Imaging Schizophrenia with Voxel Based Morphometry

Imaging Schizophrenia; Voxel Based Morphometry Reveals Differences in Superior Temporal Gyrus Volumes

Introduction

The role of magnetic resonance imaging (MRI) in assessing brain architecture has become increasingly popular among neuroscientists studying the complex nature of the brain. This non-invasive tool has enabled the in-vivo investigation of spatially varying patterns that occur over time as a function of normal development and aging as well as in the perturbation of this trajectory in the presence of disease. The multi-modal nature of MRI has allowed for the assessment of brain morphology, connectivity, and function and the precision of this technology has only increased with time. Moreover, the association of these measurements with genetic and cognitive indices has propelled our understanding of the synergistic interplay between brain structure and function. Among these implementations lies structural MRI, a modality that allows for the quantification of density, volume, and thickness of grey and white matter within the brain. These macroscopic anatomical changes are thought to be an outcome of the microscopic changes occurring at the neuronal and synaptic level ultimately shedding light on the biological processes driving a disease or function (Tardif et al., 2016). Voxel based morphometry, a technique that utilizes structural MRI, has been instrumental in comparing local gray matter concentrations between different groups of interest (Ashburner et al., 2000).

Voxel Based Morphometry

Voxel-based morphometry(VBM) involves a number of steps in order to ensure that regional differences between groups are initially comparable in their spatial location. The methodological steps described below have been described at length in Ashburner et al., 2000. To summarize, the first step of the pipeline involves pre-processing the images so that they are realigned and normalized to a standard template space where all resulting images will occupy the same 3D coordinates. This is performed by estimating the optimum 12-parameter affine transformations where images are scaled, rotated, translated, or sheared accordingly. Prior knowledge of variability in brain size is used to constrain maximum estimates. A second step is taken to account for global nonlinear shape differences by implementing a linear combination of smooth spatial basis functions and then by calculating the coefficients of the functions that minimize the difference between the template and the subject image while maximizing the smoothness of the deformations.

Next, a probabilistic classification of tissue types is calculated using a modified mixture model cluster analysis that classifies each voxel in the image as being either grey matter, white matter, or cerebrospinal fluid (CSF). Due to the heterogenous nature of image intensities across and within subjects, a correction for nonuniform image intensities is implemented prior to the segmentation. A third step is taken to ensure the data is valid for statistical testing and regions are comparable among subjects. This step involves apply a smoothing kernel where each voxel is smoothed by means of a gaussian shaped filter that should occupy the same size of the expected regional differences between groups. This also compensates for the imprecise nature of the first spatial normalization step. Lastly, statistical analysis is performed by using the general linear model to identify regions of grey matter (or white matter/CSF) concentration between groups. These voxel-wise parametric tests, typically F or t tests with a correction for multiple comparisons, are used to test hypothesis regarding group differences and are shown in terms of p-value significance.

Although innovative at the time of inception, as with all technology over time, there has been advancements and optimizations to overcome initial limitations. Since the year 2000, VBM has run on different versions of the SPM (Statistical Parametric Mapping) software. Diaz-de-Grenu et al., 2014 set out to evaluate the effects of different software versions with respect to Alzheimer’s disease and have summarized the differences as follows. Standard VBM (SPM99 and SPM2) spatially normalized the images to their respective T1 templates and classification of tissue types was performed in this spatially normalized space. There has been documented limitations with this methodology in that spurious group differences may be detected due to the nature of the spatial normalization errors (Bookstein, 2001). Subsequent “optimized” versions have therefore been implemented for both SPM99 and SPM2 that uses an iterative approach to the normalization and segmentations steps of the pipeline. In the optimized versions of the software, probabilistic tissue segmentation is performed in the native space and spatial normalization of the resulting grey matter segment is instead performed and this optimized transformation is then applied to the original structural image. This “iterative” concept is then introduced in that segmentation and modulation is applied once more to this spatially transformed image using the Jacobian determinant of the transformation field. Although the optimized versions of SPM99 and SPM2 both have this capability, SPM5 introduced a “unified segmentation” approach that implemented this process as well as bias correction all in one step.

One of the more recent renovations to the software however, occurs in SPM8’s implementation of diffeomorphic anatomical registration through exponentiated Lie algebra, also known as DARTEL. This approach is able to register the images using a “flow field that is able to generate both forward and backward deformations” (Ashburner, 2007). In this new method, grey and white matter classifications are simultaneously registered to a study-specific average image. Then, before modulation, a new custom template is created by repeatedly taking the mean of the inverse transformations used to create the average image (Diaz-de-Grenu et al., 2014). This group template is then used to create final individualized tissue maps. Nonetheless, many of the results investigated using the various VBM implementations have produced similar findings. One population that is studied quite extensively with the use of VBM are those inflicted with schizophrenia.

Schizophrenia

Schizophrenia is a serious mental illness that is characterized by long-term or recurrent psychosis and results in the chronic deterioration of many functional abilities (Modinos et al., 2012). Some of the core positive features include delusions and hallucinations whereas negative features include impaired motivation, reduction in spontaneous speech, and social withdrawal (Owen et al., 2016). Although the etiology of schizophrenia is not well elucidated, there have been great efforts in the past few decades that have contributed to our understanding of the potential mechanisms by which the neuropathology of schizophrenia manifests. One major way this has been possible has been through the association of behavioral and cognitive deficits seen in schizophrenics and the macroscopic brain changes assessed using structural MRI (Bakhshi, 2015).

Superior Temporal Gyrus

VBM, as described previously, has been widely used by researchers in identifying regional abnormalities implicated with the disease. A consistent finding among these studies has been that compared to healthy controls, patients with schizophrenia exhibit a significantly higher progressive decrease in total grey matter volume over time. Many longitudinal and cross-sectional studies and meta-analysis have revealed a pronounced effect localized to the superior temporal gyrus, a region central to language processing and perception. These studies have been associated with positive symptom severity in schizophrenia one of which includes hallucinations (Vita et al., 2012, Modinos et al., 2012). Here, I will focus on 10 cross-sectional observational studies that have associated decrements in superior temporal gyrus volume with positive symptom severity – most notably as a measure of auditory verbal hallucinations (AVHs).

Literature Review: Demographics

Seven of the studies incorporated here included both male/female participants (Gaser et al., 2004; O’Daly et al., 2007; Nenadic et al., 2010; Tang et al., 2012; van Tol et al., 2014; Narayanaswamy et al., 2015; Kim et al., 2017), two of the studies included only males (Shapleske et al., 2002; García-Martí et al., 2008) and one study did not specify the sex of the participants (Neckelmann et al., 2006). Three of the studies recruited both right and left handed subjects (Shapleske et al., 2002; O’Daly et al., 2007; van Tol et al., 2014), five had only right handed participants (Gaser et al., 2004; García-Martí et al., 2008; Tang et al., 2012;  Narayanaswamy et al., 2015; Kim et al., 2017) and two studies did not specify handedness (Neckelmann et al., 2006; Nenadic et al., 2010). All studies specified mean ages with standard deviations except for one (Neckelmann et al., 2006) where only a range of ages was given. Moreover, all studies in this review focus on schizophrenia phenotypes in adulthood except for one (Tang et al., 2012) where the specific aim of the study was to assess first-episode early on-set schizophrenia. Since there is evidence that anti-psychotic medication can affect brain structure, a stated benefit of this study was that since the participants had a shorter duration of illness as well as administration of medication, this could reduce the confounding effects of medication as a variable. To this end, one study in this review pertained to schizophrenics who were anti-psychotic naïve (Narayanaswamy et al., 2015) in order to further investigate the effects of medication during adulthood. All other studies in this review contained a either a mixture of patients on anti-psychotic medication, or all patients were on anti-psychotic medication.

Literature Review: Experimental Designs

Symptom severity in schizophrenia is often assessed by clinicians using a number of different scales. The studies described in this review had at least one population of schizophrenics classified under DSM criteria. In addition, four of these studies (Shapleske et al., 2002; Gaser et al., 2004; Nenadic et al., 2010; van Tol et al., 2014) had a subdivision of their schizophrenic population in which a subset of participants exhibited auditory verbal hallucinations, and the others did not. The purpose of this design was to assess whether or not symptom severity correlated with the degree of regional atrophy in the schizophrenic groups. Using a different and in some cases an additional experimental design, eight of these studies (Shapleske et al., 2002; Neckelmann et al., 2006; O’Daly et al., 2007; García-Martí et al., 2008; van Tol et al., 2014; Kim et al., 2017; Tang et al., 2012;  Narayanaswamy et al., 2015) used a healthy control group for a case/control assessment. In addition, many of these studies still calculated a correlation with symptom severity and volumes within specified regions within the schizophrenic groups. Some of the clinical assessments included in these studies are SAPS (Scale for Assessment of Positive Symptoms)(Shapleske et al., 2002; Gaser et al., 2004; Nenadic et al., 2010), SANS (Scale for the Assessment of Negative Symptoms) (Gaser et al., 2004), PANSS(Positive and Negative Syndrome Scale)( O’Daly et al., 2007; Tang et al., 2012;  van Tol et al., 2014; Narayanaswamy et al., 2015; Kim et al., 2017), PSYRATS(Psychotic Symptom Rating Scale)( García-Martí et al., 2008), and BPRS(Brief Psychiatric Rating Scale) (Neckelmann et al., 2006; García-Martí et al., 2008).

Literature Review: Methodology/Statistical Analysis

As described previously, there have been several different implementations/ improvements regarding VBM software. Some of the earlier studies described here use SPM99 (Shapleske et al., 2002; Gaser et al., 2004; Neckelmann et al., 2006; O’Daly et al., 2007) and SPM2 (García-Martí et al., 2008; Nenadic et al., 2010) and have variable smoothing parameters, one of which is not indicated at all (Nenadic et al., 2010). The more recent studies begin to implement SPM5 (Tang et al., 2012) and the more innovative SPM8 diffeomorphic model DARTEL (van Tol et al., 2014; Narayanaswamy et al., 2015; Kim et al., 2017). For statistical analysis, three studies did not correct for multiple comparisons (Shapleske et al., 2002; Gaser et al., 2004; Nenadic et al., 2010), three used false discovery rate (FDR) correction (Neckelmann et al., 2006; García-Martí et al., 2008; Tang et al., 2012), three used family-wise error (FWE) correction (van Tol et al., 2014; Narayanaswamy et al., 2015; Kim et al., 2017;), and one did not indicate whether or not multiple comparisons was accounted for (O’Daly et al., 2007). Many of these studies reported significant findings with and without correction as well. The p-value significance threshold varied among studies as well, please see Table 1 for details.

Literature Review: Results: Schizophrenics vs Healthy Controls

Despite the heterogeneity in the demographics, experimental design, and the methodology, a consistent decrease in volume in either the left (Neckelmann et al., 2006; Tang et al., 2012; van Tol et al., 2014) right (O’Daly et al., 2007) or bilateral (Shapleske et al., 2002; García-Martí et al., 2008; Kim et al., 2017) superior temporal gyrus is observed when comparing a schizophrenic population to a healthy control group. Although beyond the scope of this review, other consistent regions are commonly found to be affected in those who have schizophrenia including the insula, anterior cingulate, pre/cuneus, and lingual gyri. Please see Table 1 for the results of all the regions implicated in schizophrenia in these studies.

Literature Review: Results: Auditory Verbal Hallucination Associations

Four of the studies looked directly at the comparison between schizophrenics with AVHs and those without AVHs (Shapleske et al., 2002; Gaser et al., 2004; Nenadic et al., 2010; van Tol et al., 2014) and found differences in some regions, only one was specific to superior temporal gyrus (Nenadic et al., 2010). Moreover, many of studies looked at the relationship between the severity of symptoms and the degree to which differences regions are atrophied and found that superior temporal gyrus volume decreases were correlated with positive symptom scores (O’Daly et al., 2007; Tang et al., 2012; Narayanaswamy et al., 2015; Kim et al., 2017). This finding, coupled with studies that mitigate the potential of anti-psychotic medication being a confounding factor(Tang et al., 2012; Narayanaswamy et al., 2015) to the effects seen in the superior temporal gyrus, bring more evidence to the notion that the superior temporal gyrus is involved in hallucinations in schizophrenia.

Discussion

The superior temporal gyrus is a key player in language processing and perception. The left superior temporal gyrus, which includes both primary and association auditory cortices, is known to be involved in the perception and comprehension of phonological and semantic aspects of speech (Wernicke’s area). The right superior temporal gyrus, also involved in language and auditory processes, has a more integral role in the emotional salience aspect of speech and language (Modinos et al., 2012). Given this knowledge and the neuroimaging results presented here, the notion that the superior temporal gyrus is implicated in the aetiology of positive schizophrenic symptoms is indeed compelling. It has been suggested that these volume reductions, potentially driven by a reduction or disconnection of neurons, may be leading to the inability to inhibit and correctly allocate internally generated speech. Future studies incorporating a multi-modal approach using connectivity analysis (functional and structural) can begin to offer more insight into the complex mechanisms that manifest in hallucinations plaguing those with schizophrenia (Modinos et al., 2012).

Indeed there are many limitations within studies such as older methodologies, the lack of proper statistical corrections, and groupings of different populations that may actually be affected by the disease differently (i.e. right vs left handed, male vs female, and age variability).  Moreover, there is heterogeneity across studies as well, making the comparison between studies less defined. There are also limitations regarding VBM that raise concerns regarding the accuracy of regional volumetric differences due to atypical brains, as well as the validity of statistical inferences (Perlini et al., 2012). Despite the numerous limitations described here, VBM has still been able to offer compelling results regarding regional differences in patients with schizophrenia, one of which involves the superior temporal gyrus. The ability to associate these differences with different cognitive measures like symptom severity scores is what propels our understanding of disease forward – in hopes that it can help elucidate paths to treatment and prevention for diseases like schizophrenia.

References

• Ashburner, J. (2007). A fast diffeomorphic image registration algorithm. NeuroImage, 38, 95–113. https://doi.org/10.1016/j.neuroimage.2007.07.007
• Ashburner, J., & Friston, K. J. (2000). Voxel-Based Morphometry — The Methods. NeuroImage, 821, 805–821. https://doi.org/10.1006/nimg.2000.0582
• Bakhshi, K., & Chance, S. A. (2015). The neuropathology of schizophrenia: A selective review of past studies and emerging themes in brain structure and cytoarchitecture. Neuroscience, 303, 82–102.
• Bookstein, F. L. (2001). “Voxel-Based Morphometry” Should Not Be Used with Imperfectly Registered Images. NeuroImage, 1462, 1454–1462. https://doi.org/10.1006/nimg.2001.0770
• Diaz-de-grenu, L. Z., Acosta-cabronero, J., Feng, Y., Chong, V., Pereira, J. M. S., & Diseases, N. (2014). A Brief History of Voxel-Based Grey Matter Analysis in Alzheimer’s Disease. Journal of Alzheimer’s Disease, 38, 647–659. https://doi.org/10.3233/JAD-130362
• García-Martí, G., Aguilar, E. J., Lull, J. J., Martí-bonmatí, L., Escartí, M. J., Manjón, J. V, … Sanjuán, J. (2008). Schizophrenia with auditory hallucinations: A voxel-based morphometry study. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 32, 72–80. https://doi.org/10.1016/j.pnpbp.2007.07.014
• Gaser, C., Nenadic, I., & Volz, H. (2004). Neuroanatomy of ‘Hearing Voices’: A Frontotemporal Brain Structural Abnormality Associated with Auditory Hallucinations in Schizophrenia. Cerebral Cortex, (January), 91–96. https://doi.org/10.1093/cercor/bhg107
• Kim, G., Kim, Y., & Jeong, G. (2017). Whole brain volume changes and its correlation with clinical symptom severity in patients with schizophrenia: A DARTEL-based VBM study. PLoS ONE. https://doi.org/10.1371/journal.pone.0177251
• Modinos, G., Tol, V., Mcguire, P. K., Modinos, G., Costafreda, S. G., & Allen, P. (2012). Neuroanatomy of auditory verbal hallucinations in schizophrenia: A quantitative meta-analysis of voxel-based morphometry studies. Cortex, 9, 1046–1055. https://doi.org/10.1016/j.cortex.2012.01.009
• Narayanaswamy, J. C., Kalmady, S. V, Sc, M., Venkatasubramanian, G., & Ph, D. (2015). Clinical Correlates of Superior Temporal Gyrus Volume Abnormalities in Antipsychotic-Naïve Schizophrenia. Journal of Neuropsychiatry. https://doi.org/10.1176/appi.neuropsych.14030049
• Neckelmann, G., Specht, K., Lund, A., Ersland, L., Smievoll, A. L. F. I., Neckelmann, D. A. G., & Hugdahl, K. (2006). MR Morphometry Analysis of Grey Matter Volume Reduction in Schizophrenia: Association with Hallucinations. International Journal of Neuroscience, (116), 9–23. https://doi.org/10.1080/00207450690962244
• Nenadic, I., Smesny, S., Schlo, R. G. M., Sauer, H., & Gaser, C. (2010). Auditory hallucinations and brain structure in schizophrenia: voxel-based morphometric study. The British Journal of Psychiatry, 01, 412–413. https://doi.org/10.1192/bjp.bp.109.070441
• O’Daly, O. G., Frangou, S., Chitnis, X., & Shergill, S. S. (2007). Brain structural changes in schizophrenia patients with persistent hallucinations. Psychiatry Research – Neuroimaging, 156, 15–21. https://doi.org/10.1016/j.pscychresns.2007.03.001
• Owen, M. J., Sawa, A., & Mortensen, P. B. (2016). Schizophrenia. Lancet, 86–97. https://doi.org/10.1016/S0140-6736(15)01121-6
• Perlini, C., Bellani, M., & Brambilla, P. (2012). Structural imaging techniques in schizophrenia. Acta Psychiatrica Scandinavica, 235–242. https://doi.org/10.1111/j.1600-0447.2012.01868.x
• Shapleske, J., Rossell, S. L., Chitnis, X. A., Suckling, J., Simmons, A., Bullmore, E. T., … David, A. S. (2002). A Computational Morphometric MRI Study of Schizophrenia: Effects of Hallucinations. Cerebral Cortex, 1331–1341.
• Tang, J., Liao, Y., Zhou, B., Tan, C., Liu, W., Wang, D., … Chen, X. (2012). Decrease in Temporal Gyrus Gray Matter Volume in First- Episode, Early Onset Schizophrenia: An MRI Study. PLoS ONE, i(7). https://doi.org/10.1371/journal.pone.0040247
• Tardif, C. L., Joëlle, C., John, C., Bazin, P., Schäfer, A., Schaefer, A., … Villringer, A. (2016). Advanced MRI techniques to improve our understanding of experience-induced neuroplasticity. NeuroImage, 131, 55–72. https://doi.org/10.1016/j.neuroimage.2015.08.047
• van Tol, M., Meer, L. Van Der, Bruggeman, R., Modinos, G., Knegtering, H., & Aleman, A. (2014). Voxel-based gray and white matter morphometry correlates of hallucinations in schizophrenia: The superior temporal gyrus does not stand alone. NeuroImage: Clinical, 4, 249–257. https://doi.org/10.1016/j.nicl.2013.12.008
• Vita, A., Peri, L. De, Deste, G., & Sacchetti, E. (2012). Progressive loss of cortical gray matter in schizophrenia: a meta-analysis and meta-regression of longitudinal MRI studies. Translational Psychiatry, 2(11), e190-13. https://doi.org/10.1038/tp.2012.11