William Pettersson-Yeo

Using Support Vector Machine to identify imaging biomarkers of neurological and psychiatric disease: A critical review

Standard univariate analysis of neuroimaging data has revealed a host of neuroanatomical and functional differences between healthy individuals and patients suffering a wide range of neurological and psychiatric disorders. Significant only at group level however these findings have had limited clinical translation, and recent attention has turned toward alternative forms of analysis, including Support-Vector-Machine (SVM). A type of machine learning, SVM allows categorisation of an individuals previously unseen data into a predefined group using a classification algorithm, developed on a training data set. In recent years, SVM has been successfully applied in the context of disease diagnosis, transition prediction and treatment prognosis, using both structural and functional neuroimaging data. Here we provide a brief overview of the method and review those studies that applied it to the investigation of Alzheimers disease, schizophrenia, major depression, bipolar disorder, presymptomatic Huntingtons disease, Parkinsons disease and autistic spectrum disorder. We conclude by discussing the main theoretical and practical challenges associated with the implementation of this method into the clinic and possible future directions.

The effects of psychotherapy on brain function: A systematic and critical review

Over the past two decades, the development of neuroimaging techniques has allowed the non-invasive investigation of neuroplastic changes associated with psychotherapeutic treatment. The aim of the present article is to present a systematic and critical review of longitudinal studies addressing the impact of psychotherapy on the brain published to date. After summarizing the results reported in the literature for each psychiatric disorder separately (i.e. obsessive-compulsive disorder, panic disorder, unipolar major depressive disorder, posttraumatic stress disorder, specific phobia, schizophrenia), we discuss the results focusing on three questions of interest: (i) whether neurobiological changes which follow psychotherapy occur in regions that showed significant neurofunctional alteration pre-treatment; (ii) whether these neurobiological changes are similar, or different, to those observed following pharmacological treatment; and (iii) whether neurobiological changes could be used as an objective means of monitoring the progress and outcome of psychotherapy. The evidence reviewed indicates that (i) depending on the disorder under investigation, psychotherapy results in either a normalisation of abnormal patterns of activity, the recruitment of additional areas which did not show altered activation prior to treatment, or a combination of the two; (ii) the effects of psychotherapy on brain function are comparable to those of medication for some but not all disorders; and (iii) there is preliminary evidence that neurobiological changes are associated with the progress and outcome of psychotherapy. It is hoped that a better understanding of the impact of psychotherapy on brain function will eventually inform the development of new biologically informed treatments and allow clinicians to make more effective treatment decisions.

Using genetic, cognitive and multi-modal neuroimaging data to identify ultra-high-risk and first-episode psychosis at the individual level

Background Group-level results suggest that relative to healthy controls (HCs), ultra-high-risk (UHR) and first-episode psychosis (FEP) subjects show alterations in neuroanatomy, neurofunction and cognition that may be mediated genetically. It is unclear, however, whether these groups can be differentiated at single-subject level, for instance using the machine learning analysis support vector machine (SVM). Here, we used a multimodal approach to examine the ability of structural magnetic resonance imaging (sMRI), functional MRI (fMRI), diffusion tensor neuroimaging (DTI), genetic and cognitive data to differentiate between UHR, FEP and HC subjects at the single-subject level using SVM. Method Three age- and gender-matched SVM paired comparison groups were created comprising 19, 19 and 15 subject pairs for FEP versus HC, UHR versus HC and FEP versus UHR, respectively. Genetic, sMRI, DTI, fMRI and cognitive data were obtained for each participant and the ability of each to discriminate subjects at the individual level in conjunction with SVM was tested. Results Successful classification accuracies (p < 0.05) comprised FEP versus HC (genotype, 67.86%; DTI, 65.79%; fMRI, 65.79% and 68.42%; cognitive data, 73.69%), UHR versus HC (sMRI, 68.42%; DTI, 65.79%), and FEP versus UHR (sMRI, 76.67%; fMRI, 73.33%; cognitive data, 66.67%). Conclusions The results suggest that FEP subjects are identifiable at the individual level using a range of biological and cognitive measures. Comparatively, only sMRI and DTI allowed discrimination of UHR from HC subjects. For the first time FEP and UHR subjects have been shown to be directly differentiable at the single-subject level using cognitive, sMRI and fMRI data. Preliminarily, the results support clinical development of SVM to help inform identification of FEP and UHR subjects, though future work is needed to provide enhanced levels of accuracy.

Auditory verbal hallucinations and brain dysconnectivity in the perisylvian language network: A multimodal investigation

Neuroimaging studies of schizophrenia have indicated that the development of auditory verbal hallucinations (AVHs) is associated with altered structural and functional connectivity within the perisylvian language network. However, these studies focussed mainly on either structural or functional alterations in patients with chronic schizophrenia. Therefore, they were unable to examine the relationship between the 2 types of measures and could not establish whether the observed alterations would be expressed in the early stage of the illness. We used diffusion tensor imaging and functional magnetic resonance imaging to examine white matter integrity and functional connectivity within the left perisylvian language network of 46 individuals with an at risk mental state for psychosis or a first episode of the illness, including 28 who had developed AVH group and 18 who had not (nonauditory verbal hallucination [nAVH] group), and 22 healthy controls. Inferences were made at P < .05 (corrected). The nAVH group relative to healthy controls showed a reduction of both white matter integrity and functional connectivity as well as a disruption of the normal structure−function relationship along the fronto-temporal pathway. For all measures, the AVH group showed intermediate values between healthy controls and the nAVH group. These findings seem to suggest that, in the early stage of the disorder, a significant impairment of fronto-temporal connectivity is evident in patients who do not experience AVHs. This is consistent with the hypothesis that, whilst mild disruption of connectivity might still enable the emergence of AVHs, more severe alterations may prevent the occurrence of the hallucinatory experience.

Multivariate pattern classification reveals differential brain activation during emotional processing in individuals with psychosis proneness

Among the general population, individuals with subthreshold psychotic-like experiences, or psychosis proneness (PP), can be psychometrically identified and are thought to have a 10-fold increased risk of psychosis. They also show impairments in measures of emotional functioning parallel to schizophrenia. Whilst previous studies have revealed altered brain activation in patients with schizophrenia during emotional processing, it is unclear whether these alterations are also expressed in individuals with high PP. Here we used Support Vector Machine (SVM) to perform multivariate pattern classification based on brain activation during emotional processing in 20 individuals with high PP and 20 comparison subjects (low PP). In addition, we performed a standard univariate analysis based on the General Linear Model (GLM) on the same data for comparison. The experimental task involved passively viewing negative and neutral pictures from the International Affective Picture System (IAPS). SVM allowed classification of the two groups with statistically significant accuracy (p=0.017) and identified group differences within an emotional circuitry including the amygdala, insula, anterior cingulate and medial prefrontal cortex. In contrast, the standard univariate analysis did not detect any significant between-group differences. Our results reveal a distributed and subtle set of alterations in brain function within the emotional circuitry of individuals with high PP, providing neurobiological support for the notion of dysfunctional emotional circuitry in this group. In addition, these alterations are best detected using a multivariate approach rather than standard univariate methods. Further application of this approach may aid in characterising people at clinical and genetic risk of developing psychosis.

Using Structural Neuroimaging to Make Quantitative Predictions of Symptom Progression in Individuals at Ultra-High Risk for Psychosis

Neuroimaging holds the promise that it may one day aid the clinical assessment of individual psychiatric patients. However, the vast majority of studies published so far have been based on average differences between groups, which do not permit accurate inferences at the level of the individual. We examined the potential of structural Magnetic Resonance Imaging (MRI) data for making accurate quantitative predictions about symptom progression in individuals at ultra-high risk for developing psychosis. Forty people at ultra-high risk for psychosis were scanned using structural MRI at first clinical presentation and assessed over a period of 2 years using the Positive and Negative Syndrome Scale. Using a multivariate machine learning method known as relevance vector regression (RVR), we examined the relationship between brain structure at first clinical presentation, characterized in terms of gray matter (GM) volume and cortical thickness (CT), and symptom progression at 2-year follow-up. The application of RVR to whole-brain CT MRI data allowed quantitative prediction of clinical scores with statistically significant accuracy (correlation = 0.34, p = 0.026; Mean Squared-Error = 249.63, p = 0.024). This prediction was informed by regions traditionally associated with schizophrenia, namely the right lateral and medial temporal cortex and the left insular cortex. In contrast, the application of RVR to GM volume did not allow prediction of symptom progression with statistically significant accuracy. These results provide proof-of-concept that it could be possible to use structural MRI to inform quantitative prediction of symptom progression in individuals at ultra-high risk of developing psychosis. This would enable clinicians to target those individuals at greatest need of preventative interventions thereby resulting in a more efficient use of health care resources.

Using structural neuroanatomy to identify trauma survivors with and without post-traumatic stress disorder at the individual level

Background At present there are no objective, biological markers that can be used to reliably identify individuals with post-traumatic stress disorder (PTSD). This study assessed the diagnostic potential of structural magnetic resonance imaging (sMRI) for identifying trauma-exposed individuals with and without PTSD. Method sMRI scans were acquired from 50 survivors of the Sichuan earthquake of 2008 who had developed PTSD, 50 survivors who had not developed PTSD and 40 healthy controls who had not been exposed to the earthquake. Support vector machine (SVM), a multivariate pattern recognition technique, was used to develop an algorithm that distinguished between the three groups at an individual level. The accuracy of the algorithm and its statistical significance were estimated using leave-one-out cross-validation and permutation testing. Results When survivors with PTSD were compared against healthy controls, both grey and white matter allowed discrimination with an accuracy of 91% (p < 0.001). When survivors without PTSD were compared against healthy controls, the two groups could be discriminated with accuracies of 76% (p < 0.001) and 85% (p < 0.001) based on grey and white matter, respectively. Finally, when survivors with and without PTSD were compared directly, grey matter allowed discrimination with an accuracy of 67% (p < 0.001); in contrast the two groups could not be distinguished based on white matter. Conclusions These results reveal patterns of neuroanatomical alterations that could be used to inform the identification of trauma survivors with and without PTSD at the individual level, and provide preliminary support to the development of SVM as a clinically useful diagnostic aid.

Quantitative Prediction of Individual Psychopathology in Trauma Survivors Using Resting-State fMRI

Neuroimaging techniques hold the promise that they may one day aid the clinical assessment of individual psychiatric patients. However, the vast majority of studies published so far have been based on average differences between groups. This study employed a multivariate approach to examine the potential of resting-state functional magnetic resonance imaging (MRI) data for making accurate predictions about psychopathology in survivors of the 2008 Sichuan earthquake at an individual level. Resting-state functional MRI data was acquired for 121 survivors of the 2008 Sichuan earthquake each of whom was assessed for symptoms of post-traumatic stress disorder (PTSD) using the 17-item PTSD Checklist (PCL). Using a multivariate analytical method known as relevance vector regression (RVR), we examined the relationship between resting-state functional MRI data and symptom scores. We found that the use of RVR allowed quantitative prediction of clinical scores with statistically significant accuracy (correlation=0.32, P=0.006; mean squared error=176.88, P=0.001). Accurate prediction was based on functional activation in a number of prefrontal, parietal, and occipital regions. This is the first evidence that neuroimaging techniques may inform the clinical assessment of trauma-exposed individuals by providing an accurate and objective quantitative estimation of psychopathology. Furthermore, the significant contribution of parietal and occipital regions to such estimation challenges the traditional view of PTSD as a disorder specific to the fronto-limbic network.

An empirical comparison of different approaches for combining multimodal neuroimaging data with support vector machine.

In the pursuit of clinical utility, neuroimaging researchers of psychiatric and neurological illness are increasingly using analyses, such as support vector machine, that allow inference at the single-subject level. Recent studies employing single-modality data, however, suggest that classification accuracies must be improved for such utility to be realized. One possible solution is to integrate different data types to provide a single combined output classification; either by generating a single decision function based on an integrated kernel matrix, or, by creating an ensemble of multiple single modality classifiers and integrating their predictions. Here, we describe four integrative approaches: (1) an un-weighted sum of kernels, (2) multi-kernel learning, (3) prediction averaging, and (4) majority voting, and compare their ability to enhance classification accuracy relative to the best single-modality classification accuracy. We achieve this by integrating structural, functional, and diffusion tensor magnetic resonance imaging data, in order to compare ultra-high risk (n = 19), first episode psychosis (n = 19) and healthy control subjects (n = 23). Our results show that (i) whilst integration can enhance classification accuracy by up to 13%, the frequency of such instances may be limited, (ii) where classification can be enhanced, simple methods may yield greater increases relative to more computationally complex alternatives, and, (iii) the potential for classification enhancement is highly influenced by the specific diagnostic comparison under consideration. In conclusion, our findings suggest that for moderately sized clinical neuroimaging datasets, combining different imaging modalities in a data-driven manner is no “magic bullet” for increasing classification accuracy. However, it remains possible that this conclusion is dependent on the use of neuroimaging modalities that had little, or no, complementary information to offer one another, and that the integration of more diverse types of data would have produced greater classification enhancement. We suggest that future studies ideally examine a greater variety of data types (e.g., genetic, cognitive, and neuroimaging) in order to identify the data types and combinations optimally suited to the classification of early stage psychosis.

Pattern classification of brain activation during emotional processing in subclinical depression: psychosis proneness as potential confounding factor.

We used Support Vector Machine (SVM) to perform multivariate pattern classification based on brain activation during emotional processing in healthy participants with subclinical depressive symptoms. Six-hundred undergraduate students completed the Beck Depression Inventory II (BDI-II). Two groups were subsequently formed: (i) subclinical (mild) mood disturbance (n = 17) and (ii) no mood disturbance (n = 17). Participants also completed a self-report questionnaire on subclinical psychotic symptoms, the Community Assessment of Psychic Experiences Questionnaire (CAPE) positive subscale. The functional magnetic resonance imaging (fMRI) paradigm entailed passive viewing of negative emotional and neutral scenes. The pattern of brain activity during emotional processing allowed correct group classification with an overall accuracy of 77% (p = 0.002), within a network of regions including the amygdala, insula, anterior cingulate cortex and medial prefrontal cortex. However, further analysis suggested that the classification accuracy could also be explained by subclinical psychotic symptom scores (correlation with SVM weights r = 0.459, p = 0.006). Psychosis proneness may thus be a confounding factor for neuroimaging studies in subclinical depression.