Diagnostic Classification of Schizophrenia Patients on the Basis of Regional Reward-Related fMRI Signal Patterns

Stefan P. Koch; Claudia Hägele; John-Dylan Haynes; Andreas Heinz; Florian Schlagenhauf; Philipp Sterzer

doi:10.1371/journal.pone.0119089

Abstract

Functional neuroimaging has provided evidence for altered function of mesolimbic circuits implicated in reward processing, first and foremost the ventral striatum, in patients with schizophrenia. While such findings based on significant group differences in brain activations can provide important insights into the pathomechanisms of mental disorders, the use of neuroimaging results from standard univariate statistical analysis for individual diagnosis has proven difficult. In this proof of concept study, we tested whether the predictive accuracy for the diagnostic classification of schizophrenia patients vs. healthy controls could be improved using multivariate pattern analysis (MVPA) of regional functional magnetic resonance imaging (fMRI) activation patterns for the anticipation of monetary reward. With a searchlight MVPA approach using support vector machine classification, we found that the diagnostic category could be predicted from local activation patterns in frontal, temporal, occipital and midbrain regions, with a maximal cluster peak classification accuracy of 93% for the right pallidum. Region-of-interest based MVPA for the ventral striatum achieved a maximal cluster peak accuracy of 88%, whereas the classification accuracy on the basis of standard univariate analysis reached only 75%. Moreover, using support vector regression we could additionally predict the severity of negative symptoms from ventral striatal activation patterns. These results show that MVPA can be used to substantially increase the accuracy of diagnostic classification on the basis of task-related fMRI signal patterns in a regionally specific way.

Citation: Koch SP, Hägele C, Haynes J-D, Heinz A, Schlagenhauf F, Sterzer P (2015) Diagnostic Classification of Schizophrenia Patients on the Basis of Regional Reward-Related fMRI Signal Patterns. PLoS ONE 10(3): e0119089. https://doi.org/10.1371/journal.pone.0119089

Academic Editor: Adam J. Schwarz, Indiana University, UNITED STATES

Received: August 4, 2014; Accepted: December 7, 2014; Published: March 23, 2015

Copyright: © 2015 Koch et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited

Data Availability: The authors confirm that, for approved reasons, some access restrictions apply to the data underlying the findings. Due to ethical restrictions, data cannot be made publicly available. However, data will be made available to all interested researchers after receiving permission from the data protection officials of the Charité Universitätsmedizin Berlin committee. Requests for data access will be received by Stefan Koch (Stefan.Koch@charite.de).

Funding: This study was supported by the German Research Foundation (Deutsche Forschungsgemeinschaft, STE 1430/6-1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Alterations in the neural processing of reward are a key finding in schizophrenia and have been proposed to be linked to dysfunctional dopaminergic neurotransmission in the mesolimbic reward system, first and foremost the central and ventral striatum [1–5]. Over the past decade, a number of functional magnetic resonance imaging (fMRI) studies have provided consistent evidence for reduced functional activation in the ventral striatum in response to reward-predicting stimuli in schizophrenia patients compared to controls [6–9]. This reduction in ventral striatal activation has been linked predominantly to the negative symptoms of schizophrenia [7,10]. In addition, reduced activation during reward processing in schizophrenia patients has also been observed in a number of other brain regions such as the amygdala, hippocampus, nucleus accumbens, prefrontal and insular cortex and parahippocampal gyrus [7,11–14]. While such findings based on significant group differences in fMRI signal have undoubtedly provided important insights into the pathomechanisms of schizophrenia, the use of such neuroimaging results from standard univariate statistical analysis for individual diagnosis has proven difficult, mostly because of large inter-individual variance in regional fMRI activations. An approach that can be used to overcome these difficulties is the use of multivariate pattern analysis (MVPA), which can dramatically increase the sensitivity of human brain imaging by accumulating information across multiple voxels of MRI signal, i.e., by taking into account the information contained in a distributed spatial pattern of brain activity rather than a single voxel or location [15,16]. A commonly applied implementation of MVPA is the use of a classification algorithm, e.g., support vector machine classification [17,18], that is trained to distinguish between two classes of data using pattern-based information. The accuracy of the trained classifier is then probed in independent test data. Such techniques have proven extremely useful not only for the decoding of brain states from patterns of brain imaging data on the individual-subject level but also for between-subject classification of brain imaging data in a number of psychiatric and neurological diseases (for reviews, see [19–22]). In recent years schizophrenia has been studied with MVPA using various neuroimaging variables such as resting state, diffusion tensor imaging and structural morphometry [23–28]. However, few studies have used MVPA to differentiate between schizophrenia patients and healthy controls on the basis of task-related fMRI signal patterns [29,30].

Here we asked whether MVPA could be used for the diagnostic classification of patients with schizophrenia vs. healthy controls on the basis of reward-related fMRI signal patterns obtained in a previous study [31]. In contrast to earlier studies that used MVPA for diagnostic classification [29,30], we were particularly interested in the regional specificity of MVPA-based classification, especially with respect to the above-mentioned brain regions that were implicated in altered reward processing in schizophrenia patients by earlier studies. Rather than using whole-brain activation patterns for classification, we employed a ‘searchlight’ approach [32,33] that can be used to assess classification accuracy for regional fMRI signal patterns across a whole fMRI scan volume [34,35]. Under this approach the searchlight is moved through the entire brain, and at each location, combines local information of voxels within a spherical volume across subjects. As the combined information of voxels within the sphere is projected to the center of the sphere at each location this approach eventually provides a whole-brain map of local information. Compared to other whole-brain approaches, searchlight MVPA offers some advantages such as the simplicity of implementation and the intuitive interpretation of the resulting maps similar to mass-univariate statistics. Moreover, searchlight MVPA circumvents the necessity for feature selection, which is a challenge for whole-brain MVPA due to high dimensionality. Finally, the searchlight approach preserves the regional specificity, thus allowing for a comparison of multivariate results with those obtained from mass-univariate methods. Because functional imaging of schizophrenia patients during a MID task has to date been exclusively analysed with mass-univariate statistics, we reasoned that the latter aspect is of particular relevance to benchmark MVPA against the standard univariate approach. We hypothesized that regionally specific classification accuracy would be highest for those brain regions whose reward-related activation has been previously shown to be altered in patients with schizophrenia, especially the ventral striatum [6–9,11–14,31,36]. In addition, we also asked whether MVPA of regional reward-related fMRI signal patterns could be used to predict the severity of clinical symptoms.

Materials and Methods

Participants

The study was approved by the local ethics committee, Charité–Universitätsmedizin, Berlin, Germany. Written informed consent was obtained from all participants. A total of 98 participants were included in the study: 54 healthy controls and 44 patients diagnosed with schizophrenia. Patients fulfilling DSM-IV and ICD-10 criteria for schizophrenia without having other psychiatric axis I disorders, current drug abuse or past history of drug dependence (SCID interview; [37]) were recruited at the Charité University Medical Centre's Department of Psychiatry and Psychotherapy. Psychopathological symptoms were assessed with the Positive and Negative Syndrome Scale (PANSS; [38]). Healthy participants in the control group showed no psychiatric axis I or II disorders (SCID) or any family history of psychiatric disorders and no substance abuse or dependence within the previous 6 months. Equal sample sizes in both groups were obtained by excluding datasets from the group with larger samples (healthy controls) based on a matching of age and gender criteria. Thus, the groups contained 44 schizophrenia patients (mean age: 34.2±9.8, range 19–57) and 44 healthy controls (mean age: 37.1±10.9, range 18–59), respectively. A two-sample t-test revealed no age differences between the two groups (t = 1.32, p = 0.19). There were 35 male controls (M:F ratio = 3.88) and 27 male patients (M:F ratio = 1.58). A Pearsons’s chi-square test revealed no differences between the two groups with respect to gender (chi = 3.49, p = 0.06). The medication status of the patients with schizophrenia consisted of 7 patients taking atypical antipsychotics, 21 conventional antipsychotics, and 16 not receiving any medication. All participants were right-handed, as assessed with the Edinburgh Handedness Inventory [39]. For a detailed description of the sample see Table 1.

Download:

Table 1. Demographic parameters and reaction times.

https://doi.org/10.1371/journal.pone.0119089.t001

Monetary Incentive Delay Task and Data Acquisition

Participants performed a monetary incentive delay task (MID task; [40,41]) during fMRI. The task invokes anticipation of reward and punishment. Depending on the performance in a simple reaction time task (button press) to a visual target a potential monetary gain, loss or no consequence is depicted at the end of the trial. Prior to fMRI acquisition, participants received information about the meaning of the cues. Participants were informed that they receive the earned money after completion of the scanning session. During the acquisition of the anatomical scan, participants practiced the task (without monetary payment). Each trial started with the presentation of a cue indicating whether subjects could win money, avoid losing money or obtain no money (neutral cue). The different magnitudes of the incentive (0.10 €; 0.60 € or 3 €) were indicated by the number of horizontal lines presented inside the cue image. Between cues and target, a variable delay was inserted. The application of an adaptive algorithm for target duration enabled subjects to succeed in about 67% of the trials. Successful trials were defined as button presses within the time frame of the target presentation. To control for neuronal artifacts due to motor response, participants were instructed to press the button as fast as possible regardless of the cue. A feedback display was presented after each trial to indicate the trial-related success. MR acquisition comprised anatomical and functional scans. The functional scans were splitted into two runs with altogether 144 trials consisting of 54 gain, 54 loss, and 36 neutral trials, which were presented in a random sequence (trial length 8 s, jittered mean intertrial interval 4 s; for a detailed description of the task, see Hägele and colleagues [31]).

fMRI Data Acquisition

Images were acquired with a 1.5 T Magnetom VISION (Siemens) using a standard circularly polarized head coil (CP-Headcoil). Gradient-echo echo-planar imaging (GE-EPI, TR = 1.9 s, TE = 40 ms, flip angle = 90°, matrix = 64 × 64, voxel size = 4 mm × 4 mm × 3.3 mm) was used to produce eighteen slices approximately parallel to the bicommissural plane (ac-pc plane), covering the inferior part of the frontal lobe (superior border above the caudate nucleus), the entire temporal lobe, and large parts of the occipital region. fMRI volume acquisitions were time-locked to the offset of each cue and were thus acquired during anticipatory delay periods. Six fMRI volumes were acquired per trial, resulting in 450 volumes per run. High resolution anatomical images were acquired using a 3D MPRAGE sequence (Magnetization Prepared Rapid Gradient Echo, TR = 9.7 ms; TE = 4 ms; flip angle 12°; matrix = 256 × 256, voxel size 1 mm × 1 mm × 1 mm). A vacuum pad served to minimize head movements.

fMRI Data Analysis

SPM8 (http://www.fil.ion.ucl.ac.uk/spm) was used for fMRI data analysis. To avoid non-steady state effects from T1 saturation the first three volumes of each functional time series were discarded. Volumes were realigned to the first volume to correct for between-scan movements and to remove signals correlated with head motion using sinc interpolation. Motion correction confirmed that no subjects showed more than 4 mm head movement during the run and less than 1 mm translation and 1° rotation in any dimension from one volume acquisition to the next. The anatomical image was coregistered to the mean functional image. The functional data set was coregistered with the anatomical volume based on the mean functional volume of the first run and spatially normalized to the standard MNI template using the algorithm implemented in SPM8 (12-parameter affine transformation followed by a non-linear warping using 7x8x7 harmonic basis functions to compensate anatomical distortions). Subsequently, the data were resampled to a resolution of 3 × 3 × 3 mm voxel size and smoothed using a 8 mm full-width half-maximum (FWHM) isotropic kernel. Functional MRI data were analyzed using the general linear model (GLM; [42]). Data analysis was performed by modelling the onsets of the three different conditions (cues for gain, loss and no monetary) as explanatory variables convolved with hemodynamic response function (gamma-variate function; [43]). Changes in the blood-oxygen level-dependent (BOLD) response were assessed using linear combinations of the estimated GLM parameters (betas) and are contained in the individual contrast images for the seven cue conditions, the target and the five feedback conditions (successful gain, non-successful gain, successful loss avoidance, non-successful loss-avoidance, neutral condition). Movement parameters derived from image realignment were included as additional regressors of no interest. For the anticipation phase the contrast image ‘gain vs. no outcome’ was computed combining the three different values for gain. Knutson and colleagues [41] suggested that neuronal activation during reward anticipation is stronger than during loss anticipation, and indeed, several studies observed small or non-significant differences in loss anticipation between healthy participants and patients with various disorders, and furthermore during the feedback phase [31,44,45]. To use a robust and strong contrast for the MVPA approach we therefore focused on reward anticipation in both fMRI and MVPA analyses. For standard univariate analysis, the individual contrast images entered a second-level random effects analysis to investigate the between-group differences with respect to the gain vs. no outcome contrast using a two-sample t-test (FDR corrected at q = 0.05, cluster level 30 voxels).

Multivariate Pattern Analysis

MVPA was performed to investigate whether the clinical diagnosis can be determined on the basis of task-related regional activation patterns of the gain vs. no outcome contrast. Support vector machine classification (SVM; [46,47]) has been shown to be a powerful tool for statistical pattern analysis and proven to be a versatile and robust approach for analyzing functional neuroimaging data [21,48]. SVM is a binary classifier that finds the maximum margin separating hyperplane. Based on the training data the goal of SVM is to produce a model which predicts the target value y_i (label, diagnostic status) of the test instance i given only the test data attributes x_i (features, fMRI voxel).

Given a training set , where x_i ∈ R^d and y_i ∈ {+1,−1} the standard form of the SVM objective with parameter C to scale the loss is where w is the normal to the hyperplane and l_i(z) denotes the loss function. The standard SVM objective is equivalent (proportional) to the following SVM objective,

Using the hinge-loss function the following optimization problem has to be solved while an ε-accurate solution is quested by the applied optimization method. To solve the optimization problem in the primal space we used the Pegasos algorithm [49], a stochastic sub-gradient descent method (Matlab code is provided by Sebastien Paris, http://www.mathworks.com/matlabcentral/fileexchange/33621-fast-linear-binary-svm-classifier). While the traditional subgradient method uses the entire training set at each iteration step, Pegasos chooses randomly a single training example to estimate a sub-gradient of the objective, and a step with pre-determined step-size is taken in the opposite direction of the gradient. One of the advantages of the Pegasos algorithm lies in the fast final convergence of solving the optimization problem and the substitution of the cost parameter. The linear support vector classification (SVC) with Pegasos algorithm for solving the optimization problem was embedded in a searchlight approach to identify local brain patterns with informative signatures with respect to the clinical status [34,35]. For each voxel location within the scan volume, the data of the voxels within a searchlight sphere of six voxels in diameter were fed into the classifier with a leave-one-out cross-validation (LOOCV) scheme. For each cross-validation iteration data were partitioned into training and test sets, by excluding one different participant (n_test = 1), and the SVC classifier was trained on the data of the remaining participants (n_train = N-1, where N = 88). The trained classifier was then used to predict the label of the unseen test participant based on his/her data alone. This process was repeated leaving each participant out once to finally obtain an accuracy measure (percentage of correctly predicted labels) based on the number of correctly classified test samples. Note that the training set during each training iteration consisted of unequal sample sizes (43 and 44 datasets for each group, respectively). Although unequal group sizes during training can introduce a prediction bias towards the majority class, we decided to apply this procedure for the following reasons: (i) The impact of the bias can be regarded as neglible because the imbalance is small given the relatively large sample size; (ii) the direction of the bias is balanced across LOOCV because both groups contained equal sample sizes; (iii) to equalize sample sizes during training another participant from the other group has to be chosen based on an arbitrary selection scheme. Mapping this accuracy value into the center of the searchlight sphere and performing the LOOCV-searchlight procedure for all locations results in a brain map of decoding accuracies. Statistical significance of the overall classification accuracy was determined by permutation testing to generate empirical chance distributions of the accuracy measure for all decoded locations [50,51]. For this, the LOOCV procedure was repeated 10,000 times with a different random permutation of the training group labels. To maintain the spatial coherence, the permutation of the label was kept constant within each permutation step while in turn each permutation step comprised an entire LOOCV-searchlight decoding of all voxels within the scan volume. For each voxel the probability to receive the accuracy value for the actual labels by chance was estimated using the permutation-based histograms of chance accuracy values. In order to confine alpha inflation due to multiple comparisons we used the false discovery rate (FDR; [52]). FDR controls the average fraction of false positives (at q = 0.05) out of the set of all positive test results. As for the univariate case a minimum cluster size of 30 voxels was considered.

To directly compare the classification accuracies of the SVM analysis with its complement from the univariate approach, we generated accuracy maps for the univariate analysis by computing the Receiver-Operating Characteristic curve (ROC; [53]) for each brain voxel. The ROC curve displays the sensitivity versus 1-specificity at various discrimination thresholds and is used to determine the threshold with the best classification percentage over the available training set. In other words, for each voxel we estimated the optimal trade-off between misses and false positives which in turn represents the maximum reachable accuracy for correctly classifying the two groups when considering the univariate analysis. Accuracies were computed for the ventral striatum as our primary regions of interest. The ventral striatal region of interest was specified from the publication-based probabilistic Montreal Neurological Institute (MNI) atlas used as binary mask at the threshold of 0.75 probability (see http://hendrix.imm.dtu.dk/services/jerne/ninf/voi/index-alphabetic.html).

For the schizophrenia group, a linear support vector regression (SVR; [54]) was performed to test whether the severity of negative symptoms as measured with PANSS scale could be predicted from fMRI response patterns of the contrast 'monetary gain vs. no outcome' for voxels within the ventral striatal mask (for a review, see [55]). The cost parameter, which determines the influence of the misclassification on the objective function, was fixed at the default setting of C = 1. On the basis of previous findings [7] we hypothesized that voxels within the ventral striatum carry information especially on the severity of negative symptoms. For SVR we used a similar LOOCV-searchlight procedure as for the support vector classification: For each voxel of the ventral striatum, the data of the voxels within a searchlight sphere of four voxels in diameter were fed into the SVR with a leave-one-out cross-validation. Compared to SVC the searchlight sphere for SVR was smaller in diameter because the ventral striatal mask contained only 90 and 92 voxels for left and right ventral striatum, respectively. In each cross-validation step, the data were partitioned into training and test sets, by excluding a different schizophrenia patient, and the SVR was trained on the data of the other patients. The resulting regression model was then used to predict the PANSS score of the untrained patient based on his/her functional data alone. Conducting this LOOCV scheme for each patient yielded a vector with individual PANSS score estimates. Spearman correlation was used to examine the relation between PANSS score estimates and actual PANSS scores. The correlation coefficient was than mapped into the center of the searchlight sphere and the entire LOOCV-searchlight-SVR procedure was performed for all locations within the ventral striatal mask. Permutation tests were performed to obtain unbiased empirical chance distributions for the relationship of true and predicted PANSS scores within the ventral striatum. For this, the LOOCV-searchlight-SVR procedure described above was repeated 10,000 times, each time with a different random assignment of the true PANSS scores across patients in the training set ('null SVR models'). Analogous to the SVC approach, the permutation-based histogram of each voxel was used to estimate the probability to obtain the observed relationship between predicted and true PANSS scores by chance. FDR correction [52] was finally applied to curtail alpha inflation.

Results

We examined the performance of the searchlight SVM classification on fMRI activation patterns in response to reward-indicating stimuli with the aim to decode the clinical status (schizophrenia patients vs. healthy control). For completeness, we also report the univariate group differences between healthy controls and schizophrenia patients for the contrast reward anticipation versus no gain.

Behavioral Data

Mean reaction times (RT) are shown in Table 1. A mixed two-way ANOVA with the within-subject factor reward cue (neutral, gain, loss) and the between-subject factor group (control, patients) on RT showed a main effect of reward cue (F(1.26,105.63) = 48.32, p < 0.001) as reported previously [31]. There was also a main effect of group (F(1,84) = 9.89, p = 0.002), indicating shorter RT in the healthy control group (RT averaged across conditions 383 ms (STD 163 ms) in schizophrenia patients and 292 ms (STD 91 ms) in controls). There was a significant group × reward-cue interaction (F(1.26,105.63) = 4.25, p = 0.033). To further analyse the significant interaction separate one-way ANOVAS for the healthy control group and schizophrenia patients were conducted. For the healthy control group a significant main effect for the factor reward cue (F(1.19,48.58) = 34.08, p < 0.001) was revealed. Pairwise comparisons (Bonferroni corrected) showed that healthy controls responded significantly faster during gain vs. neutral trials (p < 0.001) and loss vs. neutral trials (p < 0.001). There was no significant difference between gain and loss trials (p = 1.0) The ANOVA for schizophrenia patients also revealed a significant main effect for the factor reward cue (F(1.36,58.28) = 14.74, p < 0.001). Pairwise comparisons showed that schizophrenia patients responded significantly faster during gain vs. neutral trials (p < 0.001) and loss vs. neutral trials (p = 0.001). There was no significant difference between gain and loss trials (p = 0.61). Thus the effect of reward and loss anticipation on RT revealed similar result patterns between groups, indicating that participants in both groups understood the paradigm and were engaged in the task to a similar extend. Please also note that the MID task was programmed to adjust to individual reaction times, so equal percentages of gains and losses for all participants were ensured [40,45].

Univariate group differences in reward anticipation

As previously reported [31], BOLD activation during anticipation of monetary gain versus no gain was significantly reduced in schizophrenia patients compared to healthy subjects in the bilateral ventral striatum. As can be seen in Fig. 1, schizophrenia patients also showed reduced responses in the putamen, parahippocampal gyrus, cingulate gyrus, caudate, insula, amygdala and thalamus, as well as multiple frontal, temporal and occipital regions (Table 2).

Download:

Fig 1. Group differences in reward anticipation.

Results for the contrast reward anticipation versus no outcome for healthy controls > schizophrenia patients (thresholded at p < 0.05, FDR-corrected for multiple comparisons, cluster level 30 voxels). Healthy controls displayed significant larger activations in the ventral striatum, hippocampus, caudate body and substantia nigra during reward-indicating versus neutral cues.

https://doi.org/10.1371/journal.pone.0119089.g001

Download:

Table 2. Activations for the contrast reward anticipation versus no outcome for healthy controls > schizophrenia patients.

https://doi.org/10.1371/journal.pone.0119089.t002

MVPA classification of schizophrenia patients and healthy controls

Searchlight MVPA identified a distributed cortical network of frontal, temporal, occipital and midbrain regions with high classification accuracies. Across all voxels the mean of the lowest, medium and highest accuracies scores by chance, derived from the permutation-based random distributions, were 29.1%, 48.4%, and 66.5%, respectively. By contrast, maximal accuracy for the classification of patients vs. controls was obtained in the right pallidum ([MNI: 24, −6, −6], accuracy = 93%), bilateral putamen (left: [MNI: −24, 6, −15], accuracy = 90%; right: [MNI: 24, 3, −9], accuracy = 90%), right inferior frontal gyrus ([MNI: 18, 15, −18], accuracy = 86%), right nucleus accumbens ([MNI: 12, 12, −12], accuracy = 85%), right amygdala ([MNI: 27, −3, −15], accuracy = 83%), bilateral insula (left: [MNI: −27, 15, −15], accuracy = 84%; right: [MNI: 39, −18, −3], accuracy = 82%), bilateral thalamus (left: [MNI: −18, −24, 0], accuracy = 83%; right: [MNI: 6, −15, −3], accuracy = 82%), and left inferior temporal gyrus ([MNI: 19, −54, −75, −6], accuracy = 82%; p < 10⁻⁵ for all accuracies). Interestingly, for the twelve regions with the best accuracies (mean accuracy: 85%, range: 82–93%) the sensitivity (mean: 91%, range: 73–100%) was generally larger compared to the specificity score (mean: 79%, range: 64–93%). See Table 3 and Fig. 2 for further details.

Download:

Fig 2. Brain areas that discriminated between schizophrenia patients and healthy control during reward anticipation using a multivariate classification approach.

Accuracy scores (percent correct classification) from SVM searchlight decoding were colour-coded to display the classification performance. Letters x, y, z denote the axial, coronal and sagittal planes, respectively. The maps are thresholded at a significance level of p<0.05, FDR-corrected (cluster level 30 voxels).

https://doi.org/10.1371/journal.pone.0119089.g002

Download:

Table 3. Multivariate classification of schizophrenia patients and healthy controls.

https://doi.org/10.1371/journal.pone.0119089.t003

Comparison of univariate and multivariate classification of patients and healthy controls for the ventral striatum

For our primary region of interest, the ventral striatum, we compared the accuracies of the univariate approach derived from Receiver-Operating Characteristic curve analysis (ROC; [53]) with those of MVPA using the SVM classifier. For comparability reasons, FDR correction was applied for both approaches. As expected, larger maximum accuracy scores were observed in the multivariate case for both the left and the right mask of the ventral striatum (left: [MNI: −18, 2, −11], accuracy = 87%; right: [MNI: 18, 5, −11], accuracy = 88%) when compared to the univariate approach (left: [MNI: −18, 8, −5], accuracy = 75%; right: [MNI: 15, 11, −8], accuracy = 73%). Within the ventral striatum sensitivity and specificity of the peak voxel for the two approaches were compared using McNemar's test. The marginal proportions were significantly different from each other (X² = 5.88, p = 0.015), indicating that MVPA provides a better prediction performance compared to the univariate model. With the multivariate approach a substantially greater number of voxels survived FDR correction within the mask of the ventral striatum compared to the univariate analysis (percent significant voxels with the ventral striatum mask: multivariate: L = 91%, R = 77%; univariate: L = 56%, R = 65%). See Table 4 and Fig. 3 for further details.

Download:

Fig 3. Classification performance comparison.

The top and bottom panels depict percent correct classification rates (accuracies) obtained from the multivariate (linear SVM) and univariate (ROC) classification approach, respectively. The white line denotes the mask boundary of the ventral striatum. For illustrative reasons the accuracies where thresholded at 70% thus fewer significant voxels are displayed in the figure compared to the actually survived number of voxels after FDR correction.

https://doi.org/10.1371/journal.pone.0119089.g003

Download:

Table 4. Comparison of univariate and multivariate classification performance for the ventral striatum.

https://doi.org/10.1371/journal.pone.0119089.t004

Multivariate prediction of the PANSS negative scale

Leave-one-out SVR (LIBSVM; [54]) was used to investigate the relationship of gain anticipation and the PANSS negative scale in schizophrenia patients within the ventral striatum. The ventral striatum was chosen because previous univariate analyses revealed an inverse relationship between ventral striatal activation and the severity of negative symptoms [7,10]. The severity of negative symptoms as measured with PANSS could be predicted from the left ventral striatal activation pattern in response to monetary gain vs. no outcome: Within the left ventral striatum support vector regression revealed the strongest relationship with the PANSS negative scale for the searchlight sphere with the center coordinate at MNI = [−12, 11, 1] and a Spearman correlation coefficient of R = 0.72 (p = 5.14e-5; see Fig. 4). For the same center coordinate, predictions by chance revealed a minimum, mean and maximum correlation coefficient of R = −0.83, R = −0.154 and R = 0.66, respectively. Across all voxels within the ventral striatal mask the mean of the lowest, medium and highest coefficients by chance were −0.85, −0.15, and 0.62, respectively. Note that negative coefficients represent no predictive information regarding the function estimation between PANSS scores and predicted PANSS values from multiple voxel data using SVR.

Download:

Fig 4. Support vector regression (SVR) with PANSS negative scale for the schizophrenia group.

For the fMRI contrast monetary gain vs. no outcome there was a tight relationship between PANSS negative symptom scores and those predicted with SVR from activation patterns left ventral striatum within the clinical group. The right panel shows the correlation for the voxel (MNI: −12, 11, 1) within the left ventral striatum with the strongest relationship (R = 0.72) between actual and predicted PANSS negative scores. Each dot represents a schizophrenia patient.

https://doi.org/10.1371/journal.pone.0119089.g004

Altogether, the tight relationship between the actual PANSS negative scores and those predicted by SVR indicates that voxels within the ventral striatum carry information with respect to the severity of negative symptoms in schizophrenia patients.

Discussion

In this study we used searchlight MVPA of regional fMRI activation patterns in response to anticipation of monetary reward for diagnostic classification of schizophrenia patients vs. healthy control participants. Regional activation patterns with the highest accuracy scores for the discrimination between schizophrenia patients and controls were observed in subcortical regions such as the pallidum, putamen, nucleus accumbens, as well as in the inferior frontal gyrus and insular cortex. In line with previous reports, the univariate comparison of the groups revealed a reduced BOLD activation to reward anticipation in the ventral striatum [6–9,36] and a distributed network of regions in schizophrenia patients compared to healthy controls [7,11–14]. For the left and right ventral striatum the multivariate classification revealed one of the highest class prediction rates, which where found to be larger compared to those computed on the basis of ROC-curves from univariate analysis. The lower accuracies of the univariate approach can be attributed to the fact that the mass-univariate analysis treats each voxel independently and therefore does not take into account information that reflects task-related group differences in neural activity that are spatially distributed. Conversely, the searchlight SVM incorporates redundant but also additive information from spatially correlated neighbouring voxels, thereby improving class prediction [33].

In line with previous studies, both the univariate comparison and the multivariate classification of the two groups show that the ventral striatum, a key region in reward processing and encoding of the incentive salience of rewarding stimuli [56–59], exhibits differential activation for schizophrenia patients compared to healthy controls. Previous studies have found an inverse relationship between the severity of negative symptoms and the magnitude of BOLD activation in the ventral striatum during reward anticipation [7,11,60,61,44]. Our current results corroborate the notion that the negative symptoms of schizophrenia are related to ventral striatal activation. They go beyond these previous reports by now showing a significant correlation of actual PANSS negative ratings with those predicted by support vector regression, thus indicating that not only the magnitude of the ventral striatal responses but also the activation pattern in this region is informative with regard to psychopathology. Our study thus provides additional evidence that reward-associated neural activity in the left ventral striatum is coupled to the severity of negative symptoms in schizophrenia patients and supports the hypothesis that reduced motivation or anhedonia is linked with ventral striatal dysfunction [62,63].

Importantly, we used a multivariate searchlight approach [32,33] to investigate which brain regions contain activation patterns with valuable diagnostic information for the discrimination of schizophrenia patients and healthy controls. This approach successfully exposed distinct brain regions that have been observed in previous univariate studies [6–9,11–14,31,36]. Our results therefore confirm the significance of these regions in the pathophysiology of schizophrenia and highlight the usefulness of MVPA searchlight analysis for the identification of regional activation patterns that can help the diagnostic classification of clinical groups.

Note that in general specificity was somewhat smaller than sensitivity. We attribute this to the differences in the variance between groups. While both groups group showed the same frequency of voxels with violations against the normal distribution (Shapiro-Wilk test: 8.7% and 7.5% of the voxels, respectively), the variance within the schizophrenia group was smaller compared to healthy controls: Although Levene's Test for Homogeneity of Variances rejected on average only 7.1% of the voxels, numerically, 80.2% of these voxels showed a larger variance within healthy controls compared to schizophrenia patients. We conducted SVC simulations with normally distributed random data and systematically varied the variance, skewness and kurtosis parameters in one of the groups. The results of these simulations support the observation that the trade-off between sensitivity and specificity is determined by group differences in the variance parameter rather changes in skewness and kurtosis. Accordingly, the larger variance in the healthy control group may have led to consistently larger sensitivity and smaller specificity scores.

Because the multivariate searchlight combines signals from several voxels within a region, it is more sensitive to local information and shows a larger classification performance compared to univariate analysis, but at the same time provides regionally specific information. The simplicity of implementation and interpretability of regional pattern as well as the avoidance of critical prerequisites of whole brain decoding such as the choice of feature selection or dimensionality reduction technique and optimal feature size emerged as pivotal advantages of the searchlight technique compared to whole brain decoding strategies [64,65]. Apart from advantages of SVR and SVC machines some points have to be taken into account: The choice of the kernel function, the optimal selection of the meta-parameters (weighting of misclassifications, C and size of the insensitive loss region, ε) and the kernel parameters have an impact on the generalization performance and raise the problem of empirical tuning. While a nonlinear kernel provides equal or better prediction performances, the parameters of the solved model are difficult to interpret. Finally, the support vector regression as used here yields regression estimations without providing the direction of the relationship between predicted values based on multivariate data (voxel values) and the predictor variable (severity of negative symptoms).

Our results not only show that MVPA improves classification accuracy when compared to univariate methods but also suggest that the searchlight-based analysis of local pattern information can yield classification accuracies that may be even useful for individualized clinical decisions [66]. The used multivariate approach can be seen as proof of concept for the attempt to bridge the gap between the univariate approach, which merely depicts regional differences, and the diagnostic classification of the individual based on multivariate pattern information. The goal of this approach is not to replace clinical diagnosis. However, with the advance of machine learning techniques, MVPA has the potential to identify neuroimaging–based patterns with pathophysiological relevance and may serve as a basis for improved classification and differential diagnosis in the future.

Taken together, in this proof of concept study we were able to identify neurobiological markers of high diagnostic information for schizophrenia using searchlight MVPA. Our results show that searchlight MVPA can be used to substantially increase the accuracy of diagnostic classification on the basis of task-related fMRI signal patterns in a regionally specific way. This approach might help to identify biological diagnostic markers for schizophrenia that could be integrated in diagnostic systems in the future.

Author Contributions

Conceived and designed the experiments: SPK CH AH FS PS. Performed the experiments: SPK CH. Analyzed the data: SPK CH. Contributed reagents/materials/analysis tools: SPK CH. Wrote the paper: SPK CH JDH AH FS PS.

References

1. Abi-Dargham A, Rodenhiser J, Printz D, Zea-Ponce Y, Gil R, Kegeles LS, et al. Increased baseline occupancy of D2 receptors by dopamine in schizophrenia. Proc Natl Acad Sci U S A. 2000;97: 8104–8109. pmid:10884434
- View Article
- PubMed/NCBI
- Google Scholar
2. Heinz A, Schlagenhauf F. Dopaminergic dysfunction in schizophrenia: salience attribution revisited. Schizophr Bull. 2010;36: 472–485. pmid:20453041
- View Article
- PubMed/NCBI
- Google Scholar
3. Heinz A. Dopaminergic dysfunction in alcoholism and schizophrenia—psychopathological and behavioral correlates. Eur Psychiatry J Assoc Eur Psychiatr. 2002;17: 9–16.
- View Article
- Google Scholar
4. Kapur S. Psychosis as a State of Aberrant Salience: A Framework Linking Biology, Phenomenology, and Pharmacology in Schizophrenia. Am J Psychiatry. 2003;160: 13–23. pmid:12505794
- View Article
- PubMed/NCBI
- Google Scholar
5. Winton-Brown TT, Fusar-Poli P, Ungless MA, Howes OD. Dopaminergic basis of salience dysregulation in psychosis. Trends Neurosci. 2014;37: 85–94. pmid:24388426
- View Article
- PubMed/NCBI
- Google Scholar
6. Esslinger C, Englisch S, Inta D, Rausch F, Schirmbeck F, Mier D, et al. Ventral striatal activation during attribution of stimulus saliency and reward anticipation is correlated in unmedicated first episode schizophrenia patients. Schizophr Res. 2012;140: 114–121. pmid:22784688
- View Article
- PubMed/NCBI
- Google Scholar
7. Juckel G, Schlagenhauf F, Koslowski M, Wüstenberg T, Villringer A, Knutson B, et al. Dysfunction of ventral striatal reward prediction in schizophrenia. NeuroImage. 2006;29: 409–416. pmid:16139525
- View Article
- PubMed/NCBI
- Google Scholar
8. Murray GK, Corlett PR, Clark L, Pessiglione M, Blackwell AD, Honey G, et al. Substantia nigra/ventral tegmental reward prediction error disruption in psychosis. Mol Psychiatry. 2007;13: 267–276.
- View Article
- Google Scholar
9. Schlagenhauf F, Juckel G, Koslowski M, Kahnt T, Knutson B, Dembler T, et al. Reward system activation in schizophrenic patients switched from typical neuroleptics to olanzapine. Psychopharmacology (Berl). 2008;196: 673–684. pmid:18097655
- View Article
- PubMed/NCBI
- Google Scholar
10. Morris RW, Vercammen A, Lenroot R, Moore L, Langton JM, Short B, et al. Disambiguating ventral striatum fMRI-related bold signal during reward prediction in schizophrenia. Mol Psychiatry. 2012;17: 280–289.
- View Article
- Google Scholar
11. Crespo-Facorro B, Paradiso S, Andreasen NC, O’Leary DS, Watkins GL, Ponto LL, et al. Neural mechanisms of anhedonia in schizophrenia: a PET study of response to unpleasant and pleasant odors. JAMA J Am Med Assoc. 2001;286: 427–435.
- View Article
- Google Scholar
12. Paradiso S, Andreasen NC, Crespo-Facorro B, O’Leary DS, Watkins GL, Boles Ponto LL, et al. Emotions in unmedicated patients with schizophrenia during evaluation with positron emission tomography. Am J Psychiatry. 2003;160: 1775–1783. pmid:14514490
- View Article
- PubMed/NCBI
- Google Scholar
13. Takahashi H, Koeda M, Oda K, Matsuda T, Matsushima E, Matsuura M, et al. An fMRI study of differential neural response to affective pictures in schizophrenia. NeuroImage. 2004;22: 1247–1254. pmid:15219596
- View Article
- PubMed/NCBI
- Google Scholar
14. Taylor SF, Liberzon I, Decker LR, Koeppe RA. A functional anatomic study of emotion in schizophrenia. Schizophr Res. 2002;58: 159–172. pmid:12409155
- View Article
- PubMed/NCBI
- Google Scholar
15. Haynes J-D, Rees G. Decoding mental states from brain activity in humans. Nat Rev Neurosci. 2006;7: 523–534. pmid:16791142
- View Article
- PubMed/NCBI
- Google Scholar
16. Norman KA, Polyn SM, Detre GJ, Haxby JV. Beyond mind-reading: multi-voxel pattern analysis of fMRI data. Trends Cogn Sci. 2006;10: 424–430. pmid:16899397
- View Article
- PubMed/NCBI
- Google Scholar
17. Cox DD, Savoy RL. Functional magnetic resonance imaging (fMRI) “brain reading”: detecting and classifying distributed patterns of fMRI activity in human visual cortex. NeuroImage. 2003;19: 261–270. pmid:12814577
- View Article
- PubMed/NCBI
- Google Scholar
18. Vapnik VN. The Nature of Statistical Learning Theory. New York, NY, USA: Springer-Verlag New York, Inc.; 1995.
19. Klöppel S, Abdulkadir A, Jack CR, Koutsouleris N, Mourão-Miranda J, Vemuri P. Diagnostic neuroimaging across diseases. NeuroImage. 2012;61: 457–463. pmid:22094642
- View Article
- PubMed/NCBI
- Google Scholar
20. Sundermann B, Herr D, Schwindt W, Pfleiderer B. Multivariate classification of blood oxygen level-dependent FMRI data with diagnostic intention: a clinical perspective. AJNR Am J Neuroradiol. 2014;35: 848–855. pmid:24029388
- View Article
- PubMed/NCBI
- Google Scholar
21. Pereira F, Mitchell T, Botvinick M. Machine learning classifiers and fMRI: a tutorial overview. NeuroImage. 2009;45: S199–209. pmid:19070668
- View Article
- PubMed/NCBI
- Google Scholar
22. Orrù G, Pettersson-Yeo W, Marquand AF, Sartori G, Mechelli A. Using Support Vector Machine to identify imaging biomarkers of neurological and psychiatric disease: A critical review. Neurosci Biobehav Rev. 2012;36: 1140–1152. pmid:22305994
- View Article
- PubMed/NCBI
- Google Scholar
23. Castro E, Martínez-Ramón M, Pearlson G, Sui J, Calhoun VD. Characterization of groups using composite kernels and multi-source fMRI analysis data: Application to schizophrenia. NeuroImage. 2011;58: 526–536. pmid:21723948
- View Article
- PubMed/NCBI
- Google Scholar
24. Davatzikos C, Shen D, Gur RC, Wu X, Liu D, Fan Y, et al. Whole-brain morphometric study of schizophrenia revealing a spatially complex set of focal abnormalities. Arch Gen Psychiatry. 2005;62: 1218–1227. pmid:16275809
- View Article
- PubMed/NCBI
- Google Scholar
25. Ingalhalikar M, Kanterakis S, Gur R, Roberts TPL, Verma R. DTI based diagnostic prediction of a disease via pattern classification. Med Image Comput Comput-Assist Interv MICCAI Int Conf Med Image Comput Comput-Assist Interv. 2010;13: 558–565. pmid:20879275
- View Article
- PubMed/NCBI
- Google Scholar
26. Iwabuchi SJ, Liddle PF, Palaniyappan L. Clinical utility of machine-learning approaches in schizophrenia: improving diagnostic confidence for translational neuroimaging. Front Psychiatry. 2013;4: 95. pmid:24009589
- View Article
- PubMed/NCBI
- Google Scholar
27. Sun D, van Erp TGM, Thompson PM, Bearden CE, Daley M, Kushan L, et al. Elucidating a Magnetic Resonance Imaging-Based Neuroanatomic Biomarker for Psychosis: Classification Analysis Using Probabilistic Brain Atlas and Machine Learning Algorithms. Biol Psychiatry. 2009;66: 1055–1060. pmid:19729150
- View Article
- PubMed/NCBI
- Google Scholar
28. Yu Y, Shen H, Zhang H, Zeng L-L, Xue Z, Hu D. Functional connectivity-based signatures of schizophrenia revealed by multiclass pattern analysis of resting-state fMRI from schizophrenic patients and their healthy siblings. Biomed Eng Online. 2013;12: 1–1. pmid:23289769
- View Article
- PubMed/NCBI
- Google Scholar
29. Costafreda SG, Fu CHY, Picchioni M, Toulopoulou T, McDonald C, Kravariti E, et al. Pattern of neural responses to verbal fluency shows diagnostic specificity for schizophrenia and bipolar disorder. BMC Psychiatry. 2011;11: 18. pmid:21276242
- View Article
- PubMed/NCBI
- Google Scholar
30. Yang H, Liu J, Sui J, Pearlson G, Calhoun VD. A Hybrid Machine Learning Method for Fusing fMRI and Genetic Data: Combining both Improves Classification of Schizophrenia. Front Hum Neurosci. 2010;4: 192. pmid:21119772
- View Article
- PubMed/NCBI
- Google Scholar
31. Hägele C, Schlagenhauf F, Rapp M, Sterzer P, Beck A, Bermpohl F, et al. Dimensional psychiatry: reward dysfunction and depressive mood across psychiatric disorders. Psychopharmacology (Berl). 2014; https://doi.org/10.1007/s00213-014-3662-7
32. Haynes J-D, Sakai K, Rees G, Gilbert S, Frith C, Passingham RE. Reading Hidden Intentions in the Human Brain. Curr Biol. 2007;17: 323–328. pmid:17291759
- View Article
- PubMed/NCBI
- Google Scholar
33. Kriegeskorte N, Goebel R, Bandettini P. Information-based functional brain mapping. Proc Natl Acad Sci U S A. 2006;103: 3863–3868. pmid:16537458
- View Article
- PubMed/NCBI
- Google Scholar
34. Weygandt M, Blecker CR, Schäfer A, Hackmack K, Haynes J-D, Vaitl D, et al. fMRI pattern recognition in obsessive-compulsive disorder. NeuroImage. 2012;60: 1186–1193. pmid:22281674
- View Article
- PubMed/NCBI
- Google Scholar
35. Weygandt M, Schaefer A, Schienle A, Haynes J-D. Diagnosing different binge-eating disorders based on reward-related brain activation patterns. Hum Brain Mapp. 2012;33: 2135–2146. pmid:22887826
- View Article
- PubMed/NCBI
- Google Scholar
36. Juckel G, Schlagenhauf F, Koslowski M, Filonov D, Wüstenberg T, Villringer A, et al. Dysfunction of ventral striatal reward prediction in schizophrenic patients treated with typical, not atypical, neuroleptics. Psychopharmacology (Berl). 2006;187: 222–228. pmid:16721614
- View Article
- PubMed/NCBI
- Google Scholar
37. First MB, Spitzer RL, Gibbon M, Williams JB. Structured Clinical Interview for DSM-IV-TR Axis I Disorders—Patient Edition (SCID-I/P. 2/2001 Revision). N Y Biom Res Dep N Y State Psychiatr Inst. 2001;
38. Kay SR, Flszbein A, Opfer LA. The Positive and Negative Syndrome Scale (PANSS) for Schizophrenia. Schizophr Bull. 1987;13: 261–276. pmid:3616518
- View Article
- PubMed/NCBI
- Google Scholar
39. Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia. 1971;9: 97–113. pmid:5146491
- View Article
- PubMed/NCBI
- Google Scholar
40. Knutson B, Westdorp A, Kaiser E, Hommer D. FMRI visualization of brain activity during a monetary incentive delay task. NeuroImage. 2000;12: 20–27. pmid:10875899
- View Article
- PubMed/NCBI
- Google Scholar
41. Knutson B, Adams CM, Fong GW, Hommer D. Anticipation of increasing monetary reward selectively recruits nucleus accumbens. J Neurosci. 2001;21: 1–5. pmid:11312316
- View Article
- PubMed/NCBI
- Google Scholar
42. Friston KJ, Holmes AP, Worsley KJ, Poline J-P, Frith CD, Frackowiak RSJ. Statistical parametric maps in functional imaging: A general linear approach. Hum Brain Mapp. 1994;2: 189–210.
- View Article
- Google Scholar
43. Cohen MS. Parametric analysis of fMRI data using linear systems methods. NeuroImage. 1997;6: 93–103. pmid:9299383
- View Article
- PubMed/NCBI
- Google Scholar
44. Waltz JA, Schweitzer JB, Ross TJ, Kurup PK, Salmeron BJ, Rose EJ, et al. Abnormal responses to monetary outcomes in cortex, but not in the basal ganglia, in schizophrenia. Neuropsychopharmacology. 2010;35: 2427–2439. pmid:20720534
- View Article
- PubMed/NCBI
- Google Scholar
45. Beck A, Schlagenhauf F, Wüstenberg T, Hein J, Kienast T, Kahnt T, et al. Ventral striatal activation during reward anticipation correlates with impulsivity in alcoholics. Biol Psychiatry. 2009;66: 734–742. pmid:19560123
- View Article
- PubMed/NCBI
- Google Scholar
46. Boser BE, Guyon IM, Vapnik VN. A training algorithm for optimal margin classifiers. Proceedings of the fifth annual workshop on Computational learning theory. ACM; 1992. pp. 144–152. Available: http://dl.acm.org/citation.cfm?id=130401
47. Cortes C, Vapnik V. Support-vector networks. Mach Learn. 1995;20: 273–297.
- View Article
- Google Scholar
48. Schmah T, Yourganov G, Zemel RS, Hinton GE, Small SL, Strother SC. Comparing classification methods for longitudinal fMRI studies. Neural Comput. 2010;22: 2729–2762. pmid:20804386
- View Article
- PubMed/NCBI
- Google Scholar
49. Shalev-Shwartz S, Singer Y, Srebro N, Cotter A. Pegasos: primal estimated sub-gradient solver for SVM. Math Program. 2011;127: 3–30.
- View Article
- Google Scholar
50. Chen Y, Namburi P, Elliott LT, Heinzle J, Soon CS, Chee MWL, et al. Cortical surface-based searchlight decoding. NeuroImage. 2011;56: 582–592. pmid:20656043
- View Article
- PubMed/NCBI
- Google Scholar
51. Golland P, Fischl B. Permutation tests for classification: towards statistical significance in image-based studies. Inf Process Med Imaging Proc Conf. 2003;18: 330–341. pmid:15344469
- View Article
- PubMed/NCBI
- Google Scholar
52. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Ser B Methodol. 1995;57: 289–300.
- View Article
- Google Scholar
53. Metz CE. Basic principles of ROC analysis. Semin Nucl Med. 1978;8: 283–298. pmid:112681
- View Article
- PubMed/NCBI
- Google Scholar
54. Chang C-C, Lin C-J. LIBSVM: a library for support vector machines. ACM Trans Intell Syst Technol TIST. 2011;2: 27.
- View Article
- Google Scholar
55. Smola AJ, Schölkopf B. A tutorial on support vector regression. Stat Comput. 2004;14: 199–222.
- View Article
- Google Scholar
56. Jensen J, Smith AJ, Willeit M, Crawley AP, Mikulis DJ, Vitcu I, et al. Separate brain regions code for salience vs. valence during reward prediction in humans. Hum Brain Mapp. 2007;28: 294–302. pmid:16779798
- View Article
- PubMed/NCBI
- Google Scholar
57. Jensen J, McIntosh AR, Crawley AP, Mikulis DJ, Remington G, Kapur S. Direct Activation of the Ventral Striatum in Anticipation of Aversive Stimuli. Neuron. 2003;40: 1251–1257. pmid:14687557
- View Article
- PubMed/NCBI
- Google Scholar
58. Zink CF, Pagnoni G, Chappelow J, Martin-Skurski M, Berns GS. Human striatal activation reflects degree of stimulus saliency. NeuroImage. 2006;29: 977–983. pmid:16153860
- View Article
- PubMed/NCBI
- Google Scholar
59. Zink CF, Pagnoni G, Martin-Skurski ME, Chappelow JC, Berns GS. Human striatal responses to monetary reward depend on saliency. Neuron. 2004;42: 509–517. pmid:15134646
- View Article
- PubMed/NCBI
- Google Scholar
60. Dowd EC, Barch DM. Pavlovian reward prediction and receipt in schizophrenia: relationship to anhedonia. PloS One. 2012;7: e35622. pmid:22574121
- View Article
- PubMed/NCBI
- Google Scholar
61. Simon JJ, Biller A, Walther S, Roesch-Ely D, Stippich C, Weisbrod M, et al. Neural correlates of reward processing in schizophrenia—Relationship to apathy and depression. Schizophr Res. 2010;118: 154–161. pmid:20005675
- View Article
- PubMed/NCBI
- Google Scholar
62. Goldstein RZ, Volkow ND. Drug addiction and its underlying neurobiological basis: neuroimaging evidence for the involvement of the frontal cortex. Am J Psychiatry. 2002;159: 1642–1652. pmid:12359667
- View Article
- PubMed/NCBI
- Google Scholar
63. Wise RA. Neuroleptics and operant behavior: the anhedonia hypothesis. Behav Brain Sci. 1982;5: 39–53.
- View Article
- Google Scholar
64. Guyon I, Elisseeff A. An Introduction to Variable and Feature Selection. J Mach Learn Res. 2003;3: 1157–1182.
- View Article
- Google Scholar
65. Saeys Y, Inza I, Larrañaga P. A review of feature selection techniques in. Bioinformatics. 2007;23: 2507–2517. pmid:17720704
- View Article
- PubMed/NCBI
- Google Scholar
66. Koutsouleris N, Borgwardt S, Meisenzahl EM, Bottlender R, Möller H-J, Riecher-Rössler A. Disease prediction in the at-risk mental state for psychosis using neuroanatomical biomarkers: results from the FePsy study. Schizophr Bull. 2012;38: 1234–1246. pmid:22080496
- View Article
- PubMed/NCBI
- Google Scholar

[ref1] 1. Abi-Dargham A, Rodenhiser J, Printz D, Zea-Ponce Y, Gil R, Kegeles LS, et al. Increased baseline occupancy of D2 receptors by dopamine in schizophrenia. Proc Natl Acad Sci U S A. 2000;97: 8104–8109. pmid:10884434
View Article
PubMed/NCBI
Google Scholar

[2] View Article

[3] PubMed/NCBI

[4] Google Scholar

[ref2] 2. Heinz A, Schlagenhauf F. Dopaminergic dysfunction in schizophrenia: salience attribution revisited. Schizophr Bull. 2010;36: 472–485. pmid:20453041
View Article
PubMed/NCBI
Google Scholar

[6] View Article

[7] PubMed/NCBI

[8] Google Scholar

[ref3] 3. Heinz A. Dopaminergic dysfunction in alcoholism and schizophrenia—psychopathological and behavioral correlates. Eur Psychiatry J Assoc Eur Psychiatr. 2002;17: 9–16.
View Article
Google Scholar

[10] View Article

[11] Google Scholar

[ref4] 4. Kapur S. Psychosis as a State of Aberrant Salience: A Framework Linking Biology, Phenomenology, and Pharmacology in Schizophrenia. Am J Psychiatry. 2003;160: 13–23. pmid:12505794
View Article
PubMed/NCBI
Google Scholar

[13] View Article

[14] PubMed/NCBI

[15] Google Scholar

[ref5] 5. Winton-Brown TT, Fusar-Poli P, Ungless MA, Howes OD. Dopaminergic basis of salience dysregulation in psychosis. Trends Neurosci. 2014;37: 85–94. pmid:24388426
View Article
PubMed/NCBI
Google Scholar

[17] View Article

[18] PubMed/NCBI

[19] Google Scholar

[ref6] 6. Esslinger C, Englisch S, Inta D, Rausch F, Schirmbeck F, Mier D, et al. Ventral striatal activation during attribution of stimulus saliency and reward anticipation is correlated in unmedicated first episode schizophrenia patients. Schizophr Res. 2012;140: 114–121. pmid:22784688
View Article
PubMed/NCBI
Google Scholar

[21] View Article

[22] PubMed/NCBI

[23] Google Scholar

[ref7] 7. Juckel G, Schlagenhauf F, Koslowski M, Wüstenberg T, Villringer A, Knutson B, et al. Dysfunction of ventral striatal reward prediction in schizophrenia. NeuroImage. 2006;29: 409–416. pmid:16139525
View Article
PubMed/NCBI
Google Scholar

[25] View Article

[26] PubMed/NCBI

[27] Google Scholar

[ref8] 8. Murray GK, Corlett PR, Clark L, Pessiglione M, Blackwell AD, Honey G, et al. Substantia nigra/ventral tegmental reward prediction error disruption in psychosis. Mol Psychiatry. 2007;13: 267–276.
View Article
Google Scholar

[29] View Article

[30] Google Scholar

[ref9] 9. Schlagenhauf F, Juckel G, Koslowski M, Kahnt T, Knutson B, Dembler T, et al. Reward system activation in schizophrenic patients switched from typical neuroleptics to olanzapine. Psychopharmacology (Berl). 2008;196: 673–684. pmid:18097655
View Article
PubMed/NCBI
Google Scholar

[32] View Article

[33] PubMed/NCBI

[34] Google Scholar

[ref10] 10. Morris RW, Vercammen A, Lenroot R, Moore L, Langton JM, Short B, et al. Disambiguating ventral striatum fMRI-related bold signal during reward prediction in schizophrenia. Mol Psychiatry. 2012;17: 280–289.
View Article
Google Scholar

[36] View Article

[37] Google Scholar

[ref11] 11. Crespo-Facorro B, Paradiso S, Andreasen NC, O’Leary DS, Watkins GL, Ponto LL, et al. Neural mechanisms of anhedonia in schizophrenia: a PET study of response to unpleasant and pleasant odors. JAMA J Am Med Assoc. 2001;286: 427–435.
View Article
Google Scholar

[39] View Article

[40] Google Scholar

[ref12] 12. Paradiso S, Andreasen NC, Crespo-Facorro B, O’Leary DS, Watkins GL, Boles Ponto LL, et al. Emotions in unmedicated patients with schizophrenia during evaluation with positron emission tomography. Am J Psychiatry. 2003;160: 1775–1783. pmid:14514490
View Article
PubMed/NCBI
Google Scholar

[42] View Article

[43] PubMed/NCBI

[44] Google Scholar

[ref13] 13. Takahashi H, Koeda M, Oda K, Matsuda T, Matsushima E, Matsuura M, et al. An fMRI study of differential neural response to affective pictures in schizophrenia. NeuroImage. 2004;22: 1247–1254. pmid:15219596
View Article
PubMed/NCBI
Google Scholar

[46] View Article

[47] PubMed/NCBI

[48] Google Scholar

[ref14] 14. Taylor SF, Liberzon I, Decker LR, Koeppe RA. A functional anatomic study of emotion in schizophrenia. Schizophr Res. 2002;58: 159–172. pmid:12409155
View Article
PubMed/NCBI
Google Scholar

[50] View Article

[51] PubMed/NCBI

[52] Google Scholar

[ref15] 15. Haynes J-D, Rees G. Decoding mental states from brain activity in humans. Nat Rev Neurosci. 2006;7: 523–534. pmid:16791142
View Article
PubMed/NCBI
Google Scholar

[54] View Article

[55] PubMed/NCBI

[56] Google Scholar

[ref16] 16. Norman KA, Polyn SM, Detre GJ, Haxby JV. Beyond mind-reading: multi-voxel pattern analysis of fMRI data. Trends Cogn Sci. 2006;10: 424–430. pmid:16899397
View Article
PubMed/NCBI
Google Scholar

[58] View Article

[59] PubMed/NCBI

[60] Google Scholar

[ref17] 17. Cox DD, Savoy RL. Functional magnetic resonance imaging (fMRI) “brain reading”: detecting and classifying distributed patterns of fMRI activity in human visual cortex. NeuroImage. 2003;19: 261–270. pmid:12814577
View Article
PubMed/NCBI
Google Scholar

[62] View Article

[63] PubMed/NCBI

[64] Google Scholar

[ref18] 18. Vapnik VN. The Nature of Statistical Learning Theory. New York, NY, USA: Springer-Verlag New York, Inc.; 1995.

[ref19] 19. Klöppel S, Abdulkadir A, Jack CR, Koutsouleris N, Mourão-Miranda J, Vemuri P. Diagnostic neuroimaging across diseases. NeuroImage. 2012;61: 457–463. pmid:22094642
View Article
PubMed/NCBI
Google Scholar

[67] View Article

[68] PubMed/NCBI

[69] Google Scholar

[ref20] 20. Sundermann B, Herr D, Schwindt W, Pfleiderer B. Multivariate classification of blood oxygen level-dependent FMRI data with diagnostic intention: a clinical perspective. AJNR Am J Neuroradiol. 2014;35: 848–855. pmid:24029388
View Article
PubMed/NCBI
Google Scholar

[71] View Article

[72] PubMed/NCBI

[73] Google Scholar

[ref21] 21. Pereira F, Mitchell T, Botvinick M. Machine learning classifiers and fMRI: a tutorial overview. NeuroImage. 2009;45: S199–209. pmid:19070668
View Article
PubMed/NCBI
Google Scholar

[75] View Article

[76] PubMed/NCBI

[77] Google Scholar

[ref22] 22. Orrù G, Pettersson-Yeo W, Marquand AF, Sartori G, Mechelli A. Using Support Vector Machine to identify imaging biomarkers of neurological and psychiatric disease: A critical review. Neurosci Biobehav Rev. 2012;36: 1140–1152. pmid:22305994
View Article
PubMed/NCBI
Google Scholar

[79] View Article

[80] PubMed/NCBI

[81] Google Scholar

[ref23] 23. Castro E, Martínez-Ramón M, Pearlson G, Sui J, Calhoun VD. Characterization of groups using composite kernels and multi-source fMRI analysis data: Application to schizophrenia. NeuroImage. 2011;58: 526–536. pmid:21723948
View Article
PubMed/NCBI
Google Scholar

[83] View Article

[84] PubMed/NCBI

[85] Google Scholar

[ref24] 24. Davatzikos C, Shen D, Gur RC, Wu X, Liu D, Fan Y, et al. Whole-brain morphometric study of schizophrenia revealing a spatially complex set of focal abnormalities. Arch Gen Psychiatry. 2005;62: 1218–1227. pmid:16275809
View Article
PubMed/NCBI
Google Scholar

[87] View Article

[88] PubMed/NCBI

[89] Google Scholar

[ref25] 25. Ingalhalikar M, Kanterakis S, Gur R, Roberts TPL, Verma R. DTI based diagnostic prediction of a disease via pattern classification. Med Image Comput Comput-Assist Interv MICCAI Int Conf Med Image Comput Comput-Assist Interv. 2010;13: 558–565. pmid:20879275
View Article
PubMed/NCBI
Google Scholar

[91] View Article

[92] PubMed/NCBI

[93] Google Scholar

[ref26] 26. Iwabuchi SJ, Liddle PF, Palaniyappan L. Clinical utility of machine-learning approaches in schizophrenia: improving diagnostic confidence for translational neuroimaging. Front Psychiatry. 2013;4: 95. pmid:24009589
View Article
PubMed/NCBI
Google Scholar

[95] View Article

[96] PubMed/NCBI

[97] Google Scholar

[ref27] 27. Sun D, van Erp TGM, Thompson PM, Bearden CE, Daley M, Kushan L, et al. Elucidating a Magnetic Resonance Imaging-Based Neuroanatomic Biomarker for Psychosis: Classification Analysis Using Probabilistic Brain Atlas and Machine Learning Algorithms. Biol Psychiatry. 2009;66: 1055–1060. pmid:19729150
View Article
PubMed/NCBI
Google Scholar

[99] View Article

[100] PubMed/NCBI

[101] Google Scholar

[ref28] 28. Yu Y, Shen H, Zhang H, Zeng L-L, Xue Z, Hu D. Functional connectivity-based signatures of schizophrenia revealed by multiclass pattern analysis of resting-state fMRI from schizophrenic patients and their healthy siblings. Biomed Eng Online. 2013;12: 1–1. pmid:23289769
View Article
PubMed/NCBI
Google Scholar

[103] View Article

[104] PubMed/NCBI

[105] Google Scholar

[ref29] 29. Costafreda SG, Fu CHY, Picchioni M, Toulopoulou T, McDonald C, Kravariti E, et al. Pattern of neural responses to verbal fluency shows diagnostic specificity for schizophrenia and bipolar disorder. BMC Psychiatry. 2011;11: 18. pmid:21276242
View Article
PubMed/NCBI
Google Scholar

[107] View Article

[108] PubMed/NCBI

[109] Google Scholar

[ref30] 30. Yang H, Liu J, Sui J, Pearlson G, Calhoun VD. A Hybrid Machine Learning Method for Fusing fMRI and Genetic Data: Combining both Improves Classification of Schizophrenia. Front Hum Neurosci. 2010;4: 192. pmid:21119772
View Article
PubMed/NCBI
Google Scholar

[111] View Article

[112] PubMed/NCBI

[113] Google Scholar

[ref31] 31. Hägele C, Schlagenhauf F, Rapp M, Sterzer P, Beck A, Bermpohl F, et al. Dimensional psychiatry: reward dysfunction and depressive mood across psychiatric disorders. Psychopharmacology (Berl). 2014; https://doi.org/10.1007/s00213-014-3662-7

[ref32] 32. Haynes J-D, Sakai K, Rees G, Gilbert S, Frith C, Passingham RE. Reading Hidden Intentions in the Human Brain. Curr Biol. 2007;17: 323–328. pmid:17291759
View Article
PubMed/NCBI
Google Scholar

[116] View Article

[117] PubMed/NCBI

[118] Google Scholar

[ref33] 33. Kriegeskorte N, Goebel R, Bandettini P. Information-based functional brain mapping. Proc Natl Acad Sci U S A. 2006;103: 3863–3868. pmid:16537458
View Article
PubMed/NCBI
Google Scholar

[120] View Article

[121] PubMed/NCBI

[122] Google Scholar

[ref34] 34. Weygandt M, Blecker CR, Schäfer A, Hackmack K, Haynes J-D, Vaitl D, et al. fMRI pattern recognition in obsessive-compulsive disorder. NeuroImage. 2012;60: 1186–1193. pmid:22281674
View Article
PubMed/NCBI
Google Scholar

[124] View Article

[125] PubMed/NCBI

[126] Google Scholar

[ref35] 35. Weygandt M, Schaefer A, Schienle A, Haynes J-D. Diagnosing different binge-eating disorders based on reward-related brain activation patterns. Hum Brain Mapp. 2012;33: 2135–2146. pmid:22887826
View Article
PubMed/NCBI
Google Scholar

[128] View Article

[129] PubMed/NCBI

[130] Google Scholar

[ref36] 36. Juckel G, Schlagenhauf F, Koslowski M, Filonov D, Wüstenberg T, Villringer A, et al. Dysfunction of ventral striatal reward prediction in schizophrenic patients treated with typical, not atypical, neuroleptics. Psychopharmacology (Berl). 2006;187: 222–228. pmid:16721614
View Article
PubMed/NCBI
Google Scholar

[132] View Article

[133] PubMed/NCBI

[134] Google Scholar

[ref37] 37. First MB, Spitzer RL, Gibbon M, Williams JB. Structured Clinical Interview for DSM-IV-TR Axis I Disorders—Patient Edition (SCID-I/P. 2/2001 Revision). N Y Biom Res Dep N Y State Psychiatr Inst. 2001;

[ref38] 38. Kay SR, Flszbein A, Opfer LA. The Positive and Negative Syndrome Scale (PANSS) for Schizophrenia. Schizophr Bull. 1987;13: 261–276. pmid:3616518
View Article
PubMed/NCBI
Google Scholar

[137] View Article

[138] PubMed/NCBI

[139] Google Scholar

[ref39] 39. Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia. 1971;9: 97–113. pmid:5146491
View Article
PubMed/NCBI
Google Scholar

[141] View Article

[142] PubMed/NCBI

[143] Google Scholar

[ref40] 40. Knutson B, Westdorp A, Kaiser E, Hommer D. FMRI visualization of brain activity during a monetary incentive delay task. NeuroImage. 2000;12: 20–27. pmid:10875899
View Article
PubMed/NCBI
Google Scholar

[145] View Article

[146] PubMed/NCBI

[147] Google Scholar

[ref41] 41. Knutson B, Adams CM, Fong GW, Hommer D. Anticipation of increasing monetary reward selectively recruits nucleus accumbens. J Neurosci. 2001;21: 1–5. pmid:11312316
View Article
PubMed/NCBI
Google Scholar

[149] View Article

[150] PubMed/NCBI

[151] Google Scholar

[ref42] 42. Friston KJ, Holmes AP, Worsley KJ, Poline J-P, Frith CD, Frackowiak RSJ. Statistical parametric maps in functional imaging: A general linear approach. Hum Brain Mapp. 1994;2: 189–210.
View Article
Google Scholar

[153] View Article

[154] Google Scholar

[ref43] 43. Cohen MS. Parametric analysis of fMRI data using linear systems methods. NeuroImage. 1997;6: 93–103. pmid:9299383
View Article
PubMed/NCBI
Google Scholar

[156] View Article

[157] PubMed/NCBI

[158] Google Scholar

[ref44] 44. Waltz JA, Schweitzer JB, Ross TJ, Kurup PK, Salmeron BJ, Rose EJ, et al. Abnormal responses to monetary outcomes in cortex, but not in the basal ganglia, in schizophrenia. Neuropsychopharmacology. 2010;35: 2427–2439. pmid:20720534
View Article
PubMed/NCBI
Google Scholar

[160] View Article

[161] PubMed/NCBI

[162] Google Scholar

[ref45] 45. Beck A, Schlagenhauf F, Wüstenberg T, Hein J, Kienast T, Kahnt T, et al. Ventral striatal activation during reward anticipation correlates with impulsivity in alcoholics. Biol Psychiatry. 2009;66: 734–742. pmid:19560123
View Article
PubMed/NCBI
Google Scholar

[164] View Article

[165] PubMed/NCBI

[166] Google Scholar

[ref46] 46. Boser BE, Guyon IM, Vapnik VN. A training algorithm for optimal margin classifiers. Proceedings of the fifth annual workshop on Computational learning theory. ACM; 1992. pp. 144–152. Available: http://dl.acm.org/citation.cfm?id=130401

[ref47] 47. Cortes C, Vapnik V. Support-vector networks. Mach Learn. 1995;20: 273–297.
View Article
Google Scholar

[169] View Article

[170] Google Scholar

[ref48] 48. Schmah T, Yourganov G, Zemel RS, Hinton GE, Small SL, Strother SC. Comparing classification methods for longitudinal fMRI studies. Neural Comput. 2010;22: 2729–2762. pmid:20804386
View Article
PubMed/NCBI
Google Scholar

[172] View Article

[173] PubMed/NCBI

[174] Google Scholar

[ref49] 49. Shalev-Shwartz S, Singer Y, Srebro N, Cotter A. Pegasos: primal estimated sub-gradient solver for SVM. Math Program. 2011;127: 3–30.
View Article
Google Scholar

[176] View Article

[177] Google Scholar

[ref50] 50. Chen Y, Namburi P, Elliott LT, Heinzle J, Soon CS, Chee MWL, et al. Cortical surface-based searchlight decoding. NeuroImage. 2011;56: 582–592. pmid:20656043
View Article
PubMed/NCBI
Google Scholar

[179] View Article

[180] PubMed/NCBI

[181] Google Scholar

[ref51] 51. Golland P, Fischl B. Permutation tests for classification: towards statistical significance in image-based studies. Inf Process Med Imaging Proc Conf. 2003;18: 330–341. pmid:15344469
View Article
PubMed/NCBI
Google Scholar

[183] View Article

[184] PubMed/NCBI

[185] Google Scholar

[ref52] 52. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Ser B Methodol. 1995;57: 289–300.
View Article
Google Scholar

[187] View Article

[188] Google Scholar

[ref53] 53. Metz CE. Basic principles of ROC analysis. Semin Nucl Med. 1978;8: 283–298. pmid:112681
View Article
PubMed/NCBI
Google Scholar

[190] View Article

[191] PubMed/NCBI

[192] Google Scholar

[ref54] 54. Chang C-C, Lin C-J. LIBSVM: a library for support vector machines. ACM Trans Intell Syst Technol TIST. 2011;2: 27.
View Article
Google Scholar

[194] View Article

[195] Google Scholar

[ref55] 55. Smola AJ, Schölkopf B. A tutorial on support vector regression. Stat Comput. 2004;14: 199–222.
View Article
Google Scholar

[197] View Article

[198] Google Scholar

[ref56] 56. Jensen J, Smith AJ, Willeit M, Crawley AP, Mikulis DJ, Vitcu I, et al. Separate brain regions code for salience vs. valence during reward prediction in humans. Hum Brain Mapp. 2007;28: 294–302. pmid:16779798
View Article
PubMed/NCBI
Google Scholar

[200] View Article

[201] PubMed/NCBI

[202] Google Scholar

[ref57] 57. Jensen J, McIntosh AR, Crawley AP, Mikulis DJ, Remington G, Kapur S. Direct Activation of the Ventral Striatum in Anticipation of Aversive Stimuli. Neuron. 2003;40: 1251–1257. pmid:14687557
View Article
PubMed/NCBI
Google Scholar

[204] View Article

[205] PubMed/NCBI

[206] Google Scholar

[ref58] 58. Zink CF, Pagnoni G, Chappelow J, Martin-Skurski M, Berns GS. Human striatal activation reflects degree of stimulus saliency. NeuroImage. 2006;29: 977–983. pmid:16153860
View Article
PubMed/NCBI
Google Scholar

[208] View Article

[209] PubMed/NCBI

[210] Google Scholar

[ref59] 59. Zink CF, Pagnoni G, Martin-Skurski ME, Chappelow JC, Berns GS. Human striatal responses to monetary reward depend on saliency. Neuron. 2004;42: 509–517. pmid:15134646
View Article
PubMed/NCBI
Google Scholar

[212] View Article

[213] PubMed/NCBI

[214] Google Scholar

[ref60] 60. Dowd EC, Barch DM. Pavlovian reward prediction and receipt in schizophrenia: relationship to anhedonia. PloS One. 2012;7: e35622. pmid:22574121
View Article
PubMed/NCBI
Google Scholar

[216] View Article

[217] PubMed/NCBI

[218] Google Scholar

[ref61] 61. Simon JJ, Biller A, Walther S, Roesch-Ely D, Stippich C, Weisbrod M, et al. Neural correlates of reward processing in schizophrenia—Relationship to apathy and depression. Schizophr Res. 2010;118: 154–161. pmid:20005675
View Article
PubMed/NCBI
Google Scholar

[220] View Article

[221] PubMed/NCBI

[222] Google Scholar

[ref62] 62. Goldstein RZ, Volkow ND. Drug addiction and its underlying neurobiological basis: neuroimaging evidence for the involvement of the frontal cortex. Am J Psychiatry. 2002;159: 1642–1652. pmid:12359667
View Article
PubMed/NCBI
Google Scholar

[224] View Article

[225] PubMed/NCBI

[226] Google Scholar

[ref63] 63. Wise RA. Neuroleptics and operant behavior: the anhedonia hypothesis. Behav Brain Sci. 1982;5: 39–53.
View Article
Google Scholar

[228] View Article

[229] Google Scholar

[ref64] 64. Guyon I, Elisseeff A. An Introduction to Variable and Feature Selection. J Mach Learn Res. 2003;3: 1157–1182.
View Article
Google Scholar

[231] View Article

[232] Google Scholar

[ref65] 65. Saeys Y, Inza I, Larrañaga P. A review of feature selection techniques in. Bioinformatics. 2007;23: 2507–2517. pmid:17720704
View Article
PubMed/NCBI
Google Scholar

[234] View Article

[235] PubMed/NCBI

[236] Google Scholar

[ref66] 66. Koutsouleris N, Borgwardt S, Meisenzahl EM, Bottlender R, Möller H-J, Riecher-Rössler A. Disease prediction in the at-risk mental state for psychosis using neuroanatomical biomarkers: results from the FePsy study. Schizophr Bull. 2012;38: 1234–1246. pmid:22080496
View Article
PubMed/NCBI
Google Scholar

[238] View Article

[239] PubMed/NCBI

[240] Google Scholar

Figures

Abstract

Introduction

Materials and Methods

Participants

Monetary Incentive Delay Task and Data Acquisition

fMRI Data Acquisition

fMRI Data Analysis

Multivariate Pattern Analysis

Results

Behavioral Data

Univariate group differences in reward anticipation

MVPA classification of schizophrenia patients and healthy controls

Comparison of univariate and multivariate classification of patients and healthy controls for the ventral striatum

Multivariate prediction of the PANSS negative scale

Discussion

Author Contributions

References