Machine learning fMRI classifier delineates subgroups of schizophrenia patients
Introduction
The search for neuroimaging based brain markers in psychiatry has faced two major methodological hurdles. The first is that a region-of-interest (ROI) approach requires a neuro-anatomical hypothesis, whose pathophysiological validity isn't always evident and which provides a limited scope of findings. The second is that in whole brain data-driven approach sensitivity is compromised by the large number of multiple comparisons (Hendler et al., 2014).
Further complicating this effort is the fact that current clinical diagnostic systems are category-based and not symptom-domain specific. Thus, schizophrenia and Obsessive–Compulsive Disorder (OCD) have traditionally been viewed as two distinct psychopathologies with differing underlying pathophysiologies (Kurokawa et al., 2000). Nevertheless, there is compelling evidence for the co-occurrence of obsessive–compulsive symptoms (OCS) in a substantial proportion of schizophrenia patients (roughly 25%) (Poyurovsky et al., 2012), termed “schizo-obsessive” when there is frank OCD. Schizo-obsessive patients appear to have a more deteriorative course, with poorer prognosis and treatment responses, as compared to their non-OCD schizophrenia counterparts (Poyurovsky et al., 2012). In terms of neuroimaging findings, while there is vast literature regarding the neural mechanisms (structural and functional) involved in schizophrenia and in OCD separately, there is scarce data regarding the functional neurobiological substrates involved in schizophrenia-OCD co-morbidity.
Hypofrontality, expressed as reduced neural activation in the dorsolateral prefrontal cortex (DLPFC) (Callicott, 2003), is the most replicated functional brain imaging finding in SCH patients. Therefore, in order to characterize some of the abnormalities in functional brain circuits in schizo-obsessive patients, we previously used fMRI during a working memory task, namely, the N-back task, in which schizophrenia (SCH) patients are known to display abnormal brain processing (Bleich-Cohen et al., 2014). Applying region-oriented analyses focusing on the right DLPFC and the right Caudate we demonstrated that schizo-obsessive patients exhibit remarkably similar brain activation patterns as their pure schizophrenia counterparts. Moreover, whole-brain analyses did not produce significant differences between the groups as well.
To further examine possible dissimilarities between these two schizophrenia groups in a data-driven manner, we chose to further apply a whole brain classification based on multi-voxel pattern analysis (MVPA) during the N-back task. MVPA approaches (Mur et al., 2009, Pereira et al., 2009, Orrù et al., 2012) utilize machine-learning algorithms that combine several different markers for classification and robust statistical analysis of the results. Current MVPA methods rely on either a small number of ROIs (Eger et al., 2007) or whole-brain activity (Mourão-Miranda et al., 2005, Fu et al., 2008). However, both approaches have major drawbacks: regional MVPA may achieve better classification accuracy than whole brain MVPA, but remains limited to hypothesis-driven assumptions, while whole brain MVPA may have difficulty achieving significant accuracy rates (Chu et al., 2012) and does not necessarily produce neuroscientifically interpretable results. To tackle this issue of neurological interpretability, some whole-brain studies use weights assigned by the classifier to the different markers in an attempt to produce an interpretation (Mourão-Miranda et al., 2005, Fu et al., 2008). However, this approach can lead to two equally important neural markers receiving substantially different weights, resulting in misleading interpretations (Lee et al., 2010). To overcome these limitations, we used a modified MVPA approach termed as Searchlight Based Feature Extraction (SBFE), which combines a data-driven ROI search with whole-brain classification (Jamshy et al., 2012). The goal of the present fMRI study was to evaluate whether SBFE can distinguish between these two schizophrenia subgroups and delineate the neurobiological substrates involved.
Section snippets
Subjects
Three groups were evaluated in this study. The study group included sixteen inpatients (10 men, 6 women; mean age 27, range 19–32 years) who met DSM-IV criteria for both schizophrenic disorder and OCD. The comparison group included 17 schizophrenia inpatients (11 men, 6 women; mean age 25, range19–31 years) matched for age, gender and education, who were hospitalized in the same department (Tirat HaCarmel Mental Health Center) during the same time period. The control group included twenty healthy
Results
Demographic and clinical characteristics: The two schizophrenia groups with and without OCD did not differ significantly in any of the demographic and clinical variables, with the exception of OCS-related variables. As expected, the schizo-obsessive patients scored significantly higher on the Y–BOCS total score (t = 9.72, df = 31, p < 0.001), and were more likely to be treated with anti-obsessive agents (sertraline 50–150 mg/day; paroxetine 20–40 mg/day, escitalopram 10–20 mg/day) compared to the
Discussion
The goal of this study was to explore the feasibility of using SBFE, a novel classification method that is based on a data driven ROI search coupled with whole-brain machine-learning, in classifying patients with complex psychopathology. The SBFE allowed us to delineate differences in brain activity between the two groups of schizophrenia patients: with and without OCD. While standard ROI driven analyses were not able to differentiate between the two patient groups (Bleich-Cohen et al., 2014),
Funding
This study was funded by institutional funding budgets. SJ was supported by the U.S. Department of Defense award number W81XWH-11-2-0008.
Contributors
Authors Bleich-Cohen, Poyurovsky, Weizman and Hendler designed the study and wrote the protocol. Author Bleich-Cohen participated in data collection. Authors Jamshy and Intrator invented the method. Author Bleich-Cohen, Jamshy and Sharon managed the literature searches and analyses. Authors Bleich-Cohen, and Jamshy undertook the statistical analysis, and authors Bleich-Cohen, Sharon, Jamshy and Hendler wrote the first draft of the manuscript. All authors contributed to and have approved the
Conflict of interest
All authors hereby declare having NO actual or potential conflict of interest including any financial, personal or other relationships with other people or organizations within three (3) years of beginning the work submitted that could inappropriately influence, or be perceived to influence, their work.
Acknowledgments
We thank Dr. D. Ben-Bashat for insights in physics, A. Solski for her constructive comments and all the subjects who volunteered to participate in the experiment.
References (27)
- et al.
Working memory dysfunction in schizophrenia patients with obsessive–compulsive symptoms: an fMRI study
Eur. Psychiatr.
(2014) An expanded role for functional neuroimaging in schizophrenia
Curr. Opin. Neurobiol.
(2003)- et al.
Does feature selection improve classification accuracy? Impact of sample size and feature selection on classification using anatomical magnetic resonance images
NeuroImage
(2012) - et al.
The effect of obsessive-compulsive symptomatology on executive functions in schizophrenia: a systematic review and meta-analysis
Psychiatry Res.
(2013) - et al.
Pattern classification of sad facial processing: toward the development of neurobiological markers in depression
Biol. Psychiatry
(2008) - et al.
From circuit activity to network connectivity and back: the case of obsessive–compulsive disorder
Biol. Psychiatry
(2014) - et al.
Ventricular enlargement in schizophrenia spectrum patients with prodromal symptoms of obsessive–compulsive disorder
Psychiatry Res.
(2000) - et al.
Classifying brain states and determining the discriminating activation patterns: support vector machine on functional MRI data
NeuroImage
(2005) - et al.
Using support vector machine to identify imaging biomarkers of neurological and psychiatric disease: a critical review
Neurosci. Biobehav. Rev.
(2012) - et al.
Machine learning classifiers and fMRI: a tutorial overview
NeuroImage
(2009)
Obsessive–compulsive symptoms in schizophrenia: implications for future psychiatric classifications
Compr. Psychiatry
Neural correlates of executive functions in healthy G2019S LRRK2 mutation carriers
Cortex
Scale for the Assessment of Negative Symptoms (SANS), Iowa City
Cited by (46)
Disruptions of salience network during uncertain anticipation of conflict control in anxiety
2023, Asian Journal of PsychiatryResting-state functional connectivity correlates of antipsychotic treatment in unmedicated schizophrenia
2023, Asian Journal of PsychiatryArtificial Intelligence in Clinical Psychology
2022, Comprehensive Clinical Psychology, Second EditionDeep learning model of fMRI connectivity predicts PTSD symptom trajectories in recent trauma survivors
2021, NeuroImageCitation Excerpt :For example, Galatzer-Levy et al. (2017) employed a support vector machine (SVM) for prediction of PTSD trajectories of trauma survivors and Rosellini et al. (2018) used an ensemble of multiple classifiers, including random forests (Breiman, 2001), SVM, and regularized regression, to develop a risk score for earthquake survivors. Recently, a growing number of studies have applied ML methodology on neuroimaging data to predict and characterize a variety of psychiatric disorders (Bleich-Cohen et al., 2014; Haller et al., 2014; Koutsouleris and Kambeitz, 2016; Liu et al., 2012; Mikolas et al., 2016; Mourao-Miranda et al., 2012; van der Ploeg et al., 2016; Rive et al., 2016; Zeng et al., 2012), including PTSD (Galatzer-Levy et al., 2017; Gong et al., 2014; Gradus et al., 2017; Jin et al., 2017; Karstoft et al., 2015; Liu et al., 2015; Omurca and Ekinci, 2015; Saxe et al., 2017; Schultebraucks and Galatzer-Levy, 2019). To date, however, no studies have applied a single ML method on both resting-state and task-based neuroimaging data to classify PTSD diagnosis simultaneously at three different time-points, using brain data only from the first time-point.
Assessing the information content of ERP signals in schizophrenia using multivariate decoding methods
2020, NeuroImage: ClinicalCitation Excerpt :We call this the information-based decoding approach, because it seeks to quantify the information present in the neural signals of each individual participant. A distinctly different decoding approach has been applied in several previous studies of schizophrenia (e.g., Bleich-Cohen et al., 2014; Du et al., 2012; Koch et al., 2015). In this approach, the decoder attempts to classify the group membership (PSZ vs. HCS) of each individual subject on the basis of the pattern of brain activity observed under a given set of conditions.
- 1
Shahar Jamshy and Haggai Sharon were equal contributors to this article.