MRI and VBM

MRI is a technique widely used for radiological diagnosis and produces 3-dimensional images in excellent anatomical detail, and with a great capacity to differentiate soft tissue.Reference Gore ¹³ Various neuroimaging studies have revealed structural abnormalities in individuals affected by schizophrenia.Reference Nelson, Saykin, Flashman and Riordan^14–Reference Del Re, Konishi and Bouix¹⁶ Given the association of these data with symptom profiles, alterations identified through MRI suggest that biomarkers be used in the future of clinical practice. The benefits of using these biomarkers as a clinical assessment tool in schizophrenia center around the greater independence of cognitive ability, patient collaboration, and ease of measuring brain structures using neuroimaging techniques.

Various MRI techniques of the human brain have emerged through research, enabling structural tests that may help to consolidate previous findings and lead to the discovery of new patterns of abnormality in schizophrenia. For example, studies have been developed using VBM. This technique provides an automated and impartial analysis of the whole brain by identifying structural differences in a user-independent way. VBM compares grey and white matter concentrations between 2 groups of subjects involving high-resolution voxel-based images.Reference Ashburner and Friston ¹⁷ The spatially normalizing high-resolution images of the individuals studied are used in the same stereotactic space, and the analysis of the findings occurs via the following:

• • segmentation of the white and grey matter (GM) images and CSF compartments;

• • the images are smoothed and each tissue class is compared to the size of anticipated regional differences within the groups;

• • voxel-wise statistical analysis.Reference Ashburner and Friston ¹⁷

Compared to other previously used methods for the analysis of abnormalities in the brains of individuals with schizophrenia, VBM offers several advantages such as the region of interest (ROI) analysis, which has been widely employed in neuroimaging studies of schizophrenia. In this process, volumetric measurements are carried out by the researcher in areas of interest in an attempt to provide potential explanations for a previously stated hypothesis. However, these studies may be biased as new findings in unexpected regions could be neglected.Reference Wolkin, Rusinek and Vaid ^{18 ,} Reference Kubicki, Shenton and Salisbury ¹⁹ Furthermore, the manual techniques used to investigate regions of interest are time-consuming and could prejudice the comparison between various areas of an individual patient or their use on larger samples.Reference Kubicki, Shenton and Salisbury ¹⁹ It is still not possible to identify subclinical phenotypes through biomarkers in schizophrenia because of the heterogeneity of the disease process, wide divergence of results, and little correlation between clinical symptoms and abnormal patterns with MRI.Reference Kambeitz, Kambeitz-Ilankovic and Leucht ^{20 ,} Reference Scarr, Millan and Bahn ²¹

Given this background, MRI is categorized as a noninvasive test that can be performed in vivo and, when compared to other imaging techniques, relatively economical and proficientReference Dazzan, Arango and Fleischacker ⁸—something that is especially appealing to the low to middle income demographic.Reference Bandeira and Mendoza ²² For these reasons, this literature review will address the superior temporal gyrus (STG) as a possible biomarker in schizophrenia with the use of VBM on an MRI of the brain.

VBM of the STG in schizophrenic patients

Taking into consideration previous findings, which have identified changes to the temporal lobe of patients with schizophrenia,Reference Johnstone, Owens and Crow^23, Reference Suddath, Casanova, Goldberg, Daniel, Kelsoe and Weinberger²⁴ and the fact that electrical stimulation of the STG causes hallucinations,Reference Penfield and Perot²⁵ Barta and colleagues developed one of the first studies to attempt to identify STG as a possible biomarker for this mental disorder using MRI. This study examined 15 patients diagnosed with schizophrenia and 15 healthy subjects. In patients with schizophrenia, a lower volume was identified in the STG and left amygdala and an increased volume in the third ventricle. These results were not related to other confounding factors such as smaller brain size or temporal lobe. In conclusion, there was a correlation between the reduction of STG and the severity of hallucinations.Reference Barta, Pearlson, Powers, Richards and Tune²⁶

Certain caveats should be discussed for these results. The study included only male subjects that were being treated with neuroleptics and the scale designed to evaluate assessments—Scale for the Assessment of Positive Symptoms (SAPS)—was only used retrospectively. On the other hand, the inclusion and exclusion criteria were well-defined and the results are consistent with the possible anatomical and physiological correlations found in previous studies.

Although some posterior studies have found no morphological changes in the STG in patients with schizophrenia,Reference Zipursky, Marsh and Lim^27–Reference Ananth, Popescu, Critchley, Good, Frackowiak and Dolan³⁰ many others have been developed over time, demonstrating the reproducibility of these results both when lateralized to the leftReference Barta, Pearlson and Brill^31–Reference van Tol, van der Meer, Bruggeman, Modinos, Knegtering and Aleman³⁹ and also bilaterally.Reference Hirayasu, Shenton and Salisbury^40, Reference Gur, Turetsky and Cowell⁴¹ A large study of 159 schizophrenic patients with an age range of 55 years corroborated these findings, showing less gray matter in the STG of these schizophrenic patients compared to a healthy control group.Reference Hulshoff Pol, Schnack and Mandl⁴² Furthermore, post mortem and MRI studies identified changes in various subregions of the STG, and the most significant results concern the anterior and posterior regions, as well as the total area of the STG.Reference Barta, Pearlson and Brill^31–Reference Matsumoto, Simmons and Williams^34, Reference Hirayasu, Shenton and Salisbury^40, Reference Gur, Turetsky and Cowell^41, Reference Shenton, Kikinis and Jolesz⁴³

The STG is composed of functional subregions that play important roles within the cerebral cortex. Studies analyzing these subregions in schizophrenia also pointed to structural changes involving considerable impairments to brain functions. More significant evidence indicates the primary auditory cortex (Heschl gyrus), and the neocortical language region (planum temporale).Reference Sumich, Chitnis and Fannon^44, Reference Hasan, Kremer and Gruber⁴⁵ Regarding the planum temporale, studies point to a possible role in the clinical symptoms of schizophrenia, particularly in relation to auditory hallucinations (AH) and thought disorder.Reference Shenton, Kikinis and Jolesz^43, Reference Flaum, O’Leary, Swayze, Miller, Arndt and Andreasen^46, Reference Menon, Barta and Aylward⁴⁷

Considering these more specific findings, Barta developed a new study correlating changes in planum temporale with schizophrenia in an attempt to extend and replicate his previous results and compare the new findings regarding surface area and volume of GM with other research results. The data replicated previous findings, showing evidence of a reversal of planum temporale surface area asymmetry in schizophrenia in both genders, as well as reduced volume and thickness of the planum temporale.Reference Barta, Pearlson and Brill³¹

One study examined 25 men with first-episode schizophrenia according to DSM-III criteria and who had never previously used neuroleptics, and 25 healthy men as a control group. The study subjects were matched by age and level of parental educational and clinical symptoms were studied according to 3 main dimensions (negativity, psychosis, and disorganization) using the Scale for Assessment of Negative Symptoms and the SAPS. Two major findings were identified: (1) strong correlation between psychotic symptoms and decreased left anterior STG volume; and (2) positive correlation between negative symptoms and right posterior STG volume. These results indicate pathophysiological differences leading to psychotic and negative symptoms during neurodevelopment.Reference Kim, Crespo-Facorro, Andreasen, O’Leary, Magnotta and Nopoulos⁴⁸

Another study involved 40 patients with schizophrenia (38 on antipsychotics) and 40 healthy subjects matched for age and sex. The following changes to the STG were identified through MRI: (1) patients with schizophrenia have reduced STG volumes bilaterally; and (2) considering the 5 regions of the STG in schizophrenic patients, the areas which had significantly lower volumes were the lateral region and planum temporale. The authors concluded that there was an association between volume of cortical thickness, surface area, and local gyrification, with no difference in laterality.Reference Ohi, Matsuda and Shimada⁴⁹ It should be noted that the study only included Japanese subjects, which makes the sample more homogeneous but, at the same time, there is low external validity of the results. The authors subsequently falsely conclude that the volume decrease in the subregion side is due to the cortical thickness and surface area, which is incorrect because the study design does not assert causality, merely association.

Morphological changes are also related to the type of symptoms (positive/negative) and their severity, as identified in the studies of Barta et alReference Barta, Pearlson, Powers, Richards and Tune²⁶ and Kim et al.Reference Kim, Crespo-Facorro, Andreasen, O’Leary, Magnotta and Nopoulos⁴⁸ These findings could be used to assist to predict the clinical course and therefore allow modifications to the therapeutic management of these patients. Groups of patients with the same symptoms could bring less heterogeneity to future studies and more detailed functional deficits. In this context, a recent study, also aimed to correlate the findings for VBM with positive and negative symptoms.Reference Escarti, Garcia-Marti and Sanz-Requena⁵⁰ With this objective, 215 patients with first episode psychosis were evaluated, 93 with AH and 122 without AH compared to 177 healthy controls. Lower GM was found in the STG when compared to the healthy control and there was also a significant negative correlation between the Positive and Negative Syndrome Scale (PANSS) with GM reductions in the left STG. This suggests that the higher the score on the PANSS, the lower the gray matter in this area.

Another study from 2011, proposed an interesting design regarding symptomatology: the authors compared the VBM analysis of 47 subjects, of which 12 experienced “outer space hallucinations,” 15 experienced “inner space hallucinations,” and there were 20 healthy controls.Reference Plaze, Paillere-Martinot and Penttila⁵¹ They were able to show a significant decrease in white matter volume in a cluster of voxels located in the right STG of patients with “outer space hallucinations” compared to the “inner space hallucinations” group. This result suggests that certain differences regarding symptoms of inner and outer space hallucinations can be related to differences in certain areas of the STG, mainly white matter areas, implying that this difference could be caused by auditory pathways in the brain. Another provocative approach that aimed to prove STG difference in different schizophrenia subgroups, was taken in a study in which 99 patients with schizophrenia were compared to 113 matched healthy controls. The thickening in the STG was verified during the comparison of these 2 groups. The author subsequently stratified the schizophrenic group into 3 (paranoid, disorganized, negative) to check if there were differences between them. However, in respect of the applied stratification, STG was reduced in these 3 subtypes, showing that STG is a common marker for each of these different subgroups.Reference Nenadic, Sauer and Gaser⁵²

Differences in the STG of schizophrenic patients were also found in a drug-resistant population with AH.Reference Modinos, Vercammen, Mechelli, Knegtering, McGuire and Aleman^37, Reference O’Daly, Frangou, Chitnis and Shergill⁵³ A study carried out by Modinos and colleaguesReference Modinos, Vercammen, Mechelli, Knegtering, McGuire and Aleman³⁷ explored the role of VBM in 26 schizophrenic subjects with treatment-resistant AH. The covariance analysis showed a positive function relating loss of GM in the left STG and severity of symptoms. The main drawback in this article is that they did not have a control for medication use and the study was only cross-sectional. Another study that confirmed these findings in a subpopulation of schizophrenic patients with persistent hallucinations, also showed a significant reduction of STG GM in 28 schizophrenic patients with persistent hallucination.Reference O’Daly, Frangou, Chitnis and Shergill⁵³

Previous studies have shown that individuals who misuse alcohol and cannabis but are not psychotic have volume deficits in the STG area, in a similar way to schizophrenic patients.Reference Xiao, Dai, Zhong, Zhu, Shi and Pan⁵⁴ For this reason, a study from 2018 aimed to use VBM to evaluate the impact of substance use disorder (alcohol and cannabis) in schizophrenic patients.Reference Quinn, McHugo, Armstrong, Woodward, Blackford and Heckers⁵⁵ To evaluate this, they analyzed the VBM of 92 schizophrenic patients with substance use, 66 patients without substance use, and 88 healthy controls. Despite this rationale, they were not able to find any significant difference in the loss of gray matter when comparing substance users and nonusers. This result implies that substance use of alcohol and cannabis does not play an important role in the loss of gray matter in schizophrenic subjects. However, it should be stated that this study has several limitations, as it is a cross-sectional study only, being based on patient self-report and excluded patients that had used drugs in the previous month. That said, other studies have produced similar results.Reference Malchow, Hasan and Schneider-Axmann^56, Reference Thoma and Daum⁵⁷

Despite significant differences in the findings, the overlap of structural patterns between schizophrenic patients and healthy control groups discourages the use of MRI as a diagnostic tool.Reference Boksa^9, Reference Kambeitz, Kambeitz-Ilankovic and Leucht²⁰ Analyzing samples from various studies that describe brain abnormalities in patients with schizophrenia, it was observed that multiple variables could complicate the interpretation of results. These samples included patients who develop the disease at different times in their lives, treatments that differ in both the type of drug used, as well as its dosage and duration. The presence of several types of comorbidities and the severity of the disease is also worth noting. Thus, considering the progressive nature of schizophrenia, changes in brain morphology could be associated with severity of disease and age of patients.Reference DeLisi, Sakuma, Tew, Kushner, Hoff and Grimson^58, Reference Hulshoff Pol, Schnack and Bertens⁵⁹ Such confounding factors need to be taken into account in further studies to avoid the potential for bias and provide more reproducible results. An alternative would be to use studies with more stringent inclusion criteria and more homogeneous populations.

VBM of STG in Individuals at High and Ultra-High Risk for Psychosis

It is important also to study the subjects at high risk (HR) and ultra-high risk (UHR) for psychosis as, according to meta-analysis of transition outcomes in subjects at clinical HR of psychosis, there is a tendency (of approximately 35%) to convert to psychosis in 3 years.Reference Fusar-Poli, Bonoldi and Yung⁶⁰ For this reason, in studying this population, it is important to comprehend the “natural history” of schizophrenia. A major drawback of certain VBM studies derives from the fact that they were cross sectional, and several cohort studies were made to try and identify if the VBM of the STG could be a prognostic factor in psychosis and schizophrenia within this UHR and HR population.

In 2003, Job and colleagues carried out a longitudinal study with 65 young adults at HR of schizophrenia and 19 healthy young adults as a control over a period of 18 months.Reference Job, Whalley, McConnell, Glabus, Johnstone and Lawrie⁶¹ They found significant GM decrease of the left and right STG comparing the HR and control groups. They also performed an analysis of the HR subjects that had psychotic symptoms over time, but they were not able to find any differences regarding STG. This result suggests that either this region is an initial marker that has altered since the first stages of the disease and therefore cannot be used as a prognosis marker or that they were underpowered due to the small sample size being unable generate a statistical significance.

Again in 2003, Job and colleagues published another study, in which they compared 34 first-episode psychosis subjects with 146 HR subjects.Reference Job, Whalley, McConnell, Glabus, Johnstone and Lawrie⁶¹ Although they were able to find differences in several regions, they were not able to find a significant difference regarding STG. This result could be explained by the low magnetic field of the machine (1T), the cross-sectional design, or because the group of HR subjects was only selected using the criteria of 2 or more affected first- or second-degree relatives with schizophrenia, ignoring, therefore, other risk factors.

Ziermans and colleagues also investigated this issue in a longitudinal format.Reference Ziermans, Schothorst and Schnack⁶² They performed a VBM analysis of 43 adolescents at UHR for psychosis compared with 30 healthy controls, at baseline and 2 years later (or when they became psychotic). Although they did not find any significant difference regarding the loss of GM in the STG from the ones who converted to psychosis, this could be explained by the high percentage (49%) of individuals in this group that were already taking psychotropic medications. Another possible explanation was the lower number of subjects that had converted to psychosis (16%) compared with the expected figure of 40%.

A broader multicentric case-control study was carried out in 2011 with 182 individuals at UHR from 5 different sites with 167 healthy controls. Conversely, this showed that subjects at UHR that later developed psychosis had lower GM volume in the superior temporal cortices.Reference Mechelli, Riecher-Rossler and Meisenzahl⁶³ Another study compared the VBM of 30 UHR subjects with 29 controls and found less GM in the STG of the UHR subjects: the main issue of this study being cross-sectional design.Reference Witthaus, Kaufmann and Bohner⁶⁴ These results differ from previous ones possibly because of the use of larger samples and a more diverse population, whereas the study carried out by Mechelelli was multicentric.Reference Mechelli, Riecher-Rossler and Meisenzahl⁶³

Another multicentric study was undertaken in 2015 and followed 274 individuals at clinical HR, 35 of whom had converted to psychosis, and 135 healthy subjects.Reference Cannon, Chung and He⁶⁵ The MRI was carried out at the baseline and at 12 months during follow-up or at the point of conversion to psychosis. It was found that the “converters” showed higher grades of thickness in the STG when compared to controls, although when controlled for multiple comparisons the results were not deemed significant. One interesting characteristic of this study is that it involved only drug-naive individuals. This study highlights interesting data because previous research in animal models had shown loss of GM, which was related to the use of antipsychotic drugs. However, the study performed in humans by Cannon et alReference Cannon, Chung and He⁶⁵ found that there is a loss of GM even when not using antipsychotics. Also in 2015, a further study analyzed and compared 21 individuals with an At Risk Mental State (ARMS) for psychosis, 26 patients with a First Episode of Psychosis (FEP) and 24 healthy controls; and found that thickening of the STG was present either in the ARMS and in the FEP when compared with the healthy control group, indicating a marker of vulnerability to the disease.Reference Benetti, Pettersson-Yeo and Hutton⁶⁶

The majority of the studies show that a decrease in GM of the STG is apparent in either the HR or UHR schizophrenic group. Additionally, other studies discovered a trend, which showed that, as the diseases progress, more GM is lost in the STG.Reference Cannon, Chung and He⁶⁵ However, to confirm the use of VBM in STG as a prognostic marker, further studies are necessary with a larger sample size and longer follow-up procedures (the longest being 18 months), to better delimitate and understand changes in the STG detected by VBM.

Future Perspectives

Machine learning is a new and growing field in medical research that could be used as a tool to assist diagnosis. The use of machine learning when coupled with previously used techniques such as VBM is an interesting field that should be further explored, as we have already seen findings that may be helpful in reaffirming STG as an important biomarker in schizophrenia.

A 2009 study used multivariate neuroanatomical pattern classification to identify subjects at HR for psychosis using an MRI machine-based learning technique. The results were promising regarding the prediction of the transition stage of attenuated psychotic symptoms to frank psychotic symptoms with high specificity and sensitivity, while also showing the use of the technique to identify cortical alterations in chronic stages of schizophrenia.Reference Koutsouleris, Meisenzahl and Davatzikos⁶⁷

Moreover, recent research findings have shown new ways to analyze the VBM data from schizophrenic subjects.Reference Chin, You, Meng, Zhou and Sim^68, Reference Vyskovsky, Schwarz and Kasparek⁶⁹ One study from 2018 used a different approach of “merging” the use of 2 techniques—ROI and VBM—through a machine learning system that uses a support vector machine (SVM) and fundamentally classifies data points by maximizing margins between classes in dimensional space.Reference Chin, You, Meng, Zhou and Sim⁶⁸ The SVM model accurately recognized 86.6% of the training set of 127 individuals and accurately validated 83.5% of a set of 85 individuals. A sequential ROI was then used by the machine to find optimal discriminatory subsets, finding 7 principal ones: anterior STG, occipital fusiform gyrus, middle frontal gyrus, pars opercularis of the inferior frontal gyrus, superior frontal gyrus, left thalamus and left lateral ventricle. With this combined subset, the machine was capable of achieving a 96.6% sensitivity and 74.1% specificity in the validation set with 127 individuals.

Another article used a combined approach of VBM and deformation-based morphometry performed with machine learning on a dataset of 104 subjects (52 schizophrenic and 52 healthy controls).Reference Vyskovsky, Schwarz and Kasparek⁶⁹ Regarding VBM, this was mainly used in the areas of significant interest such as temporal lobe (STG, and middle temporal gyrus), frontal lobe (inferior frontal gyrus, medial frontal gyrus, superior frontal gyrus, middle frontal gyrus, and precentral gyrus), limbic lobe (cingulate gyrus), occipital lobe, and parietal lobe. An accuracy of 73.12% for the detection of schizophrenic patients was demonstrated when using this combined approach.

Another possible approach, suggested by some authors, is through the use of machine learning in multimodal data, consisting of the integration of multiple types of data (eg, genetics, imaging, epigenetic) through algorithms to achieve a final biomarker.Reference Alam, Lin, Deng, Calhoun and Wang¹² For instance, this study used a kernel machine (variation of machine learning), that centers on statistical learning theory to create a functional analysis (12). A strong degree of reliability has been achieved using gene-derived SNPs (MAGI2), ROIs (CRBLCrus1.L), and gene-specific DNA methylation (FBXO28) at p ≤ 0.000001.

Summary and Conclusions

The individuals affected by this mental illness appear to have smaller STG volumes when compared to healthy controls and also to subjects with a diagnosis of first-episode affective psychosis or groups of individuals at HR of psychosis.Reference Kasai, Shenton and Salisbury⁷⁰ The rationale for these structural changes comes from certain findings such as higher levels of neuroinflammatory markers associated with neural tissue death in the postmortem analysis of neural tissues of schizophrenic patients, which leads to the presence of lower levels of GM in the brains of these patients.Reference Rao, Kim, Harry, Rapoport and Reese^71, Reference Milatovic, Gupta, Yu, Zaja-Milatovic and Aschner⁷²

This raises a fundamental question: Is the loss of gray matter in the STG the cause of the onset of psychosis? The obvious answer is that it can be, but there does not appear to be a critical threshold below which the psychosis is developed. It appears to be more of a conceptualized marker that contributes to the synaptic reduction and which could be involved in several processes that culminate in the emergence of the psychosis. Also, as the disease progresses it appears to play an important role as a prognostic marker, showing reduction with disease progression.Reference Cannon, Chung and He⁶⁵

One of the problems that emerges with the use of the STG as a biomarker is the discrepancy in defining parameters found in a large number of studies which, in itself, is a contributing factor to the lack of correlation between brain abnormalities and clinical symptoms. Areas of disagreement include the fact that some studies use ROI to analyze STG, whereas other studies have, as their main objective, an analysis of only STG subregions or specific supratemporal plane regions.Reference Kim, Crespo-Facorro, Andreasen, O’Leary, Magnotta and Nopoulos⁴⁸ It is therefore necessary to standardize the nomenclature of the areas to be studied in the future as this will lead to more consistent results, allowing higher clinical and morphological correlations.

Several advances in the treatment of brain and mind disorders have emerged in the past decades,Reference Bandeira, Guimaraes and Jagersbacher^73–Reference Correia-Melo, Leal and Vieira⁷⁹ although there is still a need for the development and consolidation of biomarkers in psychiatry to enable further achievements in the improvement of mental health screening and diagnosis, as well as increased health care quality and a reduction in the mental health burden inherent to mental illness.Reference Barreto, Bento and Barreto^80, Reference Kessler, Aguilar-Gaxiola and Alonso⁸¹