedia semantics and concrete expressions of high-level knowledge, but we suggest that traditional neural network techniques for classification are not architecturally adequate.