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

Abstract

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.

Translating research findings into clinical practice: a systematic and critical review of neuroimaging-based clinical tools for brain disorders

“…The same cannot be said for the neuroanatomical alterations that are typically observed in psychiatric disorders. Here, neuroanatomical alterations tend to be subtle and widespread, making the discrimination between normal heterogeneity and pathological heterogeneity more challenging 63,70,71 . This means that the adaptation of existing tools to psychiatric disorders would require careful consideration of the issue of heterogeneity 72 .…”

“…• From a region-of-interest to whole-brain approach: Existing tools for neurological disorders use a region-of-interest approach to detect localized alternations. Considering the subtle and widespread neural correlates of the psychiatric disorders 22,71,76 , the ideal clinical tool should not restrict its analysis to a single or few regions; instead, it should analyze the whole-brain to exploit all the available neuroanatomical information.…”

“…• Accounting for inter-scanner variability: As the ideal clinical tool is supposed to handle MRI scans of individuals from different clinicians/hospitals/ countries, it should be able to estimate and account for differences in scanner provider, magnetic strength field and acquisition parameters. This is especially important for psychiatric disorders, where the effects of interest are subtle and, therefore, inter-scanner variability can be much greater than disease-related variability 71,80,81 .…”

From models to tools: clinical translation of machine learning studies in psychosis

“…1 Furthermore, within the span of ten years, small local studies with few dozen participants have evolved into large multi-centre studies with several hundreds of participants. [2][3][4][5] In the midst of the search for accurate models, much attention has been given to methodological challenges including the impact of sample size, 6,7 the limitations of traditional case-control designs, 8,9 how to best deal with confounding variables 10 and the effects of heterogeneity 11,12 and inter-scanner variability, 13 just to mention a few. Although there are still important methodological challenges to overcome, substantial progress is being made, and a solution to these challenges is now considered to be a matter of when rather than if.…”

Functional MRI connectivity accurately distinguishes cases with psychotic disorders from healthy controls, based on cortical features associated with neurodevelopment

“…Several research studies have used machine learning (ML) algorithms to detect or diagnose psychotic disorders or schizophrenia based on MRI data (1)(2)(3)(4)(5)(6)(7)(8). However, in 64 casecontrol studies of schizophrenia, the ML performance accuracy ranged widely from 59% to 100% (1).…”