Integrating machining learning and multimodal neuroimaging to detect schizophrenia at the level of the individual.

Schizophrenia is a severe psychiatric disorder associated with both structural and functional brain abnormalities. In the past few years, there has been growing interest in the application of machine learning techniques to neuroimaging data for the diagnostic and prognostic assessment of this disorder. However, the vast majority of studies published so far have used either structural or functional neuroimaging data, without accounting for the multimodal nature of the disorder. Structural MRI and resting-state functional MRI data were acquired from a total of 295 patients with schizophrenia and 452 healthy controls at five research centers. We extracted features from the data including gray matter volume, white matter volume, amplitude of low-frequency fluctuation, regional homogeneity and two connectome-wide based metrics: structural covariance matrices and functional connectivity matrices. A support vector machine classifier was trained on each dataset separately to distinguish the subjects at individual level using each of the single feature as well as their combination, and 10-fold cross-validation was used to assess the performance of the model. Functional data allow higher accuracy of classification than structural data (mean 82.75% vs. 75.84%). Within each modality, the combination of images and matrices improves performance, resulting in mean accuracies of 81.63% for structural data and 87.59% for functional data. The use of all combined structural and functional measures allows the highest accuracy of classification (90.83%). We conclude that combining multimodal measures within a single model is a promising direction for developing biologically informed diagnostic tools in schizophrenia.

Detecting schizophrenia at the level of the individual: relative diagnostic value of whole-brain images, connectome-wide functional connectivity and graph-based metrics.

BACKGROUND: Previous studies using resting-state functional neuroimaging have revealed alterations in whole-brain images, connectome-wide functional connectivity and graph-based metrics in groups of patients with schizophrenia relative to groups of healthy controls. However, it is unclear which of these measures best captures the neural correlates of this disorder at the level of the individual patient. METHODS: Here we investigated the relative diagnostic value of these measures. A total of 295 patients with schizophrenia and 452 healthy controls were investigated using resting-state functional Magnetic Resonance Imaging at five research centres. Connectome-wide functional networks were constructed by thresholding correlation matrices of 90 brain regions, and their topological properties were analyzed using graph theory-based methods. Single-subject classification was performed using three machine learning (ML) approaches associated with varying degrees of complexity and abstraction, namely logistic regression, support vector machine and deep learning technology. RESULTS: Connectome-wide functional connectivity allowed single-subject classification of patients and controls with higher accuracy (average: 81%) than both whole-brain images (average: 53%) and graph-based metrics (average: 69%). Classification based on connectome-wide functional connectivity was driven by a distributed bilateral network including the thalamus and temporal regions. CONCLUSION: These results were replicated across the three employed ML approaches. Connectome-wide functional connectivity permits differentiation of patients with schizophrenia from healthy controls at single-subject level with greater accuracy; this pattern of results is consistent with the 'dysconnectivity hypothesis' of schizophrenia, which states that the neural basis of the disorder is best understood in terms of system-level functional connectivity alterations.

Effects of complement gene-set polygenic risk score on brain volume and cortical measures in patients with psychotic disorders and healthy controls.

Multiple genome-wide association studies of schizophrenia have reported associations between genetic variants within the MHC region and disease risk, an association that has been partially accounted for by alleles of the complement component 4 (C4) gene. Following on previous findings of association between both C4 and other complement-related variants and memory function, we tested the hypothesis that polygenic scores calculated based on identified schizophrenia risk alleles within the "complement" system would be broadly associated with memory function and associated brain structure. We tested this using a polygenic risk score (PRS) calculated for complement genes, but excluding C4 variants. Higher complement-based PRS scores were observed to be associated with lower memory scores for the sample as a whole (N = 620, F change = 8.25; p = .004). A significant association between higher PRS and lower hippocampal volume was also observed (N = 216, R2 change = 0.016, p = .015). However, after correcting for further testing of association with the more general indices of cortical thickness, surface area or total brain volume, none of which were associated with complement, the association with hippocampal volume became non-significant. A post-hoc analysis of hippocampal subfields suggested an association between complement PRS and several hippocampal subfields, findings that appeared to be particularly driven by the patient sample. In conclusion, our study yielded suggestive evidence of association between complement-based schizophrenia PRS and variation in memory function and hippocampal volume.

Effects of MiR-137 genetic risk score on brain volume and cortical measures in patients with schizophrenia and controls.

Multiple genome-wide association studies of schizophrenia have implicated genetic variants within the gene encoding microRNA-137. As risk variants within or regulated by MIR137 have been implicated in memory performance, we investigated the additive effects of schizophrenia-associated risk variants in genes empirically regulated by MIR137 on brain regions associated with memory function. A polygenic risk score (PRS) was calculated (at a p = 0.05 threshold), using this empirically regulated MIR137 gene set, to investigate associations between this PRS and structural brain measures. These measures included total brain volume, cortical thickness, cortical surface area, and hippocampal volume, in a sample of 216 individuals consisting of healthy participants (n = 171) and patients with psychosis (n = 45). We did not observe a significant association between MIR137 PRS and these cortical thickness, surface area or hippocampal volume measures linked to memory function; a significant association between increasing PRS and decreasing total brain volume, independent of diagnosis status (R2 = 0.008, Beta = -0.09, p = 0.029), was observed. This did not survive correction for multiple testing. In conclusion, our study yielded only suggestive evidence that risk variants interacting with MIR137 impacts on cortical structure.

Graph Convolutional Networks Reveal Network-Level Functional Dysconnectivity in Schizophrenia.

BACKGROUND AND HYPOTHESIS: Schizophrenia is increasingly understood as a disorder of brain dysconnectivity. Recently, graph-based approaches such as graph convolutional network (GCN) have been leveraged to explore complex pairwise similarities in imaging features among brain regions, which can reveal abstract and complex relationships within brain networks. STUDY DESIGN: We used GCN to investigate topological abnormalities of functional brain networks in schizophrenia. Resting-state functional magnetic resonance imaging data were acquired from 505 individuals with schizophrenia and 907 controls across 6 sites. Whole-brain functional connectivity matrix was extracted for each individual. We examined the performance of GCN relative to support vector machine (SVM), extracted the most salient regions contributing to both classification models, investigated the topological profiles of identified salient regions, and explored correlation between nodal topological properties of each salient region and severity of symptom. STUDY RESULTS: GCN enabled nominally higher classification accuracy (85.8%) compared with SVM (80.9%). Based on the saliency map, the most discriminative brain regions were located in a distributed network including striatal areas (ie, putamen, pallidum, and caudate) and the amygdala. Significant differences in the nodal efficiency of bilateral putamen and pallidum between patients and controls and its correlations with negative symptoms were detected in post hoc analysis. CONCLUSIONS: The present study demonstrates that GCN allows classification of schizophrenia at the individual level with high accuracy, indicating a promising direction for detection of individual patients with schizophrenia. Functional topological deficits of striatal areas may represent a focal neural deficit of negative symptomatology in schizophrenia.

Beyond C4: Analysis of the complement gene pathway shows enrichment for IQ in patients with psychotic disorders and healthy controls.

Variation in cognitive performance, which strongly predicts functional outcome in schizophrenia (SZ), has been associated with multiple immune-relevant genetic loci. These loci include complement component 4 (C4A), structural variation at which was recently associated with SZ risk and synaptic pruning during neurodevelopment and cognitive function. Here, we test whether this genetic association with cognition and SZ risk is specific to C4A, or extends more broadly to genes related to the complement system. Using a gene-set with an identified role in "complement" function (excluding C4A), we used MAGMA to test if this gene-set was enriched for genes associated with human intelligence and SZ risk, using genome-wide association summary statistics (IQ; N = 269 867, SZ; N = 105 318). We followed up this gene-set analysis with a complement gene-set polygenic score (PGS) regression analysis in an independent data set of patients with psychotic disorders and healthy participants with cognitive and genomic data (N = 1000). Enrichment analysis suggested that genes within the complement pathway were significantly enriched for genes associated with IQ, but not SZ. In a gene-based analysis of 90 genes, SERPING1 was the most enriched gene for the phenotype of IQ. In a PGS regression analysis, we found that a complement pathway PGS associated with IQ genome-wide association studies statistics also predicted variation in IQ in our independent sample. This association (observed across both patients and controls) remained significant after controlling for the relationship between C4A and cognition. These results suggest a robust association between the complement system and cognitive function, extending beyond structural variation at C4A.