Magnetic Resonance Imaging (MRI)

The idea of magnetic resonance imaging (MRI) was first described in 1938 by Rabi and colleagues. Since then a large amount of work has been invested in improving the image quality and acquisition speed, making it a very common imaging technique within hospitals.

The main benefit of MRI over other techniques, such as X-Ray imaging or computer tomography (CT), lies in the fact that MRI does not use ionising radiation. Ionising radiation can cause an electron to be knocked out of its orbit around a nucleus, which can then inflict damage to the patient's DNA. The human body can repair such damage to a certain extent, however, it is advised not to be exposed too often to this kind of radiation. One should keep in mind, however, that each imaging technique has its benefit. Therefore, when the doctor decides which technique to use, he or she weighs the potential risk against the benefits, making both X-Ray and CT important imaging techniques for patient care.

MRI on the other hand does not utilise ionising radiation and can subsequently be used for relatively safe and repeatable scans, in particular at the early stages of life. In general an MRI machine is just like a sophisticated camera that is somewhat similar to those found on modern mobile phones. Unlike the camera on your mobile phone, however, the MRI machine acquires different information which comes from the inside of your body, rather than from the surface. The secret behind the images acquired lies within the name, magnetic resonance imaging.

MRI vs Camera

An MRI machine is first and foremost a gigantic magnet, which is always turned on, just like the earth's magnetic field. The MRI machine, however, can be about 100'000 times more powerful.


The trick lies in the small molecules within our body, for example hydrogen (water), which are like a compass in the earth's magnetic field. The compass needles (spins) align themselves with respect to the magnetic field inside the MRI, just like the compass needles aligns itself to the magnetic field of the earth.

A second magnetic field then tips the compass needles away from their position, which makes them start to oscillate around their preferred direction (parallel to the MRI's magnetic field). The oscillation is large at the beginning, but with time becomes smaller and smaller, ultimately bringing the compass needles back into their original position. As the oscillation becomes smaller, energy is sent out which is then detected by the MRI and makes up the image.

Taking the image takes longer than with a regular camera (usually around 30 minutes). During that time, it is important to lie as still as possible. Similar to a real camera, once the object in the picture moves, the image becomes blurry.

blurry images

And just like a real camera, the MRI makes sounds while taking the image. Unfortunately it is a lot louder than the normal *click* sound of a camera. Therefore, when going into an MRI scanner, the patient uses ear protection. The sound an MRI scanner makes while taking the picture can be different every time, depending on what part of the body is going to be imaged. This is one example of the noise a patient might hear.

MR Physics

MRI uses the magnetic moment of non-zero spin nuclei, such as hydrogen, to generate a detectable signal. The spin of a nucleus can be seen as a compass needle. Normally the orientation of spins of the nuclei within the human body are not aligned. In order to create an image, MRI uses a constant magnetic field (B0), along which all the spins align.

A second magnetic field, which oscillates at radio frequencies (RF pulse), is then applied to the aligned spins. The RF pulse excites the spins, which leads them to rotate away from their orientation. This will not happen for any frequency, but is directly related to the so-called Larmor frequency (the resonance frequency), which is proportional to the magnetic field strength. Subsequently the spins start precessing about B0. Both rotation and precession start to decay to bring the spins back to their original position, during which they lose energy that can be detected by the MRI scanner. The spins' rotation and precession decays exponentially with tissue specific time constants T1 and T2 respectively. In newborns these time constants are in the order of 2.5s (T1) and 0.2s (T2).

There are two key parameters that can be tuned, in order to highlight different tissues in the image. One is called the repetition time (TR), which is related to time constant T1. It is the time that passes between the application of the first excitation pulse (RF pulse) and the next. The time related to the time constant T2 is called echo time and represents the time between the application of the excitation pulse and the peak of the response signal (detected signal).

So far, however, we have no information as to where the signal that we are recording is exactly coming from in the body. By applying a third magnetic field, the so-called gradient, we can gain spatial information. Gradients are magnetic fields which change in intensity across the MRI scanner. This additional magnetic field strength, in combination with Larmor frequency mentioned above, allows then for tissues at separate positions within the body to experience a different resonance frequency.

Gradients are not only useful to select individual slices (approximately 2-D planes) in the brain to be imaged, but also to investigate different aspects, such as the diffusion of water molecules within the brain (see next section). Commonly, MRI images are stacks of slices rather than the entire brain volume, which necessitates that the patient inside the MRI scanner does not move. Research into motion correction and MRI methods that can image entire volumes at once is therefore of particular importance.

A useful concept when talking about MRI acquisition strategies or sequences is k-space. It is based on the mathematical principle of Fourier transformation of the gradients and the spatial components x and y of the image and can be seen as a temporary image space. In k-space, measurements around the centre (low values of kx and ky) correspond to gradual changes in the intensities, i.e. constant or slowly varying areas of intensity. High values on the other hand corresponds to edges in the image. Imaging sequences are often described as paths (trajectories) in k-space, as it allows a relatively simple representation of the elements used in the sequence.

The time it takes to acquire MR images can be improved by modifying these paths. One example is called echo planar imaging (EPI), which acquires all of k-space (both slow and rapidly varying areas of the pictures) in one sweep, rather than line by line. Another method to speed up the imaging sequence is by either not acquiring all of k-space and estimating the missing parts, or using parallel imaging, which basically uses many cameras at once in a divide and conquer approach to take the image (similar to panorama pictures made from individual images stitched together).

Diffusion MRI

Diffusion MRI uses an MRI sequence that allows the doctor or scientist to follow the movement of water molecules. A simple way to look at it is by imagining that the diffusion sequence 'tags' a given water molecule, then acquires a first image, waits a moment and then acquires a second image. The movement the water molecule underwent in that time frame can then be estimated.

Tagging and motion estimating

There are many aspects of the human body that can be investigated using diffusion MRI. However, one application in particular benefits from the fact that sometimes the molecule's movement is hindered by "walls". One example can be found in the brain, where we can investigate the movement of water within axons (see the background section on The Brain - White matter).

The movement of water can be described in general by Brownian motion, or simply speaking a random walk. This happens due to collisions with other water molecules over time. Once a "wall", such as the axonal membranes, is introduced, the motion cannot take place freely, leading to a preferential diffusion direction.

Unrestricted and restricted Brownian motion

Mathematically, this preferential diffusion direction can be expressed through measurements of anisotropy. The amount of anisotropy (directionality) has been used to study patients and healthy subjects, where low levels of anisotropy can identify white matter damage. This is of particular interest, as it has been shown that the level of anisotropy is related to how well a task is performed.

The principle of diffusion MRI was introduced by Stejskal and Tanner in 1965. After the spins are tipped away from their preferred direction a bipolar gradient is applied. That means that the same gradient with opposite signs is used. In simple terms, this means that something is done (a change in phase) to the spins in a first step and then "undone" in the second. If the water molecule stays on average at the same position, there will be no effect and therefore no detectable signal. This is the case, for example, for unrestricted Brownian motion.

If the water molecule changes its position between the two steps, however, we will be able to see a net effect. This is due to the fact that the Larmor frequency is dependent on the amplitude of the magnetic field, which changes depending on the location due to the gradient. This net effect can then be related to the movement of the water molecules and the structure can be mapped.