Increased frontal gray matter volume in individuals with prodromal psychosis

Abstract Background Brain anatomical deficits associated with cognitive dysfunction have been reported in patients with schizophrenia. However, it remains unknown whether such anatomical deficits exist in individuals with prodromal psychosis. The present study is designed to investigate anatomical deficits in prodromal individuals and their associations with clinical/cognitive features. Methods Seventy‐four prodromal individuals and seventy‐six healthy controls were scanned using structural magnetic resonance imaging. Support vector machines were applied to test whether anatomical deficits might be used to discriminate prodromal individuals from healthy controls. Results Prodromal individuals showed significantly increased gray matter volume (GMV) in the right inferior frontal gyrus (IFG) and right rectus gyrus relative to healthy controls. No correlations were observed between increased GMV and clinical/cognitive characteristics. The combination of increased GMV in the right rectus gyrus and right IFG showed a sensitivity of 74.32%, a specificity of 67.11%, and an accuracy of 70.67% in differentiating prodromal individuals from healthy controls. Conclusion Our results provide evidence of increased frontal GMV in prodromal individuals. A combination of GMV values in the two frontal brain areas may serve as potential markers to discriminate prodromal individuals from healthy controls. The results thus highlight the importance of the frontal regions in the pathophysiology of psychosis.

32% (24%-35%) over the period of 3 years, and 36% (30%-43%) after 3 years. 4 However, a recent study showed a low psychosis incidence in the prodromal group after 12 months of follow-up, and the low incidence might be due to a short follow-up time. 5 The pathophysiological mechanism of prodromal psychosis still requires further exploration because of the inconsistencies in these field advancements.
Some studies observed that prodromal individuals had reduced GMV in several brain regions, particularly in the hippocampal gyrus, 6 lateral temporal lobe, 7 and prefrontal cortex(PFC), 7,8 including the medial PFC and lateral PFC. Previous review revealed that prodromal individuals exhibited reduced GMV in the temporal gyrus, PFC, and anterior cingulate cortex (ACC) before illness onset. 9 Nenadic et al found reduced GMV in the right middle/superior temporal, left superior frontal, and right postcentral cortices in prodromal individuals compared with healthy controls. Meanwhile, they also found increased GMV in the left temporal gyrus in prodromal individuals. 10 However, several neuroimaging studies found no GMV reduction between prodromal individuals and controls. 11,12 A recent study found no significant structural changes in the prodromal individuals, but patients with first-episode schizophrenia exhibited significantly decreased GMV in the bilateral superior parietal lobule and left orbital frontal cortex compared with prodromal individuals and healthy controls. 13 Hence, it remains controversial whether prodromal individuals have structural alterations.
Several factors may attribute to the inconsistent findings. First, sample size, clinical characteristics, and analysis methods differ across studies. For example, a recent study indicated that a voxelbased morphometry (VBM) analysis by Computational Anatomy Toolbox (CAT12) was more accurate and robust against volumetric alterations compared with the VBM8 toolbox. 14 Furthermore, a small sample size may confine the power to detect volumetric differences. Second, medication use can confound the results across studies. For example, previous studies have reported that antipsychotic treatment might decrease GMV in the temporal and frontal areas in early phases of psychosis. 15 Therefore, it is meaningful to conduct a structural study to examine whether prodromal individuals have GMV deficits after controlling for the abovementioned confounding factors.
The prediction of psychosis based on neuroanatomical biomarkers is possible by using multivariate pattern recognition approaches, including support vector machine (SVM). SVM has emerged as a promising tool for diagnostic purpose of various neuropsychiatric conditions. 16 Previous SVM results showed that the classification pattern included the prefrontal and temporal cortices, as well as a large bilateral cluster containing the parahippocampus and hippocampus where GMV reductions were recognized in the prodromal individuals. 17,18 SVM could successfully discriminate prodromal individuals from healthy controls with an accuracy of 68.42% based on structural magnetic resonance imaging (MRI) and diffusion tensor neuroimaging parameters. 19 Zarogianni et al 20 reported an accuracy of 74% for predicting later onset of psychosis, and the discriminative neuroanatomical pattern included many brain areas such as the temporal, frontal, and parietal regions. Therefore, SVM may be feasible in the early discrimination of psychosis using the neuroanatomical-based pattern recognition method.
In the present study, a relatively large sample of prodromal individuals was recruited. Prodromal individuals were drug-naive to eliminate the effects of medication use. Structural data were analyzed with the CAT12 method with optimized segmentation and normalization. Based on the abovementioned studies, we hypothesized that prodromal individuals would exhibit significantly decreased GMV in certain brain regions, especially in the prefrontal and temporal regions, which could be applied as potential image markers to identify prodromal individuals from healthy controls using SVM. We also hypothesized that decreased GMV would be significantly correlated with clinical/cognitive features.

| Participants
Seventy-four prodromal individuals from the Department of Test (CPT). BACS-SC is used to measure processing speed and attention. 23 HVLT-R is a list of learning verbal memory test including 12 words to assess verbal memory. 24 BVMT-R is widely utilized to evaluate visuospatial learning and memory in neuropsychological assessment. 25 SCWT is applied to evaluate the attention and working memory functions, which includes three parts. Part 1 is about reading a list of 100 words, and the words "red," "green," or "blue" are printed in black. Part 2 requires the participants to distinguish the ink color of a list of unmeaning characters. Part 3 requires the participants to report the ink color of the words "red," "green," and "blue." 26 TMT-A is applied to measure psychomotor speed, including 25 circles (numbered 1-25) distributed over a piece of paper. Participants are required to draw lines to link the numbers as quickly as possible in an ascending order. 27 CPT is a widely used measure of sustained attention. 28 These tests cover visual learning and memory, verbal processing speed, attention/ vigilance, and executive function.
Exclusion criteria for all participants were any physical illnesses, such as liver and kidney diseases, cardiovascular diseases, and any past or present neuropsychiatric disorders; any traumatic brain injury; seizures; drug or alcohol abuse or dependence; pregnancy; and any contraindications to MRI scan.

The Ethics Committee of the Second Xiangya Hospital of Central
South University approved this study. After a complete explanation, all participants (if the subject was under 18 years of age, the signature of the guardian was required) submitted their written informed consent.

| Scan acquisition
Magnetic resonance imaging scanning was conducted with a 3.0 T Siemens scanner (General Electric). The participants were told to lie supine and stay still with eyes closed. Foam pads and soft earplugs were used to reduce scanner head motion and noise. A 3D magnetization-prepared rapid acquisition gradient-echo sequence was used with the following parameters: repetition time of 2710 ms, echo time

| Statistical analysis
The clinical and demographic data of the two groups were compared by two-sample t tests or a chi-square test when necessary.
The differences of GMV between prodromal individuals and healthy controls were compared using voxel-wise two-sample t tests, with total intracranial volume, age, and years of education as covariates of no interest. The significance level was set at P < 0.05 corrected according to the Gaussian random field theory (voxel significance: P < 0.001, cluster significance: P < 0.05) for multiple comparisons with the REST software.
Once significant differences in GMV were observed in brain regions between the two groups, the mean GMV values were extracted from those brain regions. Pearson's correlation analyses between abnormal GMV and clinical/cognitive parameters were carried out with threshold of P < 0.05. Bonferroni correction was used to limit type I error.

| Classification analysis
To test the capacity of the combination of abnormal GMV in any two brain regions to discriminate the prodromal individuals from the controls, we applied a SVM ran in MATLAB using the LIBSVM software package (http://www.csie.ntu.edu.tw/~cjlin/ libsv m/). The "leave-one-out" method was employed in the study. 15.11 ± 1.8, P < 0.001). The scores of the BACS-SC, HVLT-R, CPT, BVMT-R, and SCWT are significantly lower in prodromal individuals than those in healthy controls. Prodromal individuals score significantly higher than healthy controls in the TMT-A scores.

| Differences in GMV between prodromal individuals and healthy controls
Relative to healthy controls, prodromal individuals exhibit significantly increased GMV in the right inferior frontal gyrus (IFG) (t = 4.1821) and right rectus gyrus (t = 4.0674). Table 2 and Figure 1 present the detailed information.

| Correlation analysis
No significant correlations are observed between increased GMV and clinical characteristics/cognitive function in prodromal individuals (P > 0.05, Bonferroni corrected).

| Distinguishing prodromal individuals from healthy controls
The

| D ISCUSS I ON
By analyzing the whole brain GMV with CAT12, prodromal individuals showed significantly increased GMV in the right IFG and right rectus gyrus compared with healthy controls. No correlations were noted between increased GMV in the two brain regions and clinical characteristics/cognitive function. Moreover, the SVM analysis showed that a combination of the GMV values in these two brain regions might be a potential marker to distinguish prodromal individuals from controls.
Our findings of increased GMV in the right IFG and right rec-  healthy controls. 7,30 Other studies from Asian even failed to find significant difference in the regional GMV between the prodromal individuals and healthy controls. 13 as hypertrophy or preapoptotic osmotic changes, could possibly increase regional volumes. 33 The expectation is supported by a longitudinal study with increased GMV in the right ACC, IFG, and left cerebellum in subjects at clinical risk for psychosis. 34 Third, medication use might confound previous studies. 12,35 Progressive GMV loss was reported after antipsychotic treatment. 36,37 Also, antipsychotic medications could contribute to the decline of brain tissue volumes in animal studies. 38, 39 Vernon et al 38 reported that F I G U R E 1 GMV differences between prodromal individuals and healthy controls. Increased GMV in the right IFG and right rectal gyrus were observed in the prodromal individuals. The color bar represents the t values of the group analysis of GMV. GMV, gray matter volume; IFG, inferior frontal gyrus F I G U R E 2 Visualization of classification by the method of support vector machine (SVM) using the combination of the GMV values in the abnormal brain regions. In the right of figure, dimension 1 and 2 represent the GMV values in the right IFG and right rectal gyrus, respectively. Red crosses represent the controls, and green crosses represent the prodromal individuals. GMV, gray matter volume; IFG, inferior frontal gyrus chronic exposure (8 weeks) to antipsychotic drugs in rats would induce significant decreases in the whole-brain volume, mainly in the frontal cortex volume. which could be normalized after withdrawal of the antipsychotic medications. 40 Prodromal individuals in the present study were drug-naive, and thus confounding effects induced by antipsychotic medications could be limited.
Therefore, it is expected that prodromal individuals in this study presented increased GMV. Fourth, sample size is relatively large in the present study, which may have statistical power to identify hypertrophic effects not indicated frequently in previous studies.
Fifth, CAT12 is an advanced method with optimized segmentation and normalization, which is more sensitive in the analysis of GMV than the previous version such as the VBM8 method. Other potential reasons are neuroplasticity. According to results from postmortem studies, alterations in GMV may be related to changes in dendritic density, synaptic, and neuronal, as well as increased afferentation in certain regions. 41  A few longitudinal studies revealed that prodromal subjects showed progressive GMV decreases over time. Prodromal subjects, who later developed psychosis, showed active GMV loss in several brain regions, including the prefrontal cortices, superior temporal gyrus, and parahippocampal gyrus during the transition period. 29,49 Recent multicenter study using a relatively large prodromal subjects cohort who developed psychosis, also showed progressive GMV loss mainly in the prefrontal regions, superior temporal, parietal, and parahippocampal regions. 50 Taken together, the current evidence revealed that the progressive pathological process precedes the first manifestation of overt psychosis in the prefrontal and other brain regions.
Previous study showed that more than 0.7 of specificity or sensitivity is good for establishing diagnostic index, 51 whereas less than 0.6 of specificity or sensitivity may be poor for diagnostic indicator. 52 SVM has been well applied in extensive biomedical applications for diagnostic purpose. 53 that the cognitive performance of prodromal individuals was similar to that of FES. Previous meta-analysis also suggested that prodromal individuals were significantly impaired in various cognitive function domains. 60 Together with the abovementioned studies, the present study showed cognitive deficits in prodromal individuals.
Several limitations exist in the present study. First, the study is cross-sectional. It is unclear how many of the prodromal individuals will transition to psychosis in the follow-up period, and whether increased GMV in the frontal lobes will be stable or decrease over time. Future longitudinal research is necessary to examine this possibility. Second, this study did not recruit patients with schizophrenia to compare their anatomical changes with those of prodromal individuals. This issue might limit our comprehending of the disease progression between prodromal individuals and patients with schizophrenia. Finally, level of education is unmatched between two groups. Cognitive differences may be driven by psychotic pathology of unmatched level of education. Although we tried to minimize the possible effects of unmatched level of education using it as a covariate of no interest in the analyses, the effects of unmatched level of education might not be completely eliminated. Therefore, the samples differ in terms of level of education and cognitive test performance which may have confounded the results.
In conclusion, the present study provides evidence of increased GMV in the frontal gyri in prodromal individuals. A combination of GMV values in these two brain areas may serve as potential markers to discriminate prodromal individuals from healthy controls. The results thus highlight the importance of the frontal regions in the pathophysiology of psychosis.

ACK N OWLED G M ENTS
This project was supported by grants from the National Key R&D Program of China (2016YFC1307100 and 2016YFC1306900) and the National Natural Science Foundation of China (Grant Nos. 81571310, 81771447, and 81630033).

CO N FLI C T O F I NTE R E S T
We declare that none of the authors holds any actual or potential conflict of interest for this study.