Diffusion Tensor Imaging
DTI provides a measure of the orientation (vector) of water diffusion along the axis of gross anatomical compartments.60
Mean diffusivity (MD) is a quantitative measure of directionally averaged diffusion, while fractional anisotropy (FA) is a normalized measure of diffusion anisotropy and ranges from 0 (random diffusion) to 1 (unidirectional diffusion). FA permits the representation of data in 2-dimensional maps () and allows data analysis with “traditional” analysis procedures.61
The sensitivity of the method is such that it allows an estimate of the “structural integrity” of coherent brain structures like subcortical WM tracts. Basically, reduced anisotropy can reflect a variety of phenomena and thus impaired structural integrity can mean different things, and therefore, the biologic determinants of diffusion parameters in WM are not yet fully understood. For instance, anisotropy has been shown to increase during myelination62,63
and to decrease during demyelinative processes64
suggesting that the degree of axon myelination substantially contributes to this measure. On the other hand, there is also evidence that the number or density of axon fibers and the coherence of the fibers influences anisotropy.65
For instance, several DTI studies have shown significantly reduced FA in amyotrophic lateral sclerosis (ALS) that is thought to be caused by axonal degeneration of pyramidal tracts.66,67
FA images; coronal, axial, and saggital orientation (resolution 256 × 256 × 64).
Accordingly, any finding of a FA reduction is expected to have a different neurophysiological impact depending on the cause of the reduction. When the FA reduction is due to a reduced number of axons in a fiber tract, there are simply less nerve impulses transferred, ie, less information is carried. However, when axonal myelination is impaired, a slowdown of conduction velocity of nerve impulses in the affected axons is seen because current is lost through the membrane.68
Across the fiber bundle, one would expect that impaired myelination should result in variable axonal conduction properties, ie, variable speed of impulse transmission, between neighboring and differentially affected axons, whereas a lower number of axons should only carry less information in the fiber bundle. The downstream effects at the synaptic level of these 2 alternative causes for FA reduction should also be different. Local field potentials (LFPs) are thought to reflect the weighted average of synchronized dendrosomatic components of pyramidal synaptic signals (postsynaptic potentials) and they are the physiological basis of synchronized electromagnetic oscillations and blood-oxygen level–dependent response.69
Although our current knowledge is still quite sketchy on this issue, one would theoretically predict that a lower number of axons with a decrease of nerve impulses would also decrease the LFP amplitude because less synchronous activity is present (lower “downstream activation”). In contrast, the logical consequence of a higher variability of axonal impulse transmission should be a failure to synchronize postsynaptic activity (more “downstream noise”). In this context, it may be of some interest that previous work from our group provided some evidence on the basis of electrophysiological and functional MRI studies that schizophrenia patients fail to synchronize brain activity.70–72
However, apart from myelination deficits, there are other pathophysiological mechanisms that may also account for this deficit that has been discussed by us elsewhere in more detail.73
Most DTI studies applied the single-shot echo-planar imaging method for data collection. A recent study, however, used the line-scan diffusion imaging that offers the advantage of not being as sensitive to susceptibility variations and chemical shift effects.20
For image analysis, 2 principle methods have been adopted: the ROI and the whole-brain VBA. The ROI method allows a more powerful examination of regions but there is a risk of a systematic placement bias. The VBA has still problems with image analysis, especially normalization algorithms, that were originally not designed for anisotropy images.74
In schizophrenia, a number of DTI studies have suggested reduced integrity of subcortical WM tracts (). DTI studies in schizophrenia patients have found reduced anisotropy, compared with healthy control subjects, in prefrontal WM,75
in the splenium of the corpus callosum,76,77
in whole-brain WM,94,95
in the anterior cingulum,81,87
in the left arcuate faciculus,17
bilaterally in the cingulum,19
bilaterally in the middle cerebellar peduncles.86
In addition, some investigations demonstrated reduced asymmetry in WM tracts in the uncinate fasciculus80
and the anterior cingulum.87
On the other side, several other studies did not find reduced FA in schizophrenia patients.16,78,84
In general, there is considerable inconsistency between findings—perhaps because of different sensitivity and resolution of the applied measurement technology and their vulnerability to various artifacts.
DTI Studies in Schizophrenia, Study Characteristics, and Results
As DTI findings in schizophrenia are of particular relevance with regard to the concept of structural disconnectivity, some of the most important DTI studies in schizophrenia will be briefly summarized at this point. Kubicki et al.80
investigated the uncinate fasciculus, the largest of the 3 WM tracts connecting temporal and frontal lobes. Schizophrenia patients showed a lack of normal left-greater-than-right asymmetry that was interpreted as an indication for frontotemporal disconnectivity in schizophrenia. The same study group performed an ROI investigation of the cingulum, most important cerebral fiber bundle, and showed smaller volume and reduced anisotropy in schizophrenics using line scan DTI.19
Burns et al.17
performed the study with the largest sample size to date (30 patients, 30 control subjects): FA was determined in the 3 most important subcortical fiber tracts: the anterior cingulum, the uncinate fasciculus, and the arcuate fasciculus, and analysis was based on an optimized voxel-based morphometry technique. Reduced FA was found in the left arcuate and the left uncinate fasciculus, suggesting frontotemporal and frontoparietal structural disconnectivity in schizophrenia. Two studies of WM integrity in schizophrenia used a VBA of FA of the whole brain, but both with limited sample size: Agartz et al.77
demonstrated bilaterally reduced FA in the splenium of the corpus callosum and the adjacent WM (forceps major) in schizophrenics while examination of volumes did not show any differences between the 2 groups (20 patients, 24 controls). Ardekani et al.82
examined 14 patients and 14 control subjects and showed lower FA in the corpus callosum, left superior temporal gyrus, parahippocampal gyri, middle temporal gyri, inferior parietal gyri, medial occipital lobe, and the deep frontal perigenual region. The authors emphasize that their findings in these regions are consistent with other reported structural abnormalities in schizophrenia. Wang et al.84
examined superior and middle cerebellar peduncles as main pathways of neural fibers in the cerebellum in 29 patients and 20 healthy controls with ROI analysis and did not find any difference in FA and MD in these regions.
Foong et al.76
performed a ROI-DTI study of the corpus callosum in 20 schizophrenia patients and 25 healthy controls and found a significant reduction of FA while MD was significantly increased in schizophrenics in the splenium but not in the genu of the corpus callosum. The authors suggested that these DTI findings might demonstrate a disturbance of the commissural connectivity in schizophrenia. The same study group investigated FA and MD in 6 slices of WM using a VBA.16
They did not find any differences, but it must be mentioned that sample size (14 patients/19 controls) and number of slices (6) was limited. An interesting DTI study was just recently published by Jones et al.96
They found that age affected DTI-based measures in schizophrenia patients in a different way from comparison subjects, most notably in the left superior longitudinal fasciculus. The youngest schizophrenia patients had lower diffusion anisotropy than age-matched comparison subjects, but this difference diminished with increasing age. The main conclusion of this study was that direct comparisons of absolute DTI-based measures between individuals with schizophrenia and comparison subjects may be problematic and misleading because of underlying age-related differences in brain maturation between groups. Salat et al.97
performed a DTI study investigating age-related WM changes in 38 subjects across the adult age span, showing age-related decline of FA mainly in frontal WM regions, while temporal and parietal WM was relatively preserved. Pfefferbaum et al.98
compared 10 younger and 10 older individuals, and DTI analysis confirmed that the age-related FA decline was most pronounced in frontal WM. These findings support the hypothesis that WM abnormalities in chronic schizophrenia patients might be in part caused by pronounced age-related changes, neurodevelopmental pathology like dysregulation of WM maturation and myelination might be less important in pathogenesis of schizophrenia. Long-term medication effects may also play an important role as antipsychotic medication effects on brain morphology has been shown: for instance, Lieberman et al.58
demonstrated significant decreases in gray matter volume in haloperidol-treated patients but not in olanzapine-treated patients.
To date, only a small number of DTI studies in first-episode schizophrenia have been performed:92,99–101
2 ROI studies investigated WM changes in hippocampi99
and the corpus callosum100
and did not find group differences in FA, which would suggest that neuropathological abnormalities may appear later and be progressive. Szeszko et al.92
performed a VBA study demonstrating lower FA in patients in WM of the left middle frontal gyrus and left posterior superior temporal gyrus, suggesting that WM abnormalities are already present in the very early course of the illness. Another voxel-based investigation showed decreased FA in several WM areas bilaterally.101
Kumra et al.88
investigated several ROIs in early onset schizophrenia patients and found reduced FA in the frontal WM bilaterally and in the right occipital WM. In total, these 4 cited studies present contradictory results; additional studies with higher sample sizes are needed to demonstrate whether there is already WM pathology present in the initial state of the illness. There is evidence for gray matter abnormalities in early states of the illness from several functional and structural imaging studies.102–105
This supports the hypothesis of structural WM pathology as a secondary degeneration caused by neural dysfunction in the course of the illness.
A systematic review of 19 DTI studies in schizophrenia investigated WM differences between schizophrenia patients and control subjects:74
They concluded that DTI studies of schizophrenia patients to date have not yet provided consistent findings of WM abnormalities. Group differences might be too slight to be demonstrated reliably and sample sizes of all studies were relatively small (median 15 patients) with relatively high variance of the obtained findings. Additionally, it was pointed out that there was no common area examined in most of the studies. The most frequently investigated area, the corpus callosum, showed reduced FA in 4 studies and normal FA in 4 other studies. In the cingulum, 4 studies demonstrated reduced FA and 5 did not. Kanaan et al. emphasize that effect size was shown in only 8 studies that does not allow a meta-analysis of the mentioned studies.
Kubicki et al.21
reviewed 18 DTI studies, and pointed out that decreased FA and increased MD within prefrontal and temporal areas and within the fiber tracts between these regions (cingulum, arcuate, and uncinate fasciculus) are the most frequent findings. The authors emphasize the role of DTI-neuropsychological correlations. Accordingly, correlations between anisotropy in prefrontal WM and negative symptoms, impulsiveness and aggressiveness have been shown. Anisotropy in frontotemporal connections have been demonstrated to correlate with measures of errors in executive functions. In another recent review about the evidence for WM abnormalities in schizophrenia including imaging, histological, and genetic findings, Kubicki et al.106
pointed out that discrepancies reported in DTI studies may be caused by the relatively low resolution and high noise of today's diffusion scans—compared with structural MRI.
As the contribution of possible confounding factors like treatment, age, individual variation is largely unknown, most of the DTI studies in schizophrenia appear to be underpowered; in addition, effect sizes were not provided in most cases. In this context, another critical issue is the choice of the applied analysis procedure. Most of the above mentioned studies used ROI methodology that is particularly difficult in regions of fiber tracts even if ROI methodology may be more sensitive in large fiber tracts that can be easily identified. However, it might be less suitable for WM regions due to variability and directionality of fibers. On the other hand, VBA is limited due to normalization and smoothing processes but does not have the problem of bias due to the choice of location of ROIs. Four of the mentioned DTI studies in schizophrenia adopted a voxel-based registration of the whole brain but then analyzed specific regions.16,17,75,90
These issues may in part explain why DTI studies in schizophrenia to date are inconsistent (). A promising new approach to the analysis of diffusion data is the tract-based spatial analysis107
that attempts to bring together the advantages of VBA and ROI analysis by a carefully tuned nonlinear registration, followed by projection onto an alignment-invariant tract representation, the “mean FA skeleton” (). Another methodological issue in analysis of DTI data is the size of the smoothing filter, which is highly variable across the studies. The effect of varying filter sizes has been investigated using DTI data from schizophrenia patients and controls, and within the range of smoothing from 0 to 16 mm, 4 different results have been demonstrated.108
Further investigations with standardized methodology of data acquisition and analysis and larger sample sizes are needed to better understand which fiber tracts are particularly affected in schizophrenia.
Tract-based spatial statistics WM skeleton superimposed on FA template; coronal, axial, and saggital orientation.
Magnetization Transfer Imaging
MTI is a promising new MR technique to improve contrast and tissue characterization. It visualizes the weakly bound protons in macromolecular structures that are normally invisible because of their short relaxation times.109
The mechanisms for magnetization transfer (MT) include the chemical shift of tightly bound protons with water, dipolar cross relaxation, and diffusion of water from the surface of macromolecules into the bulk environment.110
When saturation from the bound protons is transferred, signal from the pool of free protons decreases, leading to lower signal intensity in MR images. Magnetization transfer ratio (MTR) is an index for signal loss derived from the exchange of magnetization between bound protons and free water and is postulated to be a marker of macromolecular structure integrity.
In brain parenchyma, myelin protons may account for a substantial portion of the observed measurement effects.21
Studies of neurological diseases have shown large reductions of MTR in neurological diseases with myelin loss like multiple sclerosis, progressive multifocal leucoencephalopathy, and central pontine myelinolysis.111
Therefore, MTI is proposed as a promising approach to the diagnostics of neurological diseases.112
Current MTI technique is highly dependent on pulse sequence parameters that makes any comparison of findings between study sites difficult. Therefore, several research groups currently try to develop a standardized method for measuring MTR on MR images from different manufacturers. Davies et al.111
reported on a new quantitative MT technique that is less dependent of MT acquisition protocol and may provide more information about myelin affection in WM.
So far, several investigations using MTI have been conducted with schizophrenia patients, however, again providing inconsistent findings (). Foong et al.15
performed the first MTI study in schizophrenia patients, using ROI methodology and general linear mixed modeling in WM regions in 25 patients and 30 controls. They showed several WM MTR reductions bilaterally in the temporal lobe, predominantly in the middle temporal gyri. In the same subject groups, Foong et al.109
also used VBA for MTR maps and found widespread MTR reductions in the frontal and temporal cortex in schizophrenics compared with controls.
MTI Studies in Schizophrenia, Study Characteristics, and Results
Kubicki et al.20
intended to further specify DTI findings in schizophrenia by combing DTI with MTI. Line-scan diffusion imaging and T1-weighted sequences with and without saturation pulse were used to examine 21 chronic schizophrenia patients and 26 control subjects using VBA. Diffusion anisotropy and MTR maps demonstrated changes in different regions; decreased MTR was seen in several fiber tracts, including the right posterior cingulum bundle, corpus callosum, internal capsule, fornix, and superior occipitofrontal fasciculus. FA and MTR showed highest correlations in the left cingulum, corpus callosum, and fornix. The authors postulated that WM abnormalities observed in both DTI and MTI may indicate the presence of myelin abnormalities while DTI findings in regions without MTI abnormalities may be caused by differences in the organization and coherence of fibers and not by myelin pathology.20
Bagary et al.18
acquired high-resolution volumetric T1-weighted images and MT images from 30 schizophrenia patients and 30 controls, using a voxel-based whole-brain analysis. MTR was found to be reduced bilaterally in the medial prefrontal cortex (PFC) (right greater than left), insula (left greater than right), and WM incorporating the uncinate fasciculus (left greater than right) while—at the same time—significant volume differences have not been found.
Kalus et al.22
performed a multimodal MRI study including high-resolution volumetry, DTI, and quantitative qMTI in 14 patients and 14 controls using an ROI tracing of the amygdala. DTI and quantitative MTI parameters T1 and T2b (T2 relaxation time of restricted protons) of the amygdala showed significant changes in schizophrenics while the normalized amygdala volumes and the semiquantitative MTR, which was used in most of previous studies, did not show differences between groups.
Another ROI investigation of hippocampal subregions using semiquantitative MTR and quantitative MT parameters was conducted by Kiefer et al.114
Significant differences of quantitative MT parameters were demonstrated between the subregions of the hippocampal formation, possibly due to their structural and functional diversity while no significant group differences between 14 schizophrenia patients and 14 control subjects have been observed.
When MTR and DTI are conducted together, both methods may differentiate between WM fiber tract abnormalities that are due to abnormalities deriving from alterations in axonal coherence, density or thickness (when only FA is abnormal), and myelin sheath thickness or composition, when both FA and MTR are affected.20
This is a promising new approach to investigate neuropathological correlates of disconnectivity and needs further multimodal MRI examinations, including animal studies and human postmortem studies.
Magnetic Resonance Spectroscopy
Postmortem studies indicate reductions in phospholipids and essential fatty acids in the brain of schizophrenia patients.115
Phosphorus magnetic resonance spectroscopy (31
P-MRS) quantifies the resonances of phosphomonoesters (PME), phosphodiesters (PDE), inorganic orthophosphate, phosphocreatine (PCr), and nucleoside adenosine triphosphate (ATP). It can provide important information about high-energy metabolism and neuronal membranes, such as levels of PME that reflect the building blocks of neuronal membranes and PDE that reflect breakdown products.
A large number of 31
P-MRS studies in patients with schizophrenia have been performed, demonstrating a variety of alterations in neuronal membrane biochemistry. Several review articles provide an overview of 31
P-MRS study results in schizophrenia.116–120
Results of recent 31
P-MRS investigations in schizophrenia, not taken into account in the cited review articles, are represented in . However, the results of 31
P-MRS studies in schizophrenia are not consistent, perhaps related to the variations in medication, phase of illness and differences in MRS methodology.118
A decrease in the level of PME and an increase in the level of PDE have been demonstrated in the prefrontal lobe of neuroleptic-naive schizophrenia patients. Study results in medicated schizophrenia patients were less consistent and have shown mostly decreased PME and/or increased PDE.120
PDE were found to be elevated in the temporal lobes of neuroleptic-naive schizophrenia patients, but data about the temporal lobes of medicated schizophrenia patients have not been consistent. Except for the reduction in the ATP in the basal ganglia and the increase in the frontal lobe PCr, data related to changes in high-energy phosphates in 31
P-MRS are contradictory.130
31P-MRS Studies in Schizophrenia, Study Characteristics, and Results
Taken together, results of 31
P-MRS have been interpreted as reflecting a relative increase in cell membrane degradation in prefrontal cortical regions at certain phases of schizophrenia.119
This supports the hypothesis that decreased myelination and alteration of structural connectivity in schizophrenia might be related to decreased lipid membrane components.131
Issues of sensitivity, specificity, measurement reliability, and functional significance of the MRS findings need to be further clarified.118