Neuroimaging of Bipolar Disorder: Emphasis on Novel MRI Techniques


Address correspondence and reprint requests to Dr. S.D. Bruno at Institute of Neurology, Queen Square, London, U.K. E-mail:


Summary:  The explosion of neuroimaging in the last decade has contributed to a revision of the concept of bipolar disorder, which was traditionally seen as a cyclic illness with return to normality between episodes lacking “by default” permanent brain abnormalities. A conspicuous corpus of neuroimaging and neuropathology studies indicates that brain changes occur in bipolar disorder, although their etiology and their relation to clinical features are yet to be established. This review summarizes the results from magnetic resonance imaging studies using conventional and novel techniques in patients with bipolar disorder.

This summary of neuroimaging findings in bipolar disorder (BD) is structured in two parts. The first part deals mainly with conventional structural magnetic resonance imaging (MRI), and the second part is devoted to novel MRI techniques and their application to the study of BD.


Whereas the presence of generalized, nonspecific abnormalities such as ventricular enlargement and sulcal prominence has been confirmed by meta-analyses of structural MRI studies (1), converging evidence from structural and functional neuroimaging studies (2) suggests a more specific role for areas of the brain known to be involved in the regulation of mood and emotions, such as the prefrontal cortex, the anterior cingulated, and medial temporal structures. Reduced grey matter volume and decreased cerebral blood flow and metabolism in the left subgenual anterior cingulate have been described in familial BP patients (3). The presence of similar changes in familial first-episode affective psychosis (4) suggests that, rather than being a product of chronicity, they may be predisposing to the development of the disorder. Kruger et al. (5) also identified medial frontal cortex structures as possible sites of disease vulnerability by using positron emission tomography (PET). These findings are consistent with neuropathologic postmortem studies, which in the anterior cingulate have reported reduced glial density and number (6); reduced density of nonpyramidal neurons in lamina II (7), and reduced neuronal density and decreased depth of laminae III, V, and VI (8). Similar changes have been described in dorsolateral and orbital prefrontal cortices (9,10).

The abnormalities in the temporal lobes are less clear-cut and, although some (11) have found smaller volumes in BP patients compared with controls, others have found the opposite (12,13) or no differences between the two (14). The amygdala has been object of great interest, given its role in emotional regulation, but findings have been discordant. Both enlarged (15–17) and smaller (18,19) amygdala volumes have been reported.

Neuropathological studies of the hippocampus have described reduction in the number of nonpyramidal neurons (20) and decreased arborisation of subicular apical dendrites (21), the latter indicative of reduced synaptic activity.

With regard to white matter changes, much effort has been devoted to clarifying the role of white matter T2 hyperintensities in bipolar disorders (22). These are periventricular and subcortical areas of higher signal intensity on T2 MR images that are more frequent in BD patients than in schizophrenic patients and in those with other affective disorders but occur also in unipolar depression (23) and late-onset psychosis (24). They correlate to cardiovascular risk factors and have been described in BD particularly in patients with late onset (25) but also in patients at their first hospitalization (26). They have been associated with poor outcome (27) and cognitive deterioration (28).

Although the general picture emerging supports structural and functional abnormalities of the prefrontal cortex, the anterior cingulate, and temporal lobe structures, conflicting findings make compelling the use of novel MRI techniques, characterised by higher sensitivity and innovative analysis approaches for further investigation. Three techniques increasingly used in psychiatric research are diffusion tensor imaging (DTI), magnetisation transfer imaging (MTI), and voxel-based morphometry (VBM).


Diffusion tensor imaging

DTI (29) is based on the property of water molecules to move randomly in all directions when unrestricted, according to a “diffusion coefficient” expressing the magnitude of the movement. In brain tissue, the presence of cellular structures hinders the motion of the molecules, which, for example, in axons will be mainly longitudinal. In MRI, the diffusion coefficient influences the amount of signal, and it is therefore possible, by application of large magnetic field gradients, to use diffusion as an image-contrast mechanism, producing a number of parameters such as apparent diffusion coefficient, mean diffusivity, and fractional anisotropy. Loss of structure due to a pathologic process will produce an increase in mean diffusivity and a decrease in fractional anisotropy.

In psychiatry, DTI has been recently applied mainly to the study of schizophrenia (30), also investigated by our group (31a,31b). A study on BD using a region-of-interest approach has recently appeared (32), showing reduced fractional anisotropy in the prefrontal white matter of nine BD patients compared with healthy controls.

Magnetisation transfer imaging

MTI (33) is based on the possibility of preferentially saturate protons bound to macromolecular structures by using an off-resonance radiofrequency pulse. In brain tissue, water can be considered to be divided in two compartments: free water (intracellular and extracellular), having a long T2 (80 ms) and visible with conventional MRI, and bound water (cell membranes in grey matter and myelin in white matter) with a very short T2 (0.02) and therefore MRI invisible. Free water has a narrow spectral line, whereas bound water has a broad spectral line. By applying a radiofrequency pulse at a frequency inside the broad line but outside the narrow line, bound water is partially saturated, leaving the free water unaffected (34). This leads to a reduction in signal intensity, which is dependent on macromolecular density. The degree of signal loss, measured in percentage units (pu), is expressed as magnetization transfer ratio (MTR). MTR reduction correlates with myelin and axonal loss in postmortem tissue from spinal cord (35) and brain (36) in multiple sclerosis, experimental allergic encephalitis (37), toxic demyelination (38), Wallerian degeneration (39), and experimental brain injury (40). In vivo white matter MTR decreases occur in conditions characterized by myelin and/or axonal loss [e.g., multiple sclerosis (41), central pontine myelinolysis (42), cerebral ischaemia (43), systemic lupus (44), human immunodeficiency virus infection (45), and traumatic brain injury (46)]. Gray matter abnormalities have been studied less extensively, but Wallerian degeneration triggered by distant axonal damage and microscopic lesions are thought to explain cortical MTR reductions in multiple sclerosis (47). MTR reductions have been reported in normal-appearing grey matter in multiple sclerosis (48) and progressive supranuclear palsy (49). MTR correlates with in vivo measurements of N-acetylaspartate, a marker of neuronal integrity (50).

Foong et al. (51) reported widespread cortical MTR reductions in chronic schizophrenic patients, and more recently, MTR reductions have been reported in the insula and medial prefrontal cortex in first-episode schizophrenic patients (52). In BD, Bruno et al. (53) used a voxel-by-voxel approach with small volume corrections, which involves the placement of a “sphere of interest” in a region of “a priori” relevance, and demonstrated lower MTR in the right prefrontal subgenual cingulate cortex and surrounding white matter in BD patients compared with healthy matched controls (Fig. 1). These abnormalities, which can be detected in the absence of volumetric changes, may reflect changes in the neuropil density (54).

Figure 1.

Area of significantly lower magnetization transfer ratio (MTR) in the right prefrontal subgenual cingulate cortex and subgyral white matter in bipolar patients compared with healthy controls projected on orthogonal sections through a T1–weighted averaged image. The orthogonal lines cross at MNI coordinates (16, 24, -10) corresponding to the greatest MTR reduction (from ref. 53, with permission.)

Voxel-based morphometry

VBM is a fully automated whole-brain technique that involves the voxel-wise comparison of concentrations and volumes of segmented grey and white matter between two groups of subjects. It discounts gross volumetric differences by using spatial normalisation and allows the exploration of local tissue composition without using manually placed regions of interest. Spatial normalisation involves the registration of all images to the same template, and it is followed by smoothing, which minimises the residual differences. After smoothing, each voxel contains the average concentration of grey or white matter, which is called “density” or “concentration” (55,56). In optimised VBM (57), an additional step (modulation) is introduced to convert grey- and white-matter densities into grey- and white-matter volumes. The images are analysed according to the general linear model and the Gaussian Random Field theory by using SPM (statistical parametric mapping).

VBM is currently widely used in psychiatric research. Our group has used it to study schizophrenia (52) and BD (53). A small study on 11 patients with poor-outcome bipolar illness (58) reported reduced grey-matter density mainly in frontolimbic areas, particularly in the cingulate gyrus, and in the right temporal lobe. In our study (53), we found lower white-matter density bilaterally in frontal areas in 39 BD patients compared with healthy controls (Fig. 2).

Figure 2.

Areas of significantly lower white-matter density in bilateral frontal white matter in bipolar patients compared with healthy controls projected on contiguous 2-mm-thick axial T1-weighted averaged images. The numbers at the left bottom corner of each section represent the z MNI coordinate (from ref. 53, with permission.)


The pathophysiogenesis of the brain abnormalities in BD is still unclear, with a number of putative mechanisms. A possibility is glutamatergic neurotoxicity related to hypothalamic–pituitary–adrenal axis dysregulation, which has the hippocampus as a primary target (59). It is plausible that such progressive damage might happen on a substrate of neurodevelopmental, genetically driven, abnormalities. The glial dysfunction could lead to neuronal injury through deficit of neurotrophic factors (60), and synaptic alterations may be relevant for plasticity-related changes (61).

The possible role of medications certainly needs mentioning. Most BD patients are receiving long-term treatment with mood stabilisers, and some of them are likely to be exposed at some point to neuroleptics and antidepressant medications. So far, the available in vitro and in vivo evidence suggests a neuroprotective and neurotrophic role of lithium on the brain. For example, increases in grey-matter volume and in the neuronal marker N-acetylaspartate have been described in BD patients rescanned after 4 weeks of lithium treatment (62,63). Experimental evidence suggests that antidepressants may promote axonal regeneration and neurogenesis and prevent loss of dendritic spines (64). The antimanic and mood-stabilising effect demonstrated in an increasing number of antiepileptic medications [i.e., sodium valproate (VPA), carbamazepine (CBZ), lamotrigine (LTG)] has triggered interesting speculations on possible common pathophysiogenetic mechanisms between BD and epilepsy (65), that warrant further exploration.