MRI takes a new direction

New work has pounced on a quirk of the magnetic properties of nerve fibres to take MRI brain scanning to the next level

Over the last few decades, Magnetic Resonance Imaging (MRI) has transformed the field of neuroscience by allowing detailed non-invasive investigation of the human brain. One of the strengths of MRI is its sensitivity to a wide range of soft tissue properties such as water content, blood flow, iron content, and diffusion. In recent years, research groups from around the world have been puzzled by the relationship between the direction of nerve fibers and variation in image contrast observed using specialised high-resolution MRI techniques. At the University of Nottingham, we have discovered that these effects can be explained by modelling nerves as long thin hollow tubes with special (anisotropic) magnetic properties.

An MRI scanner is basically a giant superconducting solenoid that generates an incredibly strong magnetic field. To put the strength of this field in context, the earth’s magnetic field is around 50 micro Tesla (T), a standard hospital scanner can create a magnetic field of around 1.5 T (30,000 × earth’s field), and the specialised research scanner at the University of Nottingham generates a magnetic field of 7 T (140,000 × earth’s field). When a patient is placed in this strong magnetic field, the nuclei of the hydrogen atoms found in water molecules in the patient’s soft tissues become aligned with the field on average, generating a weak net magnetisation. In MRI, images are formed by resonantly manipulating this nuclear magnetisation through the application of radio frequency (RF) electromagnetic pulses and small magnetic field gradients.

Stronger magnetic effects are produced by the electrons in atoms and molecules in a material, having an amplitude that depends on the material’s molecular structure and composition. These effects are characterised by the magnetic susceptibility, which is the physical quantity that describes how easily the material can be magnetised and in what direction this magnetisation points relative to the applied magnetic field. Using a special combination of RF-pulses and field gradients, it is possible generate MR images of the human brain that are particularly sensitive to differences in susceptibility across tissues and are commonly referred to as ‘susceptibility-weighted’ images. These images are useful for investigating the iron content of blood and tissue and have already been applied in many clinical studies. Recent work on susceptibility-weighted data has shown that, as well as containing useful information on iron content, these images are highly sensitive to the direction of nerve fibers in the brain. An example of a typical susceptibility-weighted brain image from a healthy subject is shown in Figure 1: the yellow arrow highlights a region of the brain containing nerve fibers that are oriented perpendicular to the main field of the scanner. The magnetic properties of the perpendicular nerve fibers have caused this region of the brain to appear much darker than the surrounding tissue. It is this fascinating observation that motivated our work at the University of Nottingham on modelling the effects of nerve fibers in susceptibility-weighted MRI.

Brain tissue is largely made up of grey matter (GM), in which information is processed close to the surface of the brain and white matter (WM), wherein information is transferred from one part of the brain to another. WM is made up of billions of microscopic nerve fibers that transfer information in the form of tiny electrical signals. To increase the speed at which these signals travel, each nerve fibre is encased by a sheath formed from a fatty substance, called myelin. It is the whitish colour of myelin that gives rise to the term ‘white-matter’. Our starting point for modelling nerve fibers in WM was to consider carefully how the geometry and molecular structure of the myelin sheath would affect susceptibility-weighted images of the human brain acquired using MRI. The myelin sheath surrounding each nerve fibre is largely composed of long chain-like lipid molecules. These are packed together tightly so that the lipid chains point out from the centre of the nerve in a radial manner. One of the interesting properties of these lipid molecules is that their magnetic susceptibility depends on the orientation of the chain-axis relative to the direction of the magnetic field of the MRI scanner. This type of magnetic susceptibility is called ‘anisotropic’, and is common in materials that have an ordered molecular structure. Based on these observations, we formed a simple model in which each myelinated nerve fibre was represented as a long, thin hollow cylinder or ‘tube’ composed of material with a radially-anisotropic magnetic susceptibility. Figure 2A shows a schematic of the fiber model where the arrows point in the direction of the magnetic anisotropy. As mentioned above, the signal measured in MRI originates from water molecules. To make the nerve fiber model more realistic, two different types of water were considered: (i) ‘free-water’, which is water that is free to move and diffuse through tissue normally; (ii) ‘myelin-water’, which is water that is trapped in the myelin sheath, and so has restricted movement and diffusion; this leads to a rapid decay of its signal. By combining the anisotropic magnetic properties of the myelin sheath with the rapid decay of the ‘myelin-water’ signal, we were able to form a simple nerve fiber model that could be used to study fiber orientation effects in MRI.

The next stage was to use this simple tube-model to simulate the signals measured in MRI relating to nerve fiber effects and fit this predicted behaviour to the results measured from the human brain in vivo. Simulations were carried out on two different length scales. Firstly, powerful cluster computers at the University of Nottingham were used to carry out millions of simulations of the microscopic magnetic fields produced by slightly different tube-model parameters. The parameters which were varied included the thickness of the tubes and the strength of the anisotropic magnetic susceptibility. Figure 2B shows an example simulation of the magnetic field created by the tube-model. The second set of simulations was focused on predicting the large-scale bulk tissue effects due to the shape of the brain and the distribution of WM fiber bundles on the mm length scale. By combining the results of these microscopic and macroscopic simulations we were able to fine-tune the tube model and successfully explain how the directions of the nerve fibers affected the appearance of white matter in susceptibility weighted images of the human brain. The model potentially also allows information about nerve fibres, such as their size and direction, to be inferred from magnetic resonance images.

The accuracy and intrinsic simplicity of the tube model will give researchers around the world a new and powerful tool for interpreting susceptibility-weighted MRI images. While most MRI-based research focuses on tissue measurements at the millimetre length scale, our experimental scans on healthy human volunteers and modelling of the myelin sheath shows that much more detailed microscopic information relating to the size and direction of nerve fibres can be generated using fairly simple imaging techniques. These results will give clinicians more information for recognising and identifying lesions or abnormalities in the brain. This study will also give scientists a better understanding of the effects of nerve fibres and their orientation in magnetic resonance imaging and has potentially useful applications in the diagnosis and monitoring of brain and nervous system diseases like multiple sclerosis where there are known links to myelin loss.

Dr Samuel Wharton, EPSRC Research Fellow, School of Physics and Astronomy, University of Nottingham