The high technical barrier to entry in the field of neuroimaging can hinder early insight from promising results and the development of evidence-based clinical practice.

The working group focused on published literature in order to develop a new methodology in the analysis, visualization, and representation of fMRI data in the psychiatric setting.

Three valid and established measures were chosen, in order to achieve dimensionality reduction, stability and explainability of results, namely Regional-Homogeneity; fractional Amplitude of Low-Frequency Fluctuations; Eigenvector-Centrality. Each measure was color coded and individual images per subject compiled, averaging results by functional networks as described the FIND lab of the University of Stanford. 272 individual scans were processed (130 neurotypicals, 50 patients with Schizophrenia, 49 with Bipolar Disorder, 43 with ADHD).

The discriminative power between clinical groups of the novel method was significant both by human eye, and later confirmation by statistical tests, and by computer vision algorithms (Convolutional Neural Networks). The precision-recall Area Under the Curve, dividing by 80/20 proportion between train and test sets, was >84.5% for each group. The group of patients with Bipolar Disorder showed a partial overlap with the group of patients suffering from Schizophrenia – by a dominance of Eigenvector-Centrality and Regional-Homogeneity, as well as a lower prevalence of fractional Amplitude of Low-Frequency Fluctuations, for both in comparison to controls.

The present study offers preliminary evidence for the adoption of i-ECO (integrated-Explainability through Color Coding) in fMRI analyses during rest in the Psychiatric field.

No significant relationships.

One of the most perplexing and characteristic symptoms of the schizophrenia (SZ) patients is hallucination. The occurrence of hallucinations to be associated with altered activity in the auditory and visual cortex but is not well understood from the brain functional network dynamics in SZ.

To explore the brain abnormal basis of hallucinations in SZ with the dynamic functional connectivity (dFC).

Using magnetic resonance imaging for 83 SZ patients and 83 matched healthy controls and independent component analysis, 52 independent components (ICs) were identified as nodes and assigned into eight intrinsic connectivity networks (Figure 1A). Subsequently, we established dFC matrices and clustered them into four discrete states (Figure 1B) and three state transition metrics were obtained. To further explore the changes in the centrality of each component, eigenvector centrality (EC) was calculated and its time-varying was evaluated.

Compared to controls with FDR correction, we found that patients had more mean dwell times and fractional time in state 1 (P=0.0081 and P=0.0018), mainly with hypoconnectivity between auditory and visual network and other networks and hyperconnectivity between language and default-mode network (DMN). While, patients had less dwell times and fractional time in state 3 (P=0.0018 and P=0.0009), and decreased FC between visual network and executive control network (ECN) and increased FC between ECN and DMN than controls (Figure 2).

EC statistics showed that SZs displayed increased temporal dynamics in visual-related regions (Figure 3).

SZ was mainly manifested as altered dFC and temporal variability of nodal centrality in auditory and visual networks.

No significant relationships.