Updated: Oct 4, 2019
Schizophrenia is a mental disorder that is thought to affect close to 1% of the worlds population1. It is thought to occur more commonly in men rather than women at a rate of around 1.4 times2. Its economic and social costs are exceptionally high, treatment of the disorder equates to 2 billion pounds whilst indirect costs are around 4.7 billion pounds. This totals an estimated 6.7 billion pounds a year in England alone.
These huge costs and the severe mental suffering that sufferers experience makes it an incredibly important illness to understand and ultimately identify biomarkers for the disease with the aim to develop more effective methods of treatment and a cure. Schizophrenia actually means the ‘splitting of mental functions’ and was first described by Kraepelin and Bleuler. They coined the term ‘dementia praecox’ to describe the disorder and strongly believed that brain abnormalities were to blame for ‘schizophrenias’4. These views led to many academics devoting whole careers to trying to identify such neuropathological abnormalities but to no real avail. This belief therefore remained almost entirely unfounded until the development of modern day imaging techniques. These techniques, which include computer-assisted tomography (CT) scans and magnetic resonance imaging (MRI) techniques, have enabled the observation of brain structure and function in schizophrenia sufferers. Despite this there is yet to be strong clinical evidence of a single isolated source of the disorder that can be used in its diagnosis. Instead schizophrenia is diagnosed based on a myriad of psychiatric symptoms.
Although there is no clear cause of schizophrenia or indeed it action on the brains of those that have been diagnosed, one hypothesis is widely proposed as being the most likely source. This is the neural network ‘disconnection’ hypothesis. It suggests that schizophrenia comes about as a result of dysfunctional integration of a set of regions within the brain; these are misconnected leading to problems with basic coordination of mental processes.
The advances in technologies over the last two decades have played a foundational role in delineating the neural mechanisms of schizophrenia. There have been a host of studies that have used techniques such as the following to be able to do so:
Functional Magnetic Resonance Imaging (fMRI)
Structural Magnetic Resonance Imaging (sMRI)
Diffusion Tensor Imaging (DTI)
Electro-Encephalography (EEG) and Magneto-Encephalography (MEG)
Genetics – Genome Wide Association Studies (GWAS)
fMRI has been used regularly in the last decade by Pearlson, Curtis et al and recently Yang et al5–7. These studies have highlighted variations in the dorsolateral prefrontal cortex and temporal lobe in the brains of schizophrenics and thus indicate a level of association with the disease5. sMRI is a technique that allows the identification of tissue type within the brain itself. These tissue types are characteristically: gray matter (GM), white matter (WM) and cerebrospinal fluid (CF). The typical findings of this field of imaging is that GM is locally reduced in those whom display schizophrenic tendencies8. This reduction can however be interpreted in a number of ways from the consequence of lower cortical thickness levels or altered gyrification patterns. There is also anisotrophical reduction in WM in the corpus callosum, which may be linked to decreased connectivity of brain regions8. DTI by contrast works more similarly to fMRI and has been utilised to demonstrate changes in the integrity of several regions of the brain implicated in schizophrenia; for example the uncinate fasciculus and anterior thalamic radiation regions9,10. EEG is similar to fMRI but measures the electrical activity in the brain at a lower spatial resolution but a much larger temporal resolution. Whilst MEG also measures electrophisiological endophenotypes related to gamma band activity, a major new candidate for one area affected by schizophrenia, but uses magnetic fields rather than electrical to generate data11. These approaches all us direct methods for imaging the brain in the quest to discern neurobiological markers for schizophrenia. GWAS by contrast is looking for a genetic link. It is attempting to pinpoint specific changes in gene expression that may be potential triggers for the development of schizophrenia. It has already been reported that ANK3, MIR137 and DISC1 plus may others all show alterations in expression in subjects with varying onsets of the disorder.
Each of these approaches has its advantages and disadvantages. These are often related to the narrow area of the brain that is being observed. Having a narrow focus means that, especially for schizophrenia which is such a broad and far reaching mental disease, there are always pieces of the jigsaw that are ignored when using just one type of investigative tool. For this reason recently it has become commonplace to incorporate the use of multiple imaging methods to gain differing views of the brain of the same individual at the same time10. This new way of viewing the abnormalities, which could cause schizophrenia, is called the multimodal approach.
Schizophrenia is a mental illness that is predominately thought to affect cognition, but has also been shown to cause chronic issues with emotional and behavioural wellbeing10. As a disorder its many facets can result in other problems with mental health. It is believed that most sufferers also suffer from other comorbid conditions like depression and anxiety. This can ultimately lead to substance abuse, social welfare and other social or personal problems2,9. The implication of these many areas of mental health that are affected by the disease means that an appropriate multimodal methodology may be the best way to further learn about this disorder.
The benefit of a multimodal method is that it allows the assessment of different views of the brain simultaneously, be it brain function or structure. The ability to take advantage of this cross-information knowledge generated by multiple imaging techniques opens up new doors in the discovery of misconnected or dysfunctional sections of the brain that could eventually be used as biomarkers for diagnosis and treatment10,12. This increased efficiency in answering clinically and scientifically important questions means that the challenges faced in obtaining this type of data is worth it.
As mentioned multimodal methods are not easy to carry out, and can be incredibly complex and laborious to process. The collected image data can be extremely noisy and may be contain only a small number of subjects10. These difficulties however are outweighed by the diversity of the results uncovered. Evidence exists across all modalities for abnormalities that may lead to or be linked with the onset and progression of the disease. There is no ‘gold standard’ identifier in structural, functional or genetic areas. This indicates strongly that the need to combine these modalities is key to defining a ‘gold standard’.
The types of imaging methods previously outlined can be combined in multiple ways to achieve contrasting and enlightening results. I will review several combinations in the following paragraphs.
Function - Function
i) fMRI and EEG – This coupling is the most used example of multimodal fusion techniques and their applications in schizophrenia research. It combines two functional imaging methods in order to provide a more precise and accurate functional analysis of the brain13. fMRI gives the most exact spatial resolution and EEG provides the temporal precision necessary to draw meaningful conclusions about the neuronal network but also the haemodynamic activity of the brain14. There are limitations to this approach though brought about through a host of unrealistic assumptions that are made by each modality when they are constrained to one another. One way of getting around this fixed approach is through the use of multimodal canonical correlation analysis (mCCA) allowing the two modalities to have their own matrix, which isn’t constrained by the other. Even so the most effective way to use this multimodal approach to discriminate the functions of brain regions between schizophrenics and controls is by applying sMRI as well.
Structure - Structure
ii) GM and WM – When performing sMRI it was previously used to assess GM or WM individually. However if run in tandem the structure of both can be compared. Using a technique known as joint independent component analysis (jICA) it has been possible to identify GM and WM linked regions of the brain16. These regions signify a large proportion of the neuronal network of the brain and are thus give a good indication of the extent to which the disease affects the brain. jICA in this context as with the majority of multimodal models means joint analysis of GM and WM linked regions can be performed thus allowing the unified structure of a jICA model to highlight any major group differences10,16. Contrastingly taking a different route but using the same initial technique it is able to carry out the imaging of GM and WM via structural phase and magnitude methods. The latter distinguishes the contribution of both matter types, while the former gives the entire tissue concentration. This technique has been used to provide evidence for a significantly lower WM to GM concentration in the brains of schizophrenics.
Function – Structure
The utilisation of a functional imaging technique and a structural analytical method means that there is a potential to gather in depth information about the connections that may be miss firing in the brain of those in the grips of the disorder.
iii) fMRI and GM – Studies exploring the use of fMRI combined with structural sMRI analysis of GM have found differences between control groups and groups of patients of the illness. The findings were consistent with previous work looking at the levels of GM in sufferers but have also uncovered a relationship between increased functionality in motor areas but less activity in the GM temporal areas17. This multimodal technique is at present not well utilised but there is promise in it unearthing previously unrecognised aberrant functional – structural links in the brain of a patient.
iv) fMRI and DTI – In this instance fMRI is used as a framework for the newer derived DTI technique. fMRI and sMRI studies have contributed hugely to the hypotheses that the brain of schizophrenics is disordered. Also they have been behind the recent push in research that is pushing the idea that disruption in connectivity between regions of the brain is a more fundamental cause of schizophrenic cognitive issues rather than actual abnormalities of the regions themselves. The inclusion of DTI in this analysis will thus prove crucial. DTI is a capable of visualising the structure and functional organisation of WM tracts between brain regions, something that MRI’s cannot do with regular precision. One study has suggested that in some cases connectivity in the brain of schizophrenics is not only lower but also completely different18. This is in conjunction with upward of twenty studies have been carried out in the last few years citing the use of these techniques to investigate the integrity of WM tracts whilst also now supporting previously supported correlations between anatomical abnormalities and circuitry.
Imaging – Genotyping
Genetic factors undoubtedly play a key role in the development of function and structure in the brain of healthy and dysfunctional individuals. The following multimodal approaches, which combine imaging strategies like MRI’s and genetic studies, have added to the discoveries in the sphere of schizophrenia. By combining two classically differing aspects the power that it produces for classification of the disease is almost unrivalled7. It is because of this that imaging genomics is becoming an area of research that is evolving rapidly.
v) fMRI and SNP – An SNP is a single nucleotide polymorphism they typically occur in non coding DNA regions. However in GWAS they have been shown to occur in the coding regions of specific genes associated with schizophrenia. fMRI run in a parallel ICA setting lays out the system to test the brains dysfunctional regions in patients with the SNPs that have been identified using GWAS. Liu et al took an fMRI of the right temporal lobe, bilateral frontal lobe and parietal lobe and compiled the data with that of 10 relevant SNPs shown to be contributing to various genes linked to schizophrenia including DISC119. This study was only a conceptual study to show the principle behind using phenotypic imaging to investigate genotypic factors. More recent studies have furthered this and implicated up to 25 risk genes containing SNP markers in the function of directed behaviour, attention, memory and several other psychiatrically relevant areas.
vi) sMRI and SNP – This fusion of sMRI and DNP markers works in a similar way but instead of looking at function focuses on the structure of GM in relation to specifically chosen genes. One group of researchers looked at reading disability related genes such as DCDC2 and TTRAP in the attempt to explain reading and language difficulties that is typical in many forms of schizophrenia. The study pinpointed 5 SNP-GM relationships. In particular the brain regions superior prefrontal, occipital and temporal lobes were closely integrated with DCDC2 expression.
Long term CT and MRI monitoring of patients with schizophrenia has shown that brain abnormalities are present at early onset be it in adolescence or adult years. These abnormalities do not progress or worsen during the course of the illness22. This observation supports the view that unlike neurodegenerative disorders such as Alzheimer’s, schizophrenia is not a degenerative condition but instead a developmental one. The developmental link highlighted through GWAS studies and genomic imaging studies implies a degree of genetic control during development. However this is also likely to be influenced by environmental factors. Utilising a single method to identify a cause therefore will ultimately be in vain, which is why multimodal fusion research approaches are so critical for the future of this disease. They allow for a greater control of the parameters involved with the disorder both on a molecular and physiological level but also in terms of age and other variables. This is by no means the only route that may aid in explaining and understanding this mental health issue. Other multivariate methods have been used in the imaging of brain disorders, like the afore mentioned Alzheimer’s disease10. These have yet to be applied to the topic of schizophrenia and this may well be the next path to explore.
Thus to summarize it is clear that the use of multimodal models in the diagnosis and understanding of schizophrenia is of importance. It can potentially be used as stepping-stone in the treatment and controlling of the disorder. The fact that schizophrenia has many varying prongs means that it is a perfect test disorder for these fusion techniques. However it is vital that the correct choice of multimodal model is used for the specific study in order to prioritize the observations based on the limitations and advantages of each model. Unfortunately it has become increasingly evident through recent studies that the ability to delineate this illness will likely be impossible as schizophrenia appears to be more of a category of mental illnesses that are on a sliding scale and thus all stem from different biological or neurological origins. That being said with modern day technology the classification of multiple disorders under the banner schizophrenia is more and more likely. So with persistence and care more clues to such a mysterious mental disorder are waiting to be found.