[Artificial neural networks as a psychiatric instrument]. / Artificiële neurale netwerken als psychiatrisch instrument.

BACKGROUND: Artificial intelligence (AI) has evolved enormously over the past decade and is increasingly being applied to a range of domains, including psychiatry. AI encompasses several modalities, including artificial neural networks (ANNs), referring to computer models partly based on the workings of the brain. ANNs have existed since the ’50s, but only became ‘mainstream’ since the 2010s. The fact that they are inspired by the workings of the brain raises the question of whether they can also be used to model the (dys)functioning of the brain. This question led to the advent of the research field ‘computational psychiatry’. AIM: This article aims at providing an accessible introduction to artificial neural networks, and potential applications hereof in contemporary psychiatric practice. METHOD: Literature review with some examples. RESULTS: In this article we try to outline with some concrete examples what artificial neural networks are and how they can be used to model mechanisms in the brain. We successively discuss ANNs as a model of the human visual system, as a model of prosopagnosia and as a model of auditory hallucinations and finally as a model of autism spectrum disorder. We also describe a number of limitations of this approach. CONCLUSION: A computer model that models the entire brain is challenging at present, but current models can help in testing hypotheses concerning possible mechanisms that give rise to a wide range of neuropsychiatric conditions.

A longitudinal study of the association between basal ganglia volumes and psychomotor symptoms in subjects with late life depression undergoing ECT.

Psychomotor dysfunction (PMD) is a core element and key contributor to disability in late life depression (LLD), which responds well to electroconvulsive therapy (ECT). The neurobiology of PMD and its response to ECT are not well understood. We hypothesized that PMD in LLD is associated with lower striatal volume, and that striatal volume increase following ECT explains PMD improvement. We analyzed data from a two-center prospective cohort study of 110 LLD subjects (>55 years) receiving ECT. Brain MRI and assessment of mood, cognition, and PMD was performed 1 week before, 1 week after, and 6 months after ECT. Volumetry of the caudate nucleus, putamen, globus pallidus, and nucleus accumbens was derived from automatically segmented brain MRIs using Freesurfer®. Linear multiple regression analyses were used to study associations between basal ganglia volume and PMD. Brain MRI was available for 66 patients 1 week post ECT and in 22 patients also six months post ECT. Baseline PMD was associated with a smaller left caudate nucleus. One week after ECT, PMD improved and volume increases were detected bilaterally in the caudate nucleus and putamen, and in the right nucleus accumbens. Improved PMD after ECT did not relate to the significant volume increases in these structures, but was predicted by a nonsignificant volume change in the right globus pallidus. No volume differences were detected 6 months after ECT, compared to baseline. Although PMD is related to lower striatal volume in LLD, ECT-induced increase of striatal volume does not explain PMD improvement.

Audiovisual emotion recognition in schizophrenia: reduced integration of facial and vocal affect.

Since Kraepelin called dementia praecox what we nowadays call schizophrenia, cognitive dysfunction has been regarded as central to its psychopathological profile. Disturbed experience and integration of emotions are, both intuitively and experimentally, likely to be intermediates between basic, non-social cognitive disturbances and functional outcome in schizophrenia. While a number of studies have consistently proven that, as part of social cognition, recognition of emotional faces and voices is disturbed in schizophrenics, studies on multisensory integration of facial and vocal affect are rare. We investigated audiovisual integration of emotional faces and voices in three groups: schizophrenic patients, non-schizophrenic psychosis patients and mentally healthy controls, all diagnosed by means of the Schedules of Clinical Assessment in Neuropsychiatry (SCAN 2.1). We found diminished crossmodal influence of emotional faces on emotional voice categorization in schizophrenics, but not in non-schizophrenia psychosis patients. Results are discussed in the perspective of recent theories on multisensory integration.