Psychotic disorders occur as a result of pathobiochemical processes in the brain, which disrupt the central neurotransmission of molecules such as dopamine and glutamate. The dopamine hypothesis, adopted more than 2 decades ago, has repeatedly asserted its position as an etiopathogenetic substrate through the action of psychostimulants and neuroleptics on the mesolimbic and mesocortical systems, giving insight into the origin of positive and negative schizophrenic symptoms. On the other hand, cognitive impairments in schizophrenia remain not fully understood but are thought to be present during all stages of the disease, as well as in the prodromal the interictal and residual phases. Over the last decade, functional magnetic resonance imaging has focused on research of brain networks like the Default mode network, the Salience network and Central executive network, enabling a deeper understanding of cognitive deficits, as well as other phenomena such as disorganization of thought and behavior. The study of the nodes of these networks, such as the precuneus and insula, informs about their complex significant roles as structures responsible for important cognitive domains such as concentration, attention, ability to understand and reproduce information, as well as memory functions. It is suggested that the neurotransmission of dopamine and glutamate play a key role in these processes and their successful modulation in the correct brain regions through psychopharmacological and biomedical instrumental methods may lead to a significant reversal of conventional paradigms. Pharmaco-magnetic resonance imaging is a neuroimaging method that can provide the translation of scientific knowledge about the neural networks and the disruptions in and between different brain regions, into clinically applicable and effective therapeutic results in the management of severe psychotic disorders.

This study was conducted to examine whether there are quantitative or qualitative differences in the connectome between psychiatric patients and healthy controls and to delineate the connectome features of major depressive disorder (MDD), schizophrenia (SCZ), bipolar disorder (BD) and the severity of these disorders. Toward this end, we have performed effective connectivity analysis of resting state functional MRI data in these three patient groups and healthy controls. We have used spectral Dynamic Causal Modeling (spDCM), and the derived connectome features were further subjected to machine learning. The results outlined a model of 5 connections, which discriminate patients from controls, comprising major nodes of the limbic system (amygdala (AMY), hippocampus (HPC) and anterior cingulate cortex (ACC)), the salience network (anterior insula (AI), fronto-parietal and dorsal attention network (middle frontal gyrus (MFG) corresponding to dorsolateral prefrontal cortex, frontal eye field (FEF)). Notably, the alterations in the self-inhibitory connection of the anterior insula emerged as a feature of both mood disorders and SCZ. Moreover, 4 out of the 5 connectome features that discriminate mental illness from controls are features of mood disorders (both MDD and BD), namely the MFG→FEF, HPC→FEF, AI→AMY, and MFG→AMY connections, whereas one connection is a feature of SCZ, namely the AMY→SPL connectivity. A large part of the variance in the severity of depression (31.6%) and SCZ (40.6%) was explained by connectivity features. In conclusion, dysfunctions in the self-regulation of the salience network may underpin major mental disorders, while other key connectome features shape differences between mood disorders and SCZ, and can be used as potential imaging biomarkers.

In order to overcome this problem our group designed a novel machine learning technique, multivariate linear method (MLM) which can capture convergent data from voxel-based morphometry, functional resting state and task-related neuroimaging and the relevant clinical measures. In this paper we report results from convergent cross-validation of biological signatures of disease in a sample of patients with schizophrenia as compared to depression. Our model provides evidence that the combination of the neuroimaging and clinical data in MLM analysis can inform the differential diagnosis in terms of incremental validity to reach 90 % accuracy of the prediction.