Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using multi-parametric magnetic resonance imaging

SCHRES-08593; No of Pages 10 Schizophrenia Research xxx (xxxx) xxx

Contents lists available at ScienceDirect

Schizophrenia Research journal homepage: www.elsevier.com/locate/schres

Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using multi-parametric magnetic resonance imaging Jie Gong a,1, Long-Biao Cui b,1, Yi-Bin Xi c, Ying-Song Zhao a, Xue-Juan Yang a, Zi-liang Xu a, Jin-Bo Sun a, Peng Liu a, Jie Jia d, Ping Li e, Hong Yin c,⁎, Wei Qin a,⁎⁎ a

Engineering Research Center of Molecular and Neuro Imaging of Ministry of Education, School of Life Science and Technology, Xidian University, Xi'an, Shaanxi, China Department of Clinical Psychology, School of Medical Psychology, Fourth Military Medical University, Shaanxi, China c Department of Radiology, Xijing Hospital, Fourth Military Medical University, Xi'an, Shaanxi, China d Department of Early Intervention, Xi'an Mental Health Center, Xi'an, Shaanxi, China e Department of Medical Imaging, Xi'an Mental Health Center, Xi'an, Shaanxi, China b

a r t i c l e

i n f o

Article history: Received 21 May 2019 Received in revised form 4 September 2019 Accepted 25 November 2019 Available online xxxx Keywords: Schizophrenia Electroconvulsive therapy Radiomics Prediction Structural magnetic resonance imaging Diffusion magnetic resonance imaging

a b s t r a c t Electroconvulsive therapy (ECT) has been shown to be effective in schizophrenia, particularly when rapid symptom reduction is needed or in cases of resistance to drug treatment. However, there are no markers available to predict response to ECT. Here, we examine whether multi-parametric magnetic resonance imaging (MRI)-based radiomic features can predict response to ECT for individual patients. A total of 57 treatment-resistant schizophrenia patients, or schizophrenia patients with an acute episode or suicide attempts were randomly divided into primary (42 patients) and test (15 patients) cohorts. We collected T1-weighted structural MRI and diffusion MRI for 57 patients before receiving ECT and extracted 600 radiomic features for feature selection and prediction. To predict a continuous improvement in symptoms (ΔPANSS), the prediction process was performed with a support vector regression model based on a leave-one-out cross-validation framework in primary cohort and was tested in test cohort. The multi-parametric MRI-based radiomic model, including four structural MRI feature from left inferior frontal gyrus, right insula, left middle temporal gyrus and right superior temporal gyrus respectively and six diffusion MRI features from tracts connecting frontal or temporal gyrus possessed a low root mean square error of 15.183 in primary cohort and 14.980 in test cohort. The Pearson's correlation coefficients between predicted and actual values were 0.671 and 0.777 respectively. These results demonstrate that multi-parametric MRI-based radiomic features may predict response to ECT for individual patients. Such features could serve as prognostic neuroimaging biomarkers that provide a critical step toward individualized treatment response prediction in schizophrenia. © 2019 Published by Elsevier B.V.

1. Introduction Schizophrenia is a severe and highly disabling psychiatric disorder marked by delusions, hallucinations, poor motivation, and cognitive impairments (Sullivan et al., 2003; Whiteford et al., 2013). Up to 30% of patients with schizophrenia do not respond or respond poorly to standard treatment with antipsychotics (Freedman, 2003; Lieberman et al., 2005). However, there is growing evidence to suggest that ⁎ Correspondence to: H. Yin, Department of Radiology, Xijing Hospital, Fourth Military Medical University, No. 127 West Changle Road, Xi'an, Shaanxi 710032, China. ⁎⁎ Correspondence to: W. Qin, Engineering Research Center of Molecular and Neuro Imaging of the Ministry of Education, School of Life Science and Technology, Xidian University, 266 Xinglong Section of Xifeng Road, Xi'an, Shaanxi 710126, China. E-mail addresses: [email protected] (H. Yin), [email protected] (W. Qin). 1 These authors contributed equally to this work.

electroconvulsive therapy (ECT) has synergistic effects when applied in combination with antipsychotics, offering a potential safe and effective alternative for more severe and treatment-resistant patients (Ahmed et al., 2017; Sanghani et al., 2018; Weiner and Reti, 2017). Moreover, in spite of the significant limitations of data, ECT appears to have an acute benefit on suicidality in mood disorders (Sharma, 2001, 2003). Although 66% of patients with schizophrenia recover following ECT, a small proportion of these remain resistant to substantial improvement when receiving ECT alongside clozapine augmentation (Lally et al., 2016). Patients for whom treatment is ineffective must bear unnecessary healthcare costs, as well as the treatment-related toxicities of ineffective treatment (Case et al., 2013; Goodman, 2011; Prudic et al., 2004; Sackeim et al., 2007). To avoid the negative effects of ineffective treatment and to enable better treatment decisions, it is necessary to obtain

https://doi.org/10.1016/j.schres.2019.11.046 0920-9964/© 2019 Published by Elsevier B.V.

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

2

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

information about patients prior to ECT treatment that may be related to degree of recovery following treatment. There are currently no clinically available biomarkers. It would therefore be clinically useful to identify novel and non-invasive imaging biomarkers that could predict response to ECT prior to treatment in schizophrenia. Radiomics provides an unprecedented opportunity to improve individualized diagnosis and treatment by means of biological measurements for mental disorders (Chaddad et al., 2017; Cui et al., 2018). The important process of radiomics analysis includes high-throughput extraction to converse medical images into high-dimensional data and detection of crucial features for supporting decision making (Gillies et al., 2016). Previous studies have shown that objective and quantitative imaging descriptors could potentially be used as prognostic or predictive biomarkers (Gevaert et al., 2012; Kuo and Jamshidi, 2014; Mazurowski, 2015). The combined analysis of a panel of biomarkers, rather than individual analyses, as a signature is the most promising approach that is powerful enough to change clinical management (Birkhahn et al., 2007; Croner et al., 2008). Radiomics, which allows the investigation of multiple imaging features of multiple modalities in parallel, can provide a combination of features. The data mining process of the subsequent analysis usually comprises pattern recognition analysis, which can sift the complex brain patterns underlying neuropsychiatric disorders for clinically relevant predictive fingerprints (Iniesta et al., 2016; Orru et al., 2012). Many studies of mental disorders have successfully used pattern recognition for disease diagnosis and prediction of treatment response (Li et al., 2017; Redlich et al., 2014; Redlich et al., 2016; van Waarde et al., 2015). Therefore, radiomics including neuroimaging and pattern classification is a promising tool that can contribute to decision making to select appropriate treatment for psychotic disorders. Moreover, most past studies investigating prediction of ECT efficacy in mental disorders have used a single modality, such as sMRI (Redlich et al., 2016) or fMRI (Li et al., 2017), but have rarely used multimodal fusion to predict ECT efficacy. Multimodal fusion can help compensate for the limitations of single modality imaging studies and can also identify relationships between brain pathologies in psychosis because important pathophysiological questions can only be answered with cross-modal information (Calhoun and Sui, 2016; Plis et al., 2018; Schultz et al., 2012). Moreover, multimodal imaging offers a promising strategy in psychosis research to further elucidate the complex neuropathological mechanisms underlying schizophrenia (Calhoun and Sui, 2016; Lesh et al., 2015; Liang et al., 2019). Advanced machine learning techniques could be used as a multimodal fusion framework (Plis et al., 2018), with multimodal fusion used to build better models (Liang et al., 2019; Liu et al., 2017; Xi et al., 2018; Zhang et al., 2017). Previous studies have reported that regional changes in gray matter (GM) volume observed with sMRI play an important role in the pathophysiology of schizophrenia (Brugger and Howes, 2017; Dietsche et al., 2017; Jiang et al., 2019; Liu et al., 2019). Structural changes assessed with GM neuroanatomy of sMRI may also be important for the prediction of response to ECT, which has been found in many studies of depression (Oudega et al., 2014; Redlich et al., 2016; ten Doesschate et al., 2014), but is rarely studied in schizophrenia. The white matter (WM) architecture represents the structural connectivity of the brain and defects in WM integrity lead to communication defects between brain regions. Schizophrenia is thought to be a rupture syndrome (Andreasen et al., 1998; Crow, 1998; Davis et al., 2003; McGuire and Frith, 1996; Spalletta et al., 2003). Many studies have shown abnormal WM integrity in schizophrenia patients (Ardekani et al., 2003; Lim et al., 1999; Minami et al., 2003), especially the connections of frontal and temporal regions (Davidson and Heinrichs, 2003; Davis et al., 2003). A meta-analysis of 15 diffusion tensor imaging (DTI) studies with patients affected by chronic schizophrenia, observed significant fractional anisotropy (FA) reductions in frontal and temporal WM regions (Ellison-Wright and Bullmore, 2009). And another meta-analytic study found a widespread alteration of WM bundles in schizophrenia

mostly included in frontal, temporal, and limbic pathways (Vitolo et al., 2017). These studies emphasize that alterations of WM tracts involving frontal, temporal, and limbic circuits play an important role in schizophrenia. These tracts might play an important role in predicting therapeutic response. The results of studies on predicting the response to ECT with diffusion MRI (dMRI) are almost blank. We hypothesize that a multimodal neurologic signature comprising both GM neuroanatomy and WM microstructure could be a potential biomarker associated with treatment response in patients with schizophrenia. In ECT, a strong current is directly applied to the brain through scalp electrodes positioned near the temporal or frontal lobes. The brain regions with the strong electric field distribution generated by ECT electrode configurations are often most affected by ECT (simulated electrical field strength larger than 35 V/m) (Leaver et al., 2018; van Waarde et al., 2015). Analyzing imaging data from these regions that are directly stimulated by ECT may be particularly interesting. The use of ECT is recommended by Chinese guideline and consensus (2015, 2017, 2019) for treatmentresistant schizophrenia, and an acute episode and suicidal behavior/suicide attempts in schizophrenia. In this study, we aimed to investigate the ability of a multimodal fusion neurologic signature comprising GM neuroanatomical predictors from sMRI data and WM microstructural predictors from dMRI to predict ECT response in patients with schizophrenia. The aim was to identify quantitative and objective biomarkers for individualized clinical management of schizophrenia. 2. Methods 2.1. Patients A total of 57 patients with schizophrenia from the Department of Psychiatry at Mental Health Center of Xi'an City participated in this study. The consensus diagnoses were made by two experienced clinical psychiatrists on the basis of Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) using all the available information. Thirty three patients were those with treatment-resistant schizophrenia, 20 patients were those during an acute episode, and 4 patients were those with suicide attempts (Table 1). We used APA practice guidelines (Lehman et al., 2004), i.e., “Treatment resistance is defined as little or no symptomatic response to multiple (at least two) antipsychotic trials of an adequate duration (at least 6 weeks) and dose (therapeutic range).” The diagnoses and the inclusion/exclusion criteria are shown in Supplementary A1. 57 patients met the criteria were identified and divided into two cohorts at a ratio of 3:1 using computer generated random numbers. Forty-two patients were allocated to the primary cohort, while fifteen patients were allocated to the independent test cohort. Research procedures were approved by the local Institutional Ethics Committee and conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). All participants gave written informed consent after a full description of the aims and design of the study. 2.2. Electroconvulsive therapy and clinical assessment All patients received standard antipsychotic drugs combined with ECT during observation. Bitemporal ECT was conducted using an electric instrument (spECTrum 5000Q, MECTA, Tualatin, OR, USA).A total of 12 sessions of ECT were given (Standard treatment: three times a week for four weeks). The number of ECT sessions was individually determined by the clinician (mean number of ECT: 10.5 sessions, 2.4 SD). The details of ECT are described in Supplementary A2. Positive And Negative Syndrome Scale (PANSS) (Kay et al., 1987) that evaluates each patient's symptoms, mini-mental state examination (MMSE) that measures cognitive changes and Clinical Global Impression (CGI) were obtained before ECT course (T0) and after ECT course (T1). Endpoints for individualized prediction (treatment response)

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

3

Table 1 Clinical data of treatment-resistant schizophrenia patients (TRS), or schizophrenia patients with an acute episode (AE) or suicide attempts (SA). Mean and SD are given except when stated otherwise. Characteristic

TRS (n = 33)

AE (n = 20)

SA (n = 4)

p valuesa

Age (y) Gender (M/F) Education level (y) Duration of illness (y) CGIc score at baseline CGI score after ECT MMSEe at baseline MMSE after ECT PANSSf score at baseline Positive score Negative score General score Total score PANSS score after ECT Positive score Negative score General score Total score Changes in PANSS score Number of ECT Antipsychotic dose (mg/d)g

31.5 ± 10.6 24/9 10.8 ± 3.4 8.3 ± 6.9 5.5 ± 1.1 3.3 ± 1.0 (b0.001)d 21.5 ± 11.0 22.5 ± 10.5 (0.22)

27.5 ± 6.0 11/9 12.9 ± 3.9 1.7 ± 1.3 5.7 ± 0.5 3.1 ± 0.9 (b0.001) 24.1 ± 9.6 26.1 ± 7.3 (0.16)

34.5 ± 9.7 2/2 11.3 ± 4.5 6.3 ± 6.4 5.8 ± 0.5 2.8 ± 0.5 (b0.01) 23.5 ± 5.8 27.5 ± 2.1 (0.20)

0.20 0.34b 0.14 b0.001 0.78 0.51 0.68 0.33

28.8 ± 7.7 24.1 ± 11.8 41.3 ± 9.7 94.2 ± 18.0

29.4 ± 6.3 25.8 ± 11.2 46.0 ± 11.3 101.1 ± 22.6

27.3 ± 8.7 19.8 ± 12.0 41.3 ± 13.2 88.3 ± 21.9

0.87 0.62 0.29 0.35

12.6 ± 4.2 (b0.001) 15.1 ± 8.6 (b0.001) 23.9 ± 7.5 (b0.001) 51.6 ± 16.9 (b0.001) 67.4% ± 20.8% 9.8 ± 2.9 14.8 ± 7.4

12.4 ± 4.0 (b0.001) 16.2 ± 7.9 (b0.001) 24.3 ± 5.2 (b0.001) 52.8 ± 14.6 (b0.001) 67.5% ± 20.2% 11.4 ± 1.2 16.7 ± 5.1

13.5 ± 3.7 (0.08) 15.3 ± 8.5 (0.20) 25.3 ± 13.7 (0.01) 54.5 ± 23.2 (0.006) 62.8% ± 26.1% 11.5 ± 0.7 17.5 ± 6.5

0.89 0.90 0.94 0.93 0.91 0.32 0.51

a p values were obtained using analysis of variance (ANOVA) for comparison of clinical data (except gender) between the treatment-resistant schizophrenia patients (TRS), schizophrenia patients with an acute episode (AE) and schizophrenia patients with suicide attempts (SA). b p values were obtained using the chi-square. c CGI, Clinical Global Impression. d p values in parentheses were obtained using the paired 2-tailed t-test for comparison of clinical data between baseline and after treatment. e MMSE, mini-mental state examination. f PANSS, Positive And Negative Syndrome Scale. g Olanzapine equivalents based on defined daily doses method.

was assessed using percentage change of symptoms based on PANSS total scores (Cao et al., 2018a; Obermeier et al., 2010; Yu et al., 2018), which defined as follows: PANSST0 −PANSST1 ΔPANSS ¼ PANSST0 −30 The 30 in the denominator corresponds to the “non-symptomatic” score of PANSS total score. 2.3. Image acquisition A GE Discovery MR750 3.0 T scanner was used to acquire highresolution T1-weighted structural data (sMRI) and diffusion MRI data (dMRI) of all participants in the Department of Radiology at Mental Health Center of Xi'an City in China. The details of the acquisition protocol are shown in Supplementary A3. 2.4. Imaging data preprocessing Fig. 1 shows the workflow of this study. Structural MRI data and diffusion MRI data were preprocessed and used for feature extraction. Radiomic features were used for feature selection and prediction. 2.4.1. Definition of ROIs 23 regions of interest (ROIs) with strong electrical field distribution were defined and the details were described in Supplementary A4, Fig. S1 and Table S1. 2.4.2. Structural MRI data for extracting GM tissues of ROIs CAT12 toolbox (C. Gaser, Structural Brain Mapping group, Jena University Hospital, Jena, Germany) implemented in SPM12 (Statistical Parametric Mapping, Institute of Neurology, London, UK) was used to perform the voxel-based morphometry (VBM) analysis in the patients cohort (Ashburner and Friston, 2000). The details of pre-processing of

VBM analysis are shown in Supplementary A5. The modulated and normalized GM tissues of 23 ROIs were used for feature extraction. 2.4.3. Diffusion MRI data for extracting FA maps of the tracts associated with ECT treatment The PANDA pipeline (Cui et al., 2013), which incorporates FMRIB Software Library (Smith et al., 2004), Pipeline System for Octave and Matlab (PSOM) (Bellec et al., 2012), Diffusion Toolkit (Wang et al., 2007), and MRIcron (Rorden et al., 2007), was used to pre-processing the diffusion-weighted imaging data of each participant (Supplementary A6). The binary map of each WM tract was constructed from healthy participants, which are described in detail in Supplementary A7. The FA map for each patient in the MNI space was then overlaid to generate the FA weighted tracts. The 37 FA weighted WM tracts were used for feature extraction (Supplementary Table S2). 2.5. Feature extraction The GM tissues of 23 ROIs and the FA values of 37 WM tracts were used to calculate radiomic features. Ten first-order statistics radiomic features (Aerts et al., 2014) calculated from histogram were extracted from each ROI or tract of patients (Supplementary Table S3). The feature algorithms were implemented in Matlab 2014a (MathWorks, Natick, MA, USA). All features of 23 ROIs for each patient (23 × 10 = 230) were concatenated into a feature vector and were connected in parallel to form a GM feature matrix. All features of 37 WM tracts for each patient (37 × 10 = 370) were concatenated into a feature vector and were connected in parallel to form a WM feature matrix. The GM feature matrix and the WM feature matrix (230 + 370 = 600) were connected in parallel to form an overall GM and WM feature matrix, which were used for pattern classification. 2.6. Pattern recognition analysis and correlation analysis Pattern recognition analysis included three stages, each of which was performed with the GM feature matrix, the WM feature matrix,

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

4

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

Fig. 1. A flowchart for the data processing and analysis. 1. The modulated and normalized GM tissues were generated from T1-weighted images of each patient. 2. The construction of DTI from dMRI of each participant. The color-coded FA map represents the directions of first eigenvectors: red, left–right; green, anterior–posterior; blue, inferior–superior. 3. A realistic head model was conducted and ECT electrical field simulation was performed. 4. 23 brain regions were then defined as regions of interest (ROIs) with strong electrical field distribution. 5. Fiber pathways were performed using the FACT algorithm to reconstruct whole-brain tractography of each healthy subject and all tracts in the native space were transformed to the MNI space to build binary tract maps of tracts. 6. The binary maps of 37 WM tracts were constructed from healthy participants. 7. The GM tissues of 23 ROIs and the FA values of 37 WM tracts were used for feature extraction. Ten first-order statistics radiomics features were extracted from each ROI or tract of patients. The GM feature matrix and/or the WM feature matrix were used for pattern recognition analysis. 57 patients were randomly divided into the primary cohort (42 patients) and the test cohort (15 patients). At the first stage of analysis, feature selection was performed using LASSO based on the LOOCV framework in the primary cohort. At the second stage of analysis, model fitting using different feature subsets was performed based on a LOOCV framework in the primary cohort. At the last stage of analysis, the final SVR model was built using the most important features in the primary cohort and was tested in the test cohort. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

and the overall GM and WM fusion feature matrix respectively. To predict a continuous improvement in symptoms measured by the individual symptom relief according to the PANSS (ΔPANSS), a support vector regression (SVR) was conducted (Redlich et al., 2016). The SVR was applied with linear kernel and default parameter using a leave-onesubject-out cross-validation (LOOCV) in the primary cohort and was tested in the test cohort. In LOOCV, one patient is used as a testing sample, and the remaining patients are applied as training samples to select features and build the model. 2.6.1. Feature selection At the first stage of analysis, feature selection was performed based on the LOOCV framework in the primary cohort. The dimensions of the candidate features were greater than the sample size, suggesting that there were correlations among some features. Therefore, the least absolute shrinkage and selection operator (LASSO) method, commonly used for dimensional reduction, was conducted for feature selection to reduce the feature dimensions and reduce collinearity (Tibshirani, 1996). LASSO estimates the regression coefficients through an l1norm penalized least-squares criterion and minimizes the residual sum of squares with an l1 penalty on the regression coefficients (Waldmann et al., 2013). In this method, the regularization parameter λ controls the trade-off between data fitting and sparsity. The optimal value of λ was determined by a 10-fold cross-validation. Features with

non-zero regression coefficients were then selected as the most promising candidate features. Feature importance was determined by the number of times (frequency) each feature was selected by the feature selection method across 42 training sessions in the cross-validation setup. Features were subsequently ranked according to their predictive power as determined by their frequency (Dwyer et al., 2018). Ranked features were progressively added one by one from most to least important to create different feature subsets (Ortiz-Ramon et al., 2018). In other words, the two most important features were the first feature subset, the three most important features were the second feature subset, and so on.

2.6.2. Model fitting At the second stage of analysis, model fitting using different feature subsets was performed based on a LOOCV framework in the primary cohort. The training samples in LOOCV were used to build a SVR which was applied to predict the testing sample. The Pearson's correlation coefficient (r) between predicted and actual value, and the root-meansquare error (RMSE) were employed as measures of goodness-of-fit. When r = 1 and RMSE = 0, the model is the most accurate predictive model. Features of the feature subset that optimized model performance were the most important features and were used for subsequent analysis.

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

2.6.3. Model building and testing At the last stage of analysis, the final SVR model was built using the most important features in the primary cohort and was tested in the test cohort. Prediction performance of the SVR model was estimated using r and RMSE based on the LOOCV framework in the primary and was tested in the test cohort. 2.6.4. Correlation analysis Pearson correlation analysis was performed to examine the correlation between the important features in the final linear regression model and treatment response in all data samples. The significance threshold was set at p b 0.05 (false discovery rate [FDR]-corrected (Benjamini and Hochberg, 1995)). 3. Results 3.1. Patient characteristics Table 2 shows demographic and clinical characteristics of included patients in the primary cohort and the test cohort. No significant difference was found between the two cohorts. There were significant differences between pre- and post-treatment scores for the Clinical Global Impression (CGI) and PANSS (including positive, negative, general psychopathology, and total scores) in both cohorts. There were significant reductions in all scores in both cohorts, indicating overall relief of symptoms. There was no significant change in MMSE score after treatment in both cohorts, which is consistent with previous studies (Petrides et al., 2015; Rami et al., 2004). 3.2. Pattern recognition analysis using the GM feature matrix In the first two stages, all features were assessed for importance by calculating the frequency at which they were selected and they were

Table 2 Clinical data of patients in the primary and the test cohorts before and after ECT. Mean and SD are given except when stated otherwise. Characteristic

Primary cohort (n = Test cohort (n = 42) 15)

p valuesa

Age (y) Gender (M/F) Education level (y) Duration of illness (y) CGIc score at baseline CGI score after ECT MMSEe at baseline MMSE after ECT PANSSf score at baseline Positive score Negative score General score Total score PANSS score after ECT Positive score Negative score General score Total score Changes in PANSS score Number of ECT Antipsychotic dose (mg/d)g

30.1 ± 8.8 29/13 11.5 ± 4.0 5.6 ± 5.5 5.6 ± 0.9 3.1 ± 1.1 (b0.001)d 23.1 ± 10.2 24.7 ± 8.7 (0.21)

30.8 ± 10.9 8/7 11.5 ± 2.9 5.8 ± 8.3 5.7 ± 0.7 3.2 ± 0.4 (b0.001) 20.6 ± 10.4 22.9 ± 10.1 (0.08)

0.82 0.27b 0.94 0.91 0.79 0.79 0.43 0.54

29.4 ± 6.7 24.3 ± 10.9 43.3 ± 9.0 97.0 ± 18.3

27.7 ± 8.7 24.5 ± 13.7 42.7 ± 15.2 95.0 ± 24.9

0.45 0.94 0.86 0.74

12.9 ± 4.4 (b0.001) 15.0 ± 7.8 (b0.001) 24.5 ± 7.4 (b0.001) 52.4 ± 17.0 (b0.001) 67.5% ± 21.5% 10.2 ± 2.7 15.8 ± 6.6

11.7 ± 2.8 (b0.001) 15.9 ± 8.9 (b0.001) 23.5 ± 5.8 (b0.001) 51.1 ± 14.3 (b0.001) 67.6% ± 17.6% 11.2 ± 1.3 15.2 ± 6.7

0.31 0.70 0.66 0.80 0.98 0.45 0.75

a p values were obtained using the unpaired 2-tailed t-test for comparison of clinical data (except gender) between the primary cohort and the test cohort. b p values were obtained using the chi-square. c CGI, Clinical Global Impression. d p values in parentheses were obtained using the paired 2-tailed t-test for comparison of clinical data between baseline and after treatment. e MMSE, mini-mental state examination. f PANSS, Positive And Negative Syndrome Scale. g Olanzapine equivalents based on defined daily doses method.

5

ranked to construct subsets for model fitting. The performance of different feature subsets was showed in the Supplementary Fig. S2. For the GM feature matrix, optimal performance was obtained with the model using the feature subset with the four highest frequency features (Supplementary Table S4, Fig. S2, Fig. 2A). The four features were used for the final stage of the analysis. The performance of the final SVR model was shown in Table 3 and Fig. 3A and B. In the primary cohort, RMSE was 18.460 and r value was 0.385 (p = 0.012). In the test cohort, RMSE was 20.535 and r value was 0.477 (p = 0.072). 3.3. Pattern recognition analysis using the WM feature matrix For the WM feature matrix, optimal performance was obtained with the model using the feature subset with the six highest frequency features (Supplementary Table S5, Fig. S2, Fig. 2B). The six features were used for the final stage of the analysis. The performance of the final SVR model was shown in Table 3 and Fig. 3C and D. In the primary cohort, RMSE was 15.676 and r value was 0.630 (p b 0.001). In the test cohort, RMSE was 16.470 and r value was 0.571 (p = 0.007). Performance of the model using the WM feature subset was better than performance of the model using the GM feature subset. 3.4. Pattern recognition analysis using the WM and GM fusion feature matrix For the overall GM and WM feature matrix, optimal performance was obtained with the model using the feature subset with the ten highest frequency features (Supplementary Table S6, Fig. S2, Fig. 2). The ten features were used for the final stage of the analysis. The performance of the final SVR model was shown in Table 3 and Fig. 3E and F. In the primary cohort, RMSE was 15.183 and r value was 0.671 (p b 0.001). In the test cohort, RMSE was 14.980 and r value was 0.777 (p b 0.001). Performance of the model using the WM and GM feature subset was better than performance of the model using the GM feature subset and performance of the model using the WM feature subset. 3.5. Classification performance evaluation The classification performance of the model was additionally evaluated. The method and the results were described in detail in the Supplementary A8 and Table S7. 3.6. Pearson correlation analysis Significant correlations were observed between all ten features of the final predictive model and treatment response (p b 0.05 [FDRcorrected], Table S6, Fig. S3). 4. Discussion In this study, we provide evidence for a set of potential baseline MRI biomarkers that predict response to ECT treatment in patients with schizophrenia. Using machine learning techniques, we identified radiomic features with sMRI and/or dMRI that successfully predicted response to ECT. The SVR model using 6 dMRI features performed better than did the SVR model using 4 sMRI features, while the multimodal linear regression model using 10 features (4 sMRI and 6 dMRI features) performed the best. Moreover, a significant association was observed between important features and treatment response. These results demonstrate that multi-parametric MRI-based radiomic features may predict response to ECT for individual patients, and have the potential to serve as prognostic neuroimaging biomarkers. Routine assessment with sMRI and dMRI before treatment could guide personalized treatment decisions for clinical psychiatrists. Our results suggest that a model using multimodal image features (dMRI and sMRI together) showed better prediction performance than

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

6

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

Fig. 2. Imaging schematic of the location of important features. A, ROIs where the four important features used in the final GM model are located, include Frontal_Inf_Orb_L (yellow region), Temporal_Sup_R (blue region), Temporal_Pole_Mid_L (green region) and Insula_R (red region). B, The 37 WM tracts connected ROIs (red nodes) and regions in AAL (blue nodes) were used for feature extraction and pattern classification analysis and 6 features from 5 tracts (4 yellow connections and 1 cyan connection) are important for prediction and used for the final WM model. Moreover, the features from Frontal_Inf_Orb_L (yellow region in part A), Temporal_Sup_R (blue region in part A), Temporal_Mid_L (pink region in part A) and Insula_R (red region in part A) and the 6 features from 5 tracts (4 yellow connections and 1 brown connection in part B) are used for the final fusion model. This figure was constructed using the BrainNet Viewer (http://www.nitrc.org/projects/bnv/) (Xia et al., 2013). Abbreviation: L: left; R: right; Frontal_Inf_Orb_L, left inferior frontal gyrus, pars orbitalis; Temporal_Sup_R, right superior temporal gyrus; Temporal_Pole_Mid_L, left middle temporal pole; Temporal_Mid_L, left middle temporal gyrus. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

did a model using unimodal image features (dMRI or sMRI alone). These results support the assertion that multimodal fusion could help build better predictive models, which is consistent with previous studied (Liu et al., 2017; Plis et al., 2018; Xi et al., 2018; Zhang et al., 2017). Moreover, radiomics analysis of combined multiparametric MRI sequences has higher accuracy because it can provide more information. The better performance of the multimodal fusion models provides more evidence for the advantages of multimodal fusion. The source of improved model performance in these studies is the identification by multimodal imaging of important variations that may only be partially detected by a single modality. Amin and colleagues considered that two different imaging views of the same brain like two different languages conveying some common facts (Plis et al., 2018). In the current study, sMRI provides GM neuroanatomical information and dMRI provides WM microstructural information on features associated with WM (Xie et al., 2015). The combination of these data into a multimodal

Table 3 Prediction performance of SVR models in the pattern recognition classification analysis. Matrix

Number of features

GM WM GM&WM

4 6 10

Primary cohort

Test cohort

RMSE

r

p

18.460 15.676 15.183

0.385 0.630 0.671

0.012 20.535 b0.001 16.470 b0.001 14.980

RMSE

r

p

0.477 0.571 0.777

0.072 0.007 b0.001

neurologic signature provides more comprehensive and richer information, providing a potential biomarker predicting individual response to ECT in patients with schizophrenia. The features of the inferior frontal gyrus, the temporal gyrus and the right insula were important for GM model and multimodal model, which means that these regions are particularly important for predicting the response to ECT. Temporal and frontal lobes have been reported as common locations of pathological brain aberrations affected by schizophrenia (Shenton et al., 2001; Shenton et al., 2010). The presence of gray matter (GM) volume abnormalities in schizophrenia patients, primarily the reductions in the frontal and temporal cortex, insula, etc., have been reported in numerous imaging studies (Honea et al., 2005; Shepherd et al., 2012; Wright et al., 2000; Wylie and Tregellas, 2010). These studies indicate that patients with schizophrenia have abnormal GM volumes of the frontal gyrus, the temporal gyrus and the insula, and that these regions are important in schizophrenia and have pathophysiological significance. Moreover, a neuroimaging study has reported that the inferior frontal gyrus and the temporal cortex might be associated with the efficacy of ECT (Oudega et al., 2014). Several previous studies have also illustrated a relationship between insula and ECT treatment response in patients with late-life depression and schizophrenia (Bouckaert et al., 2016; Jiang et al., 2019), suggesting that it may be a useful as a predictive biomarker. These studies suggest that the features of these regions may be predictive biomarkers for prediction of response to ECT, and this hypothesis is supported by this study.

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

7

Fig. 3. The results of Pearson correlation analysis between the predicted values of the models and the actual values in the primary and test cohorts. A and B, the Pearson's correlation coefficient (r) between the predicted values from GM model and the actual values is 0.385 (p = 0.012) in the primary cohort and 0.477 (p = 0.072) in the test cohort. C and D, the r between the predicted values from WM model and the actual values is 0.630 (p b 0.001) in the primary cohort and 0.571 (p = 0.007) in the test cohort, which is higher than the r of GM model. E and F, the r between the predicted values from the multimodal model and the actual values is 0.671 (p b 0.001) in the primary cohort and 0.777 (p b 0.001) in the test cohort, which is higher than the other two values of r. The higher the r value, the more accurate the model predicts.

The results also indicated that up to 6 dMRI features were used in the WM and fusion models, reflecting the importance of dMRI features for prediction. Schizophrenia has been considered as a disconnection syndrome (Andreasen et al., 1998; Davis et al., 2003; McGuire and Frith, 1996; Spalletta et al., 2003; Sullivan et al., 2003), with a deficit in WM integrity leading to defective communication among brain regions (Spoletini et al., 2009). Fractional anisotropy (FA) is usually interpreted as reflecting WM integrity (Assaf and Pasternak, 2008; Basser and Pierpaoli, 1996; Gomez-Pilar et al., 2018). FA computed from diffusion tensor imaging quantifies the degree to which water diffusion is directionally hindered (anisotropic) (Ashtari et al., 2007; BarneaGoraly et al., 2005; Ben Bashat et al., 2005; Spoletini et al., 2009). WM features can be calculated from FA of WM tracts, and in some sense the features also represent something about WM integrity. Lower FA has been observed across the whole brain in patients with schizophrenia (Ardekani et al., 2003; Lim et al., 1999; Minami et al., 2003), and particularly in frontal and temporal lobes in which reduced FA is associated

with clinical symptoms (Davidson and Heinrichs, 2003; Davis et al., 2003; Hyde et al., 1992; Shin et al., 2006). An meta-analytic study also found a widespread alteration of WM bundles in schizophrenia mostly included in frontal, temporal, and limbic pathways (Vitolo et al., 2017), which may be the most commonly involved bundles in schizophrenia and may represent the “core tracts” of the illness. Taken together, these studies suggest that there is an abnormality in WM integrity in patients with schizophrenia, especially the connections of frontal and temporal regions. Moreover, ECT appears to have an impact on distinct structural networks in schizophrenia (Wolf et al., 2016), providing a basis for potential prediction of response to ECT using WM features. In addition, the results of the current study indicate that WM features can predict response to ECT well, providing new evidence for the predictive power of dMRI. In the current study, we accurately predicted the quantitative efficacy of ECT for each patient by using pattern recognition analysis based on multi-parametric MRI-based radiomic features (r = 0.777, p

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

8

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

b 0.001). A previous study suggested that functional connectivity patterns are predictive for therapeutic outcomes in schizophrenia patients (Li et al., 2017). They found that the treatment response measured based on PANSS scores was negatively correlated with the baseline classification scores in the ECT group (total score, r = −0.75, p b 0.002). There are few studies on the prediction of the response to ECT in schizophrenia, but there are many studies in depression (Cao et al., 2018b; Moreno-Ortega et al., 2019; Pinna et al., 2018; Redlich et al., 2016; van Diermen et al., 2018). Cao and his colleagues applied a machine learning algorithm to the hippocampal subfield volumes at baseline and were able to predict the quantitative efficacy of ECT for each patient in depressive symptoms (n = 24, r = 0.81) (Cao et al., 2018b). In the nonrandomized prospective study of patients with severe depression, Redlich and his colleagues achieved a successful prediction of ECT response, with accuracy rates as great as 78.3%, using structural MRI obtained before therapy. The SVR results were significantly associated with continuous symptom relief using the whole ECT sample (n = 23, r = 0.67, p b 0.001) (Redlich et al., 2016). In the present study, we used multi-parametric MRI consisting of dMRI and sMRI for quantitative predictive analysis based on a larger sample size. Multimodal fusion could help build better predictive models. Patients with partial responses are classified as non-responders, which affects the classification results (Redlich et al., 2016). Support vector regression might be a suitable approach to overcome this issue because it provides a continuous prediction of symptom improvement (Redlich et al., 2016). An independent test group was included in this study, which increased the robustness and generalization capability of the results. Despite these encouraging results, some limitations must be acknowledged. First, patients have a variable duration of illness owing to the naturalistic study design. Given the established effects of illness chronicity on brain structure (Bora et al., 2011; van Erp et al., 2016; van Erp et al., 2018), this may have an impact on the results. In future research, investigators should control for chronicity to more accurately investigate how MRI biomarkers predict ECT efficacy in patients with schizophrenia. Second, fMRI was not measured in this study. fMRI measures dynamic changes in the hemodynamic response related to neural activity. Several neuroimaging studies have indicated that the positive effects of ECT in patients with schizophrenia may relate to modulation of both aberrant brain function and structure (Gan et al., 2017; Jiang et al., 2019; Li et al., 2017; Wolf et al., 2016). Therefore, it would be of value to consider both function and structure in future research. Moreover, it is necessary to assess the duration of response to ECT, but the follow-up data was not collected in this study. Finally, radiogenomics, which focuses on the relationship between MRI and genomics, has emerged in the field of cancer research and attracted increasing interest. However, there has been little radiogenomic research in the field of schizophrenia (Kerns et al., 2014; Krajewski and Pedrosa, 2018). This might provide an interesting and worthwhile approach to understanding treatment response in schizophrenia.

5. Conclusions In conclusion, our results demonstrate that multi-parametric MRIbased radiomic features may serve as alternative prognostic neuroimaging biomarkers to predict response to ECT for individual patients with schizophrenia. These findings may be helpful for clinicians to determine the extent of response to ECT in schizophrenia and help them make treatment decisions.

Contributors W Qin, YB Xi and H Yin designed the study. P Li, J Jia, P Liu, JB Sun and LB Cui acquired the data, which J Gong, YS Zhao, LB Cui and ZL Xu analyzed. J Gong and XJ Yang wrote the article, which all authors reviewed. All authors approved the final version to be published and can certify that no other individuals not listed as authors have made substantial contributions to the paper.

Declaration of competing interest The authors have declared that there are no conflicts of interest in relation to the subject of this study.

Acknowledgments We thank Liwen Bianji, Edanz Group China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.

Funding This work was financially supported by National Basic Research Program of China [grant number 2015CB856403 and 2014CB543203], the Science and Technology Projects of Xi'an, China [grant number 201809170CX11JC12], and the National Natural Science Foundation of China [grant number 81771918, 81801675, 81471811 and 81471738].

Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.schres.2019.11.046.

References Aerts, H.J.W.L., Velazquez, E.R., Leijenaar, R.T.H., Parmar, C., Grossmann, P., Cavalho, S., Bussink, J., Monshouwer, R., Haibe-Kains, B., Rietveld, D., Hoebers, F., Rietbergen, M.M., Leemans, C.R., Dekker, A., Quackenbush, J., Gillies, R.J., Lambin, P., 2014. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat. Commun. 5. Ahmed, S., Khan, A.M., Mekala, H.M., Venigalla, H., Ahmed, R., Etman, A., Esang, M., Qureshi, M., 2017. Combined use of electroconvulsive therapy and antipsychotics (both clozapine and non-clozapine) in treatment resistant schizophrenia: a comparative meta-analysis. Heliyon 3 (11), e00429. Andreasen, N.C., Paradiso, S., O'Leary, D.S., 1998. “Cognitive dysmetria” as an integrative theory of schizophrenia: a dysfunction in cortical subcortical-cerebellar circuitry? Schizophr. Bull. 24 (2), 203–218. Ardekani, B.A., Nierenberg, J., Hoptman, M.J., Javitt, D.C., Lim, K.O., 2003. MRI study of white matter diffusion anisotropy in schizophrenia. Neuroreport 14 (16), 2025–2029. Ashburner, J., Friston, K.J., 2000. Voxel-based morphometry - the methods. Neuroimage 11 (6), 805–821. Ashtari, M., Cervellione, K.L., Hasan, K.M., Wu, J., McIlree, C., Kester, H., Ardekani, B.A., Roofeh, D., Szeszko, P.R., Kumra, S., 2007. White matter development during late adolescence in healthy males: a cross-sectional diffusion tensor imaging study. Neuroimage 35 (2), 501–510. Assaf, Y., Pasternak, O., 2008. Diffusion tensor imaging (DTI)-based white matter mapping in brain research: a review. J. Mol. Neurosci. 34 (1), 51–61. Barnea-Goraly, N., Menon, V., Eckert, M., Tamm, L., Bammer, R., Karchemskiy, A., Dant, C.C., Reiss, A.L., 2005. White matter development during childhood and adolescence: a cross-sectional diffusion tensor imaging study. Cereb. Cortex 15 (12), 1848–1854. Basser, P.J., Pierpaoli, C., 1996. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 111 (3), 209–219. Bellec, P., Lavoie-Courchesne, S., Dickinson, P., Lerch, J.P., Zijdenbos, A.P., Evans, A.C., 2012. The pipeline system for Octave and Mat lab (PSOM): a lightweight scripting framework and execution engine for scientific workflows. Frontiers in Neuroinformatics 6. Ben Bashat, D., Ben Sira, L., Graif, M., Pianka, P., Hendler, T., Cohen, Y., Assaf, Y., 2005. Normal white matter development from infancy to adulthood: comparing diffusion tensor and high b value diffusion weighted MR images. J. Magn. Reson. Imaging 21 (5), 503–511. Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 12. Birkhahn, M., Mitra, A.P., Cote, R.J., 2007. Molecular markers for bladder cancer: the road to a multimarker approach. Expert Rev Anticanc 7 (12), 1717–1727. Bora, E., Fornito, A., Radua, J., Walterfang, M., Seal, M., Wood, S.J., Yucel, M., Velakoulis, D., Pantelis, C., 2011. Neuroanatomical abnormalities in schizophrenia: a multimodal voxelwise meta-analysis and meta-regression analysis. Schizophr. Res. 127 (1–3), 46–57. Bouckaert, F., De Winter, F.L., Emsell, L., Dols, A., Rhebergen, D., Wampers, M., Sunaert, S., Stek, M., Sienaert, P., Vandenbulcke, M., 2016. Grey matter volume increase following electroconvulsive therapy in patients with late life depression: a longitudinal MRI study. J. Psychiatry Neurosci. 41 (2), 105–114. Brugger, S.P., Howes, O.D., 2017. Heterogeneity and homogeneity of regional brain structure in schizophrenia a meta-analysis. Jama Psychiatry 74 (11), 1104–1111. Calhoun, V.D., Sui, J., 2016. Multimodal fusion of brain imaging data: a key to finding the missing link(s) in complex mental illness. Biol Psychiatry Cogn Neurosci Neuroimaging 1 (3), 230–244. Cao, B., Cho, R.Y., Chen, D., Xiu, M., Wang, L., Soares, J.C., Zhang, X.Y., 2018a. Treatment response prediction and individualized identification of first-episode drug-naive schizophrenia using brain functional connectivity. Mol. Psychiatry https://doi.org/ 10.1038/s41380-018-0106-5 [Epub ahead of print]. Cao, B., Luo, Q., Fu, Y., Du, L., Qiu, T., Yang, X., Chen, X., Chen, Q., Soares, J.C., Cho, R.Y., Zhang, X.Y., Qiu, H., 2018b. Predicting individual responses to the electroconvulsive

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx therapy with hippocampal subfield volumes in major depression disorder. Sci. Rep. 8 (1), 5434. Case, B.G., Bertollo, D.N., Laska, E.M., Price, L.H., Siegel, C.E., Olfson, M., Marcus, S.C., 2013. Declining use of electroconvulsive therapy in United States general hospitals. Biol. Psychiatry 73 (2), 119–126. Chaddad, A., Desrosiers, C., Hassan, L., Tanougast, C., 2017. Hippocampus and amygdala radiomic biomarkers for the study of autism spectrum disorder. BMC Neurosci. 18. Chinese Association of Neurological Regulation Committee for Electroconvulsive Therapy and Nerve Stimulation, Chinese Association of Sleep Committee for Mental Psychology, Chinese Association of Anesthesiology (Eds.), 2019. Expert consensus on modified electroconvulsive therapy. Transl Med J 6. Chinese Medical Doctor Association ECT and Neurostimulation Specialized Committee (Ed.), 2017. Expert consensus on methods of electroconvulsive therapy. Clin Focus 4. Chinese Society of Psychiatry (Ed.), 2015. Chinese Guidelines for Prevention and Treatment of Schizophrenia, 2nd ed. Chinese Medical Multimedia Press, Beijing. Croner, R.S., Fortsch, T., Bruckl, W.M., Rodel, F., Rodel, C., Papadopoulos, T., Brabletz, T., Kirchner, T., Sachs, M., Behrens, J., Klein-Hitpass, L., Stuerzl, M., Hohenberger, W., Lausen, B., 2008. Molecular signature for lymphatic metastasis in colorectal carcinomas. Ann. Surg. 247 (5), 803–810. Crow, T.J., 1998. Schizophrenia as a transcallosal misconnection syndrome. Schizophr. Res. 30 (2), 111–114. Cui, Z.X., Zhong, S.Y., Xu, P.F., He, Y., Gong, G.L., 2013. PANDA: a pipeline toolbox for analyzing brain diffusion images. Front. Hum. Neurosci. 7. Cui, L.B., Liu, L., Wang, H.N., Wang, L.X., Guo, F., Xi, Y.B., Liu, T.T., Li, C., Tian, P., Liu, K., Wu, W.J., Chen, Y.H., Qin, W., Yin, H., 2018. Disease definition for schizophrenia by functional connectivity using radiomics strategy. Schizophr. Bull. 44 (5), 1053–1059. Davidson, L.L., Heinrichs, R.W., 2003. Quantification of frontal and temporal lobe brainimaging findings in schizophrenia: a meta-analysis. Psychiat Res-Neuroim 122 (2), 69–87. Davis, K.L., Stewart, D.G., Friedman, J.I., Buchsbaum, M., Harvey, P.D., Hof, P.R., Buxbaum, J., Haroutunian, V., 2003. White matter changes in schizophrenia - evidence for myelinrelated dysfunction. Arch. Gen. Psychiatry 60 (5), 443–456. Dietsche, B., Kircher, T., Falkenberg, I., 2017. Structural brain changes in schizophrenia at different stages of the illness: a selective review of longitudinal magnetic resonance imaging studies. Aust Nz J Psychiat 51 (5), 500–508. Dwyer, D.B., Cabral, C., Kambeitz-Ilankovic, L., Sanfelici, R., Kambeitz, J., Calhoun, V., Falkai, P., Pantelis, C., Meisenzahl, E., Koutsouleris, N., 2018. Brain subtyping enhances the neuroanatomical discrimination of schizophrenia. Schizophr. Bull. 44 (5), 1060–1069. Ellison-Wright, I., Bullmore, E.T., 2009. Meta-analysis of diffusion tensor imaging studies in schizophrenia. Schizophr. Res. 108 (1–3), 3–10. Freedman, R., 2003. Schizophrenia. N. Engl. J. Med. 349 (18), 1738–1749. Gan, J.L., Duan, H.F., Cheng, Z.X., Yang, J.M., Zhu, X.Q., Gao, C.Y., Zhao, L.M., Liang, X.J., 2017. Neuroprotective effect of modified electroconvulsive therapy for schizophrenia a proton magnetic resonance spectroscopy study. J. Nerv. Ment. Dis. 205 (6), 480–486. Gevaert, O., Xu, J., Hoang, C.D., Leung, A.N., Xu, Y., Quon, A., Rubin, D.L., Napel, S., Plevritis, S.K., 2012. Non-small cell lung cancer: identifying prognostic imaging biomarkers by leveraging public gene expression microarray data–methods and preliminary results. Radiology 264 (2), 387–396. Gillies, R.J., Kinahan, P.E., Hricak, H., 2016. Radiomics: images are more than pictures, they are data. Radiology 278 (2), 563–577. Gomez-Pilar, J., de Luis-Garcia, R., Lubeiro, A., de la Red, H., Poza, J., Nunez, P., Hornero, R., Molina, V., 2018. Relations between structural and EEG-based graph metrics in healthy controls and schizophrenia patients. Hum. Brain Mapp. 39 (8), 3152–3165. Goodman, W.K., 2011. Electroconvulsive therapy in the spotlight. N. Engl. J. Med. 364 (19), 1785–1787. Honea, R., Crow, T.J., Passingham, D., Mackay, C.E., 2005. Regional deficits in brain volume in schizophrenia: a meta-analysis of voxel-based morphometry studies. Am J Psychiat 162 (12), 2233–2245. Hyde, T.M., Ziegler, J.C., Weinberger, D.R., 1992. Psychiatric disturbances in metachromatic leukodystrophy. Insights into the neurobiology of psychosis. Arch. Neurol. 49 (4), 401–406. Iniesta, R., Stahl, D., McGuffin, P., 2016. Machine learning, statistical learning and the future of biological research in psychiatry. Psychol. Med. 46 (12), 2455–2465. Jiang, Y.C., Xia, M.Q., Li, X.K., Tang, Y.Y., Huan, C.B., Huang, H., Dong, D.B., Jiang, S.S., Wang, J.J., Xu, J., Luo, C., Yao, D.Z., 2019. Insular changes induced by electroconvulsive therapy response to symptom improvements in schizophrenia. Prog. NeuroPsychopharmacol. Biol. Psychiatry 89, 254–262. Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr. Bull. 13 (2), 261–276. Kerns, S.L., Ostrer, H., Rosenstein, B.S., 2014. Radiogenomics: using genetics to identify cancer patients at risk for development of adverse effects following radiotherapy. Cancer Discovery 4 (2), 155–165. Krajewski, K.M., Pedrosa, I., 2018. Imaging advances in the management of kidney cancer. J. Clin. Oncol. 36 (36), 3582 (-+). Kuo, M.D., Jamshidi, N., 2014. Behind the numbers: decoding molecular phenotypes with radiogenomics—guiding principles and technical considerations. Radiology 270 (2), 320–325. Lally, J., Tully, J., Robertson, D., Stubbs, B., Gaughran, F., MacCabe, J.H., 2016. Augmentation of clozapine with electroconvulsive therapy in treatment resistant schizophrenia: a systematic review and meta-analysis. Schizophr. Res. 171 (1–3), 215–224. Leaver, A.M., Wade, B., Vasavada, M., Hellemann, G., Joshi, S.H., Espinoza, R., Narr, K.L., 2018. Fronto-temporal connectivity predicts ECT outcome in major depression. Frontiers in Psychiatry 9. Lehman, A.F., Lieberman, J.A., Dixon, L.B., McGlashan, T.H., Miller, A.L., Perkins, D.O., Kreyenbuhl, J., American Psychiatric, A., Steering Committee on Practice, G, 2004.

9

Practice guideline for the treatment of patients with schizophrenia, second edition. Am. J. Psychiatry 161 (2 Suppl), 1–56. Lesh, T.A., Tanase, C., Geib, B.R., Niendam, T.A., Yoon, J.H., Minzenberg, M.J., Ragland, J.D., Solomon, M., Carter, C.S., 2015. A multimodal analysis of antipsychotic effects on brain structure and function in first-episode schizophrenia. Jama Psychiatry 72 (3), 226–234. Li, P., Jing, R.X., Zhao, R.J., Ding, Z.B., Shi, L., Sun, H.Q., Lin, X., Fan, T.T., Dong, W.T., Fan, Y., Lu, L., 2017. Electroconvulsive therapy-induced brain functional connectivity predicts therapeutic efficacy in patients with schizophrenia: a multivariate pattern recognition study. NPJ Schizophr. 3, 21. Liang, S., Li, Y., Zhang, Z., Kong, X., Wang, Q., Deng, W., Li, X., Zhao, L., Li, M., Meng, Y., Huang, F., Ma, X., Li, X.M., Greenshaw, A.J., Shao, J., Li, T., 2019. Classification of firstepisode schizophrenia using multimodal brain features: a combined structural and diffusion imaging study. Schizophr. Bull. 45 (3), 591–999. https://doi.org/10.1093/ schbul/sby091. Lieberman, J.A., Stroup, T.S., McEvoy, J.P., Swartz, M.S., Rosenheck, R.A., Perkins, D.O., Keefe, R.S.E., Davis, S.M., Davis, C.E., Lebowitz, B.D., Severe, J., Hsiao, J.K., Investigators, C., 2005. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. N. Engl. J. Med. 353 (12), 1209–1223. Lim, K.O., Hedehus, M., Moseley, M., de Crespigny, A., Sullivan, E.V., Pfefferbaum, A., 1999. Compromised white matter tract integrity in schizophrenia inferred from diffusion tensor imaging. Arch. Gen. Psychiatry 56 (4), 367–374. Liu, Z.Y., Zhang, X.Y., Shi, Y.J., Wang, L., Zhu, H.T., Tang, Z.C., Wang, S., Li, X.T., Tian, J., Sun, Y.S., 2017. Radiomics analysis for evaluation of pathological complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer. Clin. Cancer Res. 23 (23), 7253–7262. Liu, S.F., Wang, H.Y., Song, M., Lv, L.X., Cui, Y., Liu, Y., Fan, L.Z., Zuo, N.M., Xu, K.B., Du, Y.H., Yu, Q.B., Luo, N., Qi, S.L., Yang, J., Xie, S.M., Li, J., Chen, J., Chen, Y.C., Wang, H.N., Guo, H., Wan, P., Yang, Y.F., Li, P., Lu, L., Yan, H., Yan, J., Wang, H.L., Zhang, H.X., Zhang, D., Calhoun, V.D., Jiang, T.Z., Sui, J., 2019. Linked 4-way multimodal brain differences in schizophrenia in a large Chinese Han population. Schizophr. Bull. 45 (2), 436–449. Mazurowski, M.A., 2015. Radiogenomics: what it is and why it is important. J. Am. Coll. Radiol. 12 (8), 862–866. McGuire, P.K., Frith, C.D., 1996. Disordered functional connectivity in schizophrenia. Psychol. Med. 26 (4), 663–667. Minami, T., Nobuhara, K., Okugawa, G., Takase, K., Yoshida, T., Sawada, S., Ha-Kawa, S., Ikeda, K., Kinoshita, T., 2003. Diffusion tensor magnetic resonance imaging of disruption of regional white matter in schizophrenia. Neuropsychobiology 47 (3), 141–145. Moreno-Ortega, M., Prudic, J., Rowny, S., Patel, G.H., Kangarlu, A., Lee, S., Grinband, J., Palomo, T., Perera, T., Glasser, M.F., Javitt, D.C., 2019. Resting state functional connectivity predictors of treatment response to electroconvulsive therapy in depression. Sci. Rep. 9 (1), 5071. Obermeier, M., Mayr, A., Schennach-Wolff, R., Seemuller, F., Moller, H.J., Riedel, M., 2010. Should the PANSS be rescaled? Schizophr. Bull. 36 (3), 455–460. Orru, G., Pettersson-Yeo, W., Marquand, A.F., Sartori, G., Mechelli, A., 2012. Using Support Vector Machine to identify imaging biomarkers of neurological and psychiatric disease: a critical review. Neurosci. Biobehav. Rev. 36 (4), 1140–1152. Ortiz-Ramon, R., Larroza, A., Ruiz-Espana, S., Arana, E., Moratal, D., 2018. Classifying brain metastases by their primary site of origin using a radiomics approach based on texture analysis: a feasibility study. Eur. Radiol. 28 (11), 4514–4523. Oudega, M.L., van Exel, E., Stek, M.L., Wattjes, M.P., van der Flier, W.M., Comijs, H.C., Dols, A., Scheltens, P., Barkhof, F., Eikelenboom, P., van den Heuvel, O.A., 2014. The structure of the geriatric depressed brain and response to electroconvulsive therapy. Psychiat Res-Neuroim 222 (1–2), 1–9. Petrides, G., Malur, C., Braga, R.J., Bailine, S.H., Schooler, N.R., Malhotra, A.K., Kane, J.M., Sanghani, S., Goldberg, T.E., John, M., Mendelowitz, A., 2015. Electroconvulsive therapy augmentation in clozapine-resistant schizophrenia: a prospective, randomized study. Am J Psychiat 172 (1), 52–58. Pinna, M., Manchia, M., Oppo, R., Scano, F., Pillai, G., Loche, A.P., Salis, P., Minnai, G.P., 2018. Clinical and biological predictors of response to electroconvulsive therapy (ECT): a review. Neurosci. Lett. 669, 32–42. Plis, S.M., Amin, M.F., Chekroud, A., Hjelm, D., Damaraju, E., Lee, H.J., Bustillo, J.R., Cho, K., Pearlson, G.D., Calhoun, V.D., 2018. Reading the (functional) writing on the (structural) wall: multimodal fusion of brain structure and function via a deep neural network based translation approach reveals novel impairments in schizophrenia. Neuroimage 181, 734–747. Prudic, J., Olfson, M., Marcus, S.C., Fuller, R.B., Sackeim, H.A., 2004. Effectiveness of electroconvulsive therapy in community settings. Biol. Psychiatry 55 (3), 301–312. Rami, L., Bernardo, M., Valdes, M., Boget, T., Portella, M.J., Ferrer, J., Salamero, M., 2004. Absence of additional cognitive impairment in schizophrenia patients during maintenance electroconvulsive therapy. Schizophr. Bull. 30 (1), 185–189. Redlich, R., Almeida, J.J.R., Grotegerd, D., Opel, N., Kugel, H., Heindel, W., Arolt, V., Phillips, M.L., Dannlowski, U., 2014. Brain morphometric biomarkers distinguishing unipolar and bipolar depression a voxel-based morphometry-pattern classification approach. Jama Psychiatry 71 (11), 1222–1230. Redlich, R., Opel, N., Grotegerd, D., Dohm, K., Zaremba, D., Burger, C., Munker, S., Muhlmann, L., Wahl, P., Heindel, W., Arolt, V., Alferink, J., Zwanzger, P., Zavorotnyy, M., Kugel, H., Dannlowski, U., 2016. Prediction of individual response to electroconvulsive therapy via machine learning on structural magnetic resonance imaging data. Jama Psychiatry 73 (6), 557–564. Rorden, C., Karnath, H.O., Bonilha, L., 2007. Improving lesion-symptom mapping. J. Cogn. Neurosci. 19 (7), 1081–1088. Sackeim, H.A., Prudic, J., Fuller, R., Keilp, J., Lavori, P.W., Olfson, M., 2007. The cognitive effects of electroconvulsive therapy in community settings. Neuropsychopharmacol 32 (1), 244–254.

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046

10

J. Gong et al. / Schizophrenia Research xxx (xxxx) xxx

Sanghani, S.N., Petrides, G., Kellner, C.H., 2018. Electroconvulsive therapy (ECT) in schizophrenia: a review of recent literature. Current Opinion in Psychiatry 31 (3), 213–222. Schultz, C.C., Fusar-Poli, P., Wagner, G., Koch, K., Schachtzabel, C., Gruber, O., Sauer, H., Schlosser, R.G.M., 2012. Multimodal functional and structural imaging investigations in psychosis research. Eur. Arch. Psychiatry Clin. Neurosci. 262, S97–S106. Sharma, V., 2001. The effect of electroconvulsive therapy on suicide risk in patients with mood disorders. Can. J. Psychiatr. 46 (8), 704–709. Sharma, V., 2003. Atypical antipsychotics and suicide in mood and anxiety disorders. Bipolar Disord. 5 (Suppl. 2), 48–52. Shenton, M.E., Dickey, C.C., Frumin, M., McCarley, R.W., 2001. A review of MRI findings in schizophrenia. Schizophr. Res. 49 (1–2), 1–52. Shenton, M.E., Whitford, T.J., Kubicki, M., 2010. Structural neuroimaging in schizophrenia: from methods to insights to treatments. Dialogues Clin. Neurosci. 12 (3), 317–332. Shepherd, A.M., Matheson, S.L., Laurens, K.R., Carr, V.J., Green, M.J., 2012. Systematic metaanalysis of insula volume in schizophrenia. Biol. Psychiatry 72 (9), 775–784. Shin, Y.W., Kwon, J.S., Ha, T.H., Park, H.J., Kim, D.J., Hong, S.B., Moon, W.J., Lee, J.M., Kim, I.Y., Kim, S.I., Chung, E.C., 2006. Increased water diffusivity in the frontal and temporal cortices of schizophrenic patients. Neuroimage 30 (4), 1285–1291. Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E.J., Johansen-Berg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y.Y., De Stefano, N., Brady, J.M., Matthews, P.M., 2004. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 23, S208–S219. Spalletta, G., Tomaiuolo, F., Marino, V., Bonaviri, G., Trequattrini, A., Caltagirone, C., 2003. Chronic schizophrenia as a brain misconnection syndrome: a white matter voxelbased morphometry study. Schizophr. Res. 64 (1), 15–23. Spoletini, I., Cherubini, A., Di Paola, M., Banfi, G., Rusch, N., Martinotti, G., Bria, P., Rubino, I.A., Siracusano, A., Caltagirone, C., Spalletta, G., 2009. Reduced fronto-temporal connectivity is associated with frontal gray matter density reduction and neuropsychological deficit in schizophrenia. Schizophr. Res. 108 (1–3), 57–68. Sullivan, P.F., Kendler, K.S., Neale, M.C., 2003. Schizophrenia as a complex trait - evidence from a meta-analysis of twin studies. Arch. Gen. Psychiatry 60 (12), 1187–1192. ten Doesschate, F., van Eijndhoven, P., Tendolkar, I., van Wingen, G.A., van Waarde, J.A., 2014. Pre-treatment amygdala volume predicts electroconvulsive therapy response. Frontiers in Psychiatry 5. Tibshirani, R., 1996. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. 58 (1), 267–288. van Diermen, L., van den Ameele, S., Kamperman, A.M., Sabbe, B.C.G., Vermeulen, T., Schrijvers, D., Birkenhager, T.K., 2018. Prediction of electroconvulsive therapy response and remission in major depression: meta-analysis. Br. J. Psychiatry 212 (2), 71–80. van Erp, T.G.M., Hibar, D.P., Rasmussen, J.M., Glahn, D.C., Pearlson, G.D., Andreassen, O.A., Agartz, I., Westlye, L.T., Haukvik, U.K., Dale, A.M., Melle, I., Hartberg, C.B., Gruber, O., Kraemer, B., Zilles, D., Donohoe, G., Kelly, S., McDonald, C., Morris, D.W., Cannon, D.M., Corvin, A., Machielsen, M.W.J., Koenders, L., de Haan, L., Veltman, D.J., Satterthwaite, T.D., Wolf, D.H., Gur, R.C., Gur, R.E., Potkin, S.G., Mathalon, D.H., Mueller, B.A., Preda, A., Macciardi, F., Ehrlich, S., Walton, E., Hass, J., Calhoun, V.D., Bockholt, H.J., Sponheim, S.R., Shoemaker, J.M., van Haren, N.E.M., Pol, H.E.H., Ophoff, R.A., Kahn, R.S., Roiz-Santianez, R., Crespo-Facorro, B., Wang, L., Alpert, K.I., Jonsson, E.G., Dimitrova, R., Bois, C., Whalley, H.C., McIntosh, A.M., Lawrie, S.M., Hashimoto, R., Thompson, P.M., Turner, J.A., Grp, E.S.W., 2016. Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium. Mol. Psychiatry 21 (4), 547–553. van Erp, T.G.M., Walton, E., Hibar, D.P., Schmaal, L., Jiang, W.H., Glahn, D.C., Pearlson, G.D., Yao, N.L., Fukunaga, M., Hashimoto, R., Okada, N., Yamamori, H., Bustillo, J.R., Clark, V.P., Agartz, I., Mueller, B.A., Cahn, W., de Zwarte, S.M.C., Pol, H.E.H., Kahn, R.S., Ophoff, R.A., van Haren, N.E.M., Andreassen, O.A., Dale, A.M., Doan, N.T., Gurholt, T.P., Hartberg, C.B., Haukvik, U.K., Jorgensen, K.N., Lagerberg, T.V., Melle, I., Westlye, L.T., Gruber, O., Kraemer, B., Richter, A., Zilles, D., Calhoun, V.D., Crespo-Facorro, B., Roiz-Santianez, R., Tordesillas-Gutierrez, D., Loughland, C., Carr, V.J., Catts, S., Cropley, V.L., Fullerton, J.M., Green, M.J., Henskens, F.A., Jablensky, A., Lenroot, R.K., Mowry, B.J., Michie, P.T., Pantelis, C., Quide, Y., Schall, U., Scott, R.J., Cairns, M.J., Seal, M., Tooney, P.A., Rasser, P.E., Cooper, G., Weickert, C.S., Weickert, T.W., Morris, D.W., Hong, E., Kochunov, P., Beard, L.M., Gur, R.E., Gur, R.C., Satterthwaite, T.D., Wolf, D.H., Belger, A., Brown, G.G., Ford, J.M., Macciardi, F., Mathalon, D.H., O'Leary, D.S., Potkin, S.G., Preda, A., Voyvodic, J., Lim, K.O., McEwen, S., Yang, F.D., Tan, Y.L., Tan, S.P., Wang, Z.R., Fan, F.M., Chen, J.X., Xiang, H., Tang, S.Y., Guo, H., Wan, P., Wei, D.,

Bockholt, H.J., Ehrlich, S., Wolthusen, R.P.F., King, M.D., Shoemaker, J.M., Sponheim, S.R., De Haan, L., Koenders, L., Machielsen, M.W., van Amelsvoort, T., Veltman, D.J., Assogna, F., Banaj, N., de Rossi, P., Iorio, M., Piras, F., Spalletta, G., McKenna, P.J., Pomarol-Clotet, E., Salvador, R., Corvin, A., Donohoe, G., Kelly, S., Whelan, C.D., Dickie, E.W., Rotenberg, D., Voineskos, A.N., Ciufolini, S., Radua, J., Dazzan, P., Murray, R., Marques, T.R., Simmons, A., Borgwardt, S., Egloff, L., Harrisberger, F., Riecher-Rossler, A., Smieskova, R., Alpert, K.I., Wang, L., Jonsson, E.G., Koops, S., Sommer, I.E.C., Bertolino, A., Bonvino, A., Di Giorgio, A., Neilson, E., Mayer, A.R., Stephen, J.M., Kwon, J.S., Yun, J.Y., Cannon, D.M., McDonald, C., Lebedeva, I., Tomyshev, A.S., Akhadov, T., Kaleda, V., Fatouros-Bergman, H., Flyckt, L., Busatto, G.F., Rosa, P.G.P., Serpa, M.H., Zanetti, M.V., Hoschl, C., Skoch, A., Spaniel, F., Tomecek, D., Hagenaars, S.P., McIntosh, A.M., Whalley, H.C., Lawrie, S.M., Knochel, C., Oertel-Knochel, V., Stablein, M., Howells, F.M., Stein, D.J., Temmingh, H.S., Uhlmann, A., Lopez-Jaramillo, C., Dima, D., McMahon, A., Faskowitz, J.I., Gutman, B.A., Jahanshad, N., Thompson, P.M., Turner, J.A., Project, K.S, 2018. Cortical brain abnormalities in 4474 individuals with schizophrenia and 5098 control subjects via the Enhancing Neuro Imaging Genetics Through Meta Analysis (ENIGMA) Consortium. Biol. Psychiatry 84 (9), 644–654. van Waarde, J.A., Scholte, H.S., van Oudheusden, L.J.B., Verwey, B., Denys, D., van Wingen, G.A., 2015. A functional MRI marker may predict the outcome of electroconvulsive therapy in severe and treatment-resistant depression. Mol. Psychiatry 20 (5), 609–614. Vitolo, E., Tatu, M.K., Pignolo, C., Cauda, F., Costa, T., Ando', A., Zennaro, A., 2017. White matter and schizophrenia: a meta-analysis of voxel-based morphometry and diffusion tensor imaging studies. Psychiat Res-Neuroim 270, 8–21. Waldmann, P., Meszaros, G., Gredler, B., Fuerst, C., Solkner, J., 2013. Evaluation of the lasso and the elastic net in genome-wide association studies. Front. Genet. 4, 270. Wang, R., Benner, T., Sorensen, A.G., Wedeen, V.J., 2007. Diffusion toolkit: a software package for diffusion imaging data processing and tractography. Proc Intl Soc Mag Reson Med 15 (3720). Weiner, R.D., Reti, I.M., 2017. Key updates in the clinical application of electroconvulsive therapy. International Review of Psychiatry 29 (2), 54–62. Whiteford, H.A., Degenhardt, L., Rehm, J., Baxter, A.J., Ferrari, A.J., Erskine, H.E., Charlson, F.J., Norman, R.E., Flaxman, A.D., Johns, N., Burstein, R., Murray, C.J.L., Vos, T., 2013. Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. Lancet 382 (9904), 1575–1586. Wolf, R.C., Nolte, H.M., Hirjak, D., Hofer, S., Seidl, U., Depping, M.S., Stieltjes, B., Maier-Hein, K., Sambataro, F., Thomann, P.A., 2016. Structural network changes in patients with major depression and schizophrenia treated with electroconvulsive therapy. Eur. Neuropsychopharmacol. 26 (9), 1465–1474. Wright, I.C., Rabe-Hesketh, S., Woodruff, P.W., David, A.S., Murray, R.M., Bullmore, E.T., 2000. Meta-analysis of regional brain volumes in schizophrenia. Am. J. Psychiatry 157 (1), 16–25. Wylie, K.P., Tregellas, J.R., 2010. The role of the insula in schizophrenia. Schizophr. Res. 123 (2–3), 93–104. Xi, Y.B., Guo, F., Xu, Z.L., Li, C., Wei, W., Tian, P., Liu, T.T., Liu, L., Chen, G., Ye, J., Cheng, G., Cui, L.B., Zhang, H.J., Qin, W., Yin, H., 2018. Radiomics signature: a potential biomarker for the prediction of MGMT promoter methylation in glioblastoma. J. Magn. Reson. Imaging 47 (5), 1380–1387. Xia, M., Wang, J., He, Y., 2013. BrainNet Viewer: a network visualization tool for human brain connectomics. PLoS One 8 (7), e68910. Xie, Y.Y., Cui, Z.X., Zhang, Z.M., Sun, Y., Sheng, C., Li, K.C., Gong, G.L., Han, Y., Jia, J.P., 2015. Identification of amnestic mild cognitive impairment using multi-modal brain features: a combined structural MRI and diffusion tensor imaging study. J. Alzheimers Dis. 47 (2), 509–522. Yu, H., Yan, H., Wang, L.F., Li, J., Tan, L.W., Deng, W., Chen, Q., Yang, G.G., Zhang, F.Q., Lu, T.L., Yang, J.L., Li, K.Q., Lv, L.X., Tan, Q.R., Zhang, H.Y., Xiao, X., Li, M., Ma, X., Yang, F.D., Li, L.J., Wang, C.Y., Li, T., Zhang, D., Yue, W.H., Pharmacogen, C.A, 2018. Five novel loci associated with antipsychotic treatment response in patients with schizophrenia: a genome-wide association study. Lancet Psychiat 5 (4), 327–338. Zhang, B., Tian, J., Dong, D., Gu, D.S., Dong, Y.H., Zhang, L., Lian, Z.Y., Liu, J., Luo, X.N., Pei, S.F., Mo, X.K., Huang, W.H., Ouyang, F.S., Guo, B.L., Liang, L., Chen, W.B., Liang, C.H., Zhang, S.X., 2017. Radiomics features of multiparametric MRI as novel prognostic factors in advanced nasopharyngeal carcinoma. Clin. Cancer Res. 23 (15), 4259–4269.

Please cite this article as: J. Gong, L.-B. Cui, Y.-B. Xi, et al., Predicting response to electroconvulsive therapy combined with antipsychotics in schizophrenia using..., Schizophrenia Research, https://doi.org/10.1016/j.schres.2019.11.046