Medical Imaging Magnetic Particle Imaging Biology Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Medical imaging is the technique and process used to create images of the human body for clinical purposes; seeking to reveal, diagnose or examine disease, or for medical science (including the study of normal anatomy and physiology).

The goal of medical imaging is to provide a picture of the inside of the body in a way which is as non-invasive as possible. As a discipline and in its widest sense, it is part of biological imaging and incorporates radiology, nuclear medicine, investigative radiological sciences, endoscopy, (medical) thermography, medical photography and microscopy. Measurement and recording techniques which are not primarily designed to produce images, such as electroencephalography (EEG), magnetoencephalography (MEG), Electrocardiography (ECG) and others, but which produce data susceptible to be represented as maps can be seen as forms of medical imaging.

The process involved in acquiring images varies, depending on the technology being used and the area of the body which is being imaged. In the clinical context two types of imaging technologies are available; visible light and the more advanced invisible light imaging. Visible light medical imaging technology involves digital video or still pictures that can be seen without special equipment. Dermatology and wound care are two modalities that utilise visible light imagery. Invisible light medical imaging is generally equated to radiology or "clinical imaging". Diagnostic radiography designates the technical aspects of medical imaging and in particular the acquisition of medical images.

Medical imaging is often perceived to designate the set of techniques that non-invasively produce images of the internal aspect of the body. In this restricted sense, medical imaging can be seen as the solution of mathematical inverse problems. This means that cause (the properties of living tissue) is inferred from effect (the observed signal). In the case of ultrasonography the probe consists of ultrasonic pressure waves and echoes inside the tissue show the internal structure. In the case of projection radiography, the probe is X-ray radiation which is absorbed at different rates in different tissue types such as bone, muscle and fat.

The term non-invasive is a term based on the fact that following medical imaging modalities do not penetrate the skin physically. But on the electromagnetic and radiation level, they are quite invasive. From the high energy photons in X-Ray Computed Tomography, to the 2+ Tesla coils of an MRI device, these modalities alter the physical and chemical environment of the body in order to obtain data. New technology for medical imaging is being developed all the time, introducing machines which are less invasive and technology which reduces the need for radioactive materials and other harmful substances in medical imaging.

Some imaging studies simply require a capture of an image, while others involve the introduction of a contrast material to the body. Contrast materials are swallowed or injected, and they are designed to be highly visible in the picture, allowing a doctor to follow their progress through the body. A barium swallow, for example, may be used in an x-ray of the digestive tract to look for ulcers and perforations, while radioactive contrasts may be injected to look for signs of thyroid cancer.

There is a long history of using contrast agents and tracers in medicine dating back as far as 1913. They provide important information for diagnoses and therapy but for some applications, they require a higher resolution than can currently be obtained. One well known imaging method is Magnetic Resonance Imaging (MRI) which makes use of a powerful magnetic field to align the magnetisation of some atoms in the body, and radio frequency fields to systematically alter the alignment of this magnetisation. This causes the nuclei to produce a rotating magnetic field detectable by the scanner and this information is recorded to construct an image of the scanned area of the body. Contrast agents that incorporate magnetic particles are routinely used in clinical MRI examinations. These biocompatible particles are typically based on rare-earth elements or iron oxides and can highlight specific anatomical structures such as blood vessels or tumours. Furthermore they can serve as markers for processes at the molecular level. The spatial distribution of the magnetic contrast agent is recorded indirectly. In these MRI studies, the magnetic particle magnetisation is used to alter the signal of the intrinsic nuclear magnetisation of the body. Detection thresholds are such that due the strength of the contrast agent magnetisation and the orders of magnitude weaker background signal from the host tissue is a critical limiting factor. Looking directly at the contrast could potentially infer a stronger signal or reduce the amount of contrast agent required.

Resonance methods such as those used for MRI studies are often unsuitable for imaging magnetic particles, and 'inversion methods' that detect the magnetic field outside the object do not provide high spatial resolution. This report presents a method for obtaining a high-resolution image of such tracers that takes advantage of the nonlinear magnetisation curve of small magnetic particles known as Magnetic Particle Imaging (MPI).

Gleich and Weizenecker first presented this approach for capturing and localising the signal of magnetic particles inside the body in 2005 (Gleich et al, (2005)).

Magnetic Particle Imaging

Magnetic particle imaging is a tomographic imaging technique that measures the magnetic fields generated by magnetic particles in a tracer. In the simplest terms it is imaging of the magnetisation of magnetic nanoparticles. Ferromagnetic material shows a nonlinear response to a modulated magnetisation field. Particle magnetisation saturates at some magnetic field strength and this fact makes it possible to determine spatial selection.

If an oscillating magnetic field, Hac(t)= H0sin(2Ï€f0t), is applied to ferromagnetic material with frequency f1 and sufficiently high amplitude A, the magnetic material will exhibit a magnetisation M(t), where t is time. A series of harmonic frequencies is contained in M(t) which can be separated from the received signal by using appropriating filtering. The following figure describes the application of a modulation field to a material.

Figure : An oscillating magnetic field (H, modulation field, green curve) is applied to the magnetic material at a single frequency f1. As the magnetization curve (M, black curve) is nonlinear, the resulting time-dependent magnetization (red curve) exhibits higher harmonics, as is shown in the Fourier transformed signal (S, red bars). (Gleich et al (2005)

The saturation of the particles occurs with exposure to a time constant magnetic field with a sufficiently large magnitude. With saturation the generation of harmonics is suppressed as shown in Figure 2. Selective suppression of the harmonics can then be employed for spatial encoding.

Figure : A time-independent field is added to the modulation field. The oscillating field does not significantly change the magnetization of the material, as it is always in saturation. In this state, harmonics of the oscillating field are almost non-existent. The grey box indicates those harmonics used for image formation. The signal at f 1 is not used, as it is small compared to the superimposed induced modulation field signal, and therefore difficult to isolate. (Gleich et al, (2005))

A time-independent field is superimposed that decreases in magnitude moving from the edges and disappears in the centre of the imaging device at a point called the field-free point (FFP). This field is called the selection field. If there is any magnetic material positioned at the FFP a signal will be produced containing higher harmonics. Only the material at the FFP will respond to the modulation field while all other magnetic material will remain in the saturation state. A topographic image can be obtained by moving (spatial encoding) the FFP through the volume of interest and recording the magnitudes of the harmonics. By changing the location of this field-free spot (either mechanically or with auxiliary magnetic fields) the sample can be scanned bit by bit; the image is generated by mapping the magnitude of the harmonics resulting in a map of the spatial distribution of the magnetic particles.

Figure Principle of the imaging technique devised by Gleich and Weizenecker. a, The object to be imaged is immersed in an external field whose strength varies with location. In most regions the magnetization of a magnetic particle sitting inside the object is saturated (dark areas).

b,d, An additional weak radio-frequency field - oscillating between a minimum and a maximum value - cannot change this state.

c, However, in regions where the external field has a value close to zero, the additional field is able to alter the magnetization, which will start to oscillate and therefore induce a signal in a detection circuit. This signal can be unambiguously assigned to the narrow field-free region. By systematically varying the position of the field-free area in the object, a map can be created that gives the spatial distribution of the magnetic particles Trabesinger, A. (2005).

Methods of MPI Image Acquisition

There are two fundamental methods used to spatially encode an MPI image; mechanical movement and field-induced movement. The mechanical movement leads to low scanning speed and a low signal to noise ratio (SNR) due to the weak modulation field. Alternatively additional orthogonal homogenous magnetic fields can be provided (called drive fields). The three components of the selection field can be cancelled by appropriate adjustment of three fields. By driving with a predefined current waveform, the FFP can be moved on a continuous trajectory over the object.

By using drive fields, it is possible to accelerate the movement of the FFP dramatically. For this purpose, a different sinusoidal current with a high frequency is applied. The amplitudes of the currents must be large enough to generate magnetic fields capable of cancelling the selection field at the border of the desired region of interest. The fast FFP movement leads to a rapid local change in magnetisation as soon as the FFP passes a location containing magnetic material. The magnetization change induces a signal in the recording coil that exhibits higher harmonics of the drive field frequencies. This induced signal is sufficient for image reconstruction. The modulation field with low amplitude is now obsolete. Consequently, the introduction of the drive fields overcomes both drawbacks mentioned above, namely the low encoding speed and the low SNR.

The following figure demonstrates the experimental set up Gleich and Weizenecker used in their work.

Figure : The main components of the experiment, and an MPI scanner concept. a, The two large rings generate the selection field. Hence, a d.c. current with opposite direction in the upper and lower coil produces the sketched field (field lines and colour coded field magnitude) with the field free point (FFP) in the centre. The same two rings serve as drive field coils, as an a.c. current is superimposed on the d.c. current. A pair of quadratic recording coils in the centre records the generated a.c. response (harmonics).

b, The field-generating components are sketched schematically for an MPI scanner capable of encoding purely by drive fields. Two field generators produce the selection field. For each direction in space, two opposing drive field coils are used. These coils produce a more or less homogeneous field in the centre of the scanner and can therefore move the FFP. (Gleich et al, (2005))

Proof of Principle

Gleich and Weizenecker successfully obtained images in the initial experiments which have a resolution of well below 1mm. This is remarkable considering that the size of the recording coils (squares with 16-mm sides) and the wavelength of the applied radio frequency field (around 1 km) are both much larger than the size of the resolved features.

Nobel laureate Paul Lauterbur coined the term 'zeugmatography' in his introduction of MRI as a concept for image formation: when two fields are combined, the first one (here, the radio-frequency field) induces an interaction with the body, and the second one (the inhomogeneous magnetic field) restricts this interaction to a limited region. MPI can be seen as a form of zeugmatography. In this way, there is no imposed wavelength limit and MPI can use harmless radio waves that pass through the body without significant attenuation. Furthermore, the detectors can be much larger than the smallest resolved structure, thereby opening the door to depth resolution and, ultimately, three-dimensional imaging.

The following images are those generated by Gleich and Weizenecker in their initial experimentation. Using a drive field leads to a contribution of neighbouring points to the recorded signal at a given position. This means that the simple method of mapping the magnitude of the harmonics is not appropriate for generating an image and a reconstruction is necessary.

Figure Reconstructed images of the object for two different encoding types: The true size of the holes is indicated in the lower right corner of the large images. The drawings on the right side sketch the robot positions used for measurement (bottom) and a true scale image (top). In a, the data at all 52X52 robot positions were used, whereas in b only the data of 3X52 robot positions contribute to the reconstruction. In a, encoding is purely done by robot movement, although the FFP moves a considerable distance in the vertical direction. In b, this movement is exploited and the encoding is achieved partly by the drive field. The total measurement time was about 50 min, including a pure data acquisition time of 18 min for a and 1 min for b. Those spots with low intensity reflect imperfections of the object. (Gleich et al, (2005))

Image Resolution & Reconstruction

There are two sets of data acquired in this method that are used for reconstruction. The nth harmonic Vn(y) of the induced signal is written as;


Where C(x) is the magnetic particle concentration in the object (unknown for the image reconstruction) and Gn(r) denotes the delta response of the system representing the induced signal in the nth harmonic of the set-up if an infinitesimally small object is placed at position r. This function includes all the complex dynamics of the magnetic tracer, as well as the shape of the drive field and the recording coils. A Fourier transform equation is used to deconvolute the delta response from the reference response. The response function obtained becomes;


After division by the known concentration function c(k) of the reference object, a Fourier back-transformation yields Gn(r). For image reconstruction, a direct inversion of the discretised equation is used, as it gives more flexibility with respect to data reduction.

It had been shown so far that high resolved images can be obtained by moving a field-free point over an object using additional homogenous oscillating magnetic fields (drive fields). This provides a proof of principle, whereas predictions of the possible performance have been estimated only coarsely. Gleich, Weizenecker and Borgert carried out further work in order to provide a more detailed theoretical analysis of the properties of MPI by simulating the complete image acquisition process. The physics of the signal chain was modelled and approximated by numerical simulations choosing parameters representing conditions expected in clinical applications and imaging in humans. Two main steps were performed in order to generate images. The first is to simulate the recapture of the data of all the necessary aspects of an actual image. The second is the processing of this data to form a reconstruction. A set of data (called the system function) describing the signal formation and the dynamics of the magnetic material in the scanner set up is provided constituting a delta-like probe. The reconstruction uses both the data from the object and reference simulation.

The following figure demonstrates the method used.

Figure Schematic overview of the simulation and reconstruction process. The simulation algorithm uses the geometrical setup and the object geometries as input. In addition to the phantom (object data) a small spot-like object (reference object) is simulated at various positions in space in order to obtain the system and tracer response (system function). In the reconstruction step, the system function is inverted, allowing for a reconstruction of the image of the phantom. (Gleich et al, (2007))

The simulation study showed that images with a resolution sufficient for many applications can be obtained using MPI. The analysis of the simulation allowed the introduction and verification of different scaling laws, allowing for the prediction of image quality for various imaging conditions. Previously, encoding and acquisition were slow but this simulation showed that methods based on the principle described show a strong increase in acquisition time with dimensionality. More advanced encoding schemes would allow signal detection from larger regions, thus drastically increasing the performance.

The following figure shows images obtained using this method.

Figure : Reconstructed images for various particle sizes and concentrations. The images are sized 20 Ã- 20 mm2 containing 128 Ã- 128 pixel. The acquisition time was assumed to be 40 s. The regularisation parameter was chosen by the best visual impression. It could be observed that the choice of the regularisation parameter did not change with the particle size for a given concentration.

The regularisation parameters λ4, λ6, λ9 and λ10 have been chosen for row one to four. The resolutions in the horizontal and vertical direction are given in brackets. The bottom and right boxes present alternative choices of parameters as an application of scaling laws. For an acquisition time of 4ms instead of 40s, the concentration has to be changed according to the scale given in the right box. From the scaling presented in the bottom, it can be observed that for a fixed particle size of 50 nm the drive and selection field gradient strength can be adjusted to obtain the according image quality for different concentrations. (Gleich et al, (2007))

The significance of this study is that the simulated scanner is large enough to accept human bodies. Together with the choice of field strength and noise the setup is representative for clinical applications. Good resolution, fast image acquisition and high sensitivity are demonstrated for various tracer concentrations, acquisition times, tracer properties and fields of view. Scaling laws for the simple prediction of image quality under the variation of these parameters are derived proving Gleich and Weizenecker earlier work is transferable to a larger scale. Furthermore it has been shown that the quality of the image depends directly on the iron-core diameter of the nanoparticles. The rise in magnetisation curve is increasingly steeper with the increasing size of the diameter. This results in a greater amount of harmonics frequencies that can be read before the signals spectrum drops into the noise level. However complexity of the particles size and shape also increases and as a result the magnetisation becomes anisotropic.

Two important scaling laws were determined by the simulation study; increasing of the particle diameter allows the decrease of the field strength by a power of three leading to the same induced signal in the recording coils, and also the signal to noise ratio (SNR) is proportional to the particle concentration, the square root of the acquisition time, the excitation frequency, the coil sensitivity and inversely proportional to the third power of the gradient strength and the square root of the noise resistance. By varying the gradient strength it is possible to trade-off between the resolution and the detection limit of the particle concentration.

This simulation work was then compared to experimental work carried out again by Gleich and Weizenecker and Borgert who demonstrated experimentally that fast 2D imaging is indeed feasible in magnetic particle imaging using two drive fields and a Lissajous trajectory. It was possible to encode a full image in 4ms and achieve good image quality within 40ms. Videos of moving objects with 25 frames s−1 were presented. They also determined that it was possible to improve image resolution by a factor of 500. This improvement in acquisition time allows for real-time imaging using MPI.

In order to provide mandatory information for the design of a MPI scanner Knopp et al performed a simulation study on different trajectories moving the FFP through the field of view. In the experimental verification of Gleich et al's simulation study a Lissajous trajectory was presented. This trajectory was compared to four other types in the work by Knoop et al. Trajectories are compared with respect to density, speed and image quality when applied in data acquisition. The simulation tested the Lissajous curve, Cartesian, radial and spiral trajectories. Sinusoidal and triangular excitations were used in the simulation, summarised in the following figure.

Figure : FFP-trajectories used in simulations for sinusoidal and triangular excitation, the corresponding currents in the drive field coils, and the trajectory speed in the x- and y-directions encoded as grey values, where dark values denote fast and bright values slow FFP movement.

Overall, the Lissajous trajectory produced admirable image quality even for low densities, and overtook the Cartesian and spiral sampling patterns. However, compared to the Lissajous trajectory the radial sampling showed better results in the interior while performing worse in peripheral regions. Hence, both the Lissajous and the radial trajectory may find a use in MPI. In MPI application, Lissajous and Cartesian can be appreciated with only two dedicated frequencies. This has the advantage that the signals for both drive field coils can be band-pass filtered to compensate for harmonic distortions of available amplifiers. In contrast, radial and spiral trajectories may need an advanced filter stage for this compensation. The Cartesian improved trajectory requires additional logic for the switch infrequencies. For each trajectory, different sampling densities were compared by changing the repetition time while keeping the total acquisition time constant. Increasing density yielded improved image quality as a result. Following this an increase the repetition time to a value that permits further pre-processing is proposed and simultaneously increase the number of repetitions.

The triangular excitation provided better image quality than sinusoidal excitation for all trajectories. This is due to the constant speed of the FFP and the improvements in the uniformity of sampling point density. In practice, an implementation of the analogue signal chain for triangular excitation is a challenging task, since the filter must be able to separate the nanoparticle signal from the drive field signal. In contrast to magnetic resonance imaging, where data are collected in Fourier-space, the trajectory is sampled in image space in MPI. Hence, trajectory design for MPI is fundamentally different from that for magnetic resonance imaging.

As a result of the simulation and experimental work it became clear that particles used in MRI which were previously used in MPI may be unsuitable. To allow for more accurate simulations using MPI, Biederer et al developed a spectrometer capable of measuring the remagnetisation of ferritic nanoparticles. This allows the classification of the suitability of particles for MPI. The approximate particle size distribution can also be obtained to facilitate further simulation accuracy. There are measurement techniques which can be used to characterise nanoparticles but none previously which measured directly the spectral magnetic moment at field strengths and frequencies used in MPI. Biederer et al developed a magnetic particle spectrometer (MPS) which operates at a frequency, f0=25Hz, consistent with that used in MPI. The spectrometer works on the principal that the resolution in an MPI system will be higher if more harmonics can be detected, however field strength is limited sue to the specific absorption rate (SAR) i.e., patient heating. When it comes to the design of a MPI scanner, the chemical synthesis process to design nanoparticles is important and the presented spectrometer is a useful tool to predict the suitability of particles and ensure constant image quality. Based on the measured spectral magnetic moments, particle size distributions were also determined.

Simulations based on particle size distributions show a much higher association with measurements than simulations using monodisperse particles. The determined distribution can be used to improve the estimation of the imaging performance of MPI. However, the Langevin theory of paramagnetism does not take magnetic anisotropy and relaxation into account. As a result, the estimated particle size distribution does not match the geometrical one. Therefore, they should not be called iron-core diameters. Biederer et al suggested a better description could be magnetic effective diameters for MPI.

3 D imaging

Up to this point only 2 dimensional imaging has been presented. Weizenecker along with others demonstrated the first 3D in vivo magnetic particle imaging revealing a beating mouse heart using a clinically approved concentration of a commercially available MRI contrast in 2008. They achieved a resolution of 21.5ms at a 3D field of view of 20.4 X 12 X 16.8mm3. The spatial resolution was sufficient to resolve all heart chambers thus showing that MPI had taken a huge step toward medical application. Theoretically MPI could be used in medical imaging but despite the evidence proving this, agglomeration of the nano-particles due to contact with tissue could not be disqualified. This occurrence would degrade the MPI signal while leaving the MRI performance almost unchanged. To understand the level of speed and sensitivity required for volumetric in vivo imaging, several innovations and improvements had to be introduced into the scanner concept previously used for dynamic 2D imaging.

The following figure represents a schematic of the three dimensional scanner used to image a mouse heart.

Figure : Schematic scanner setup. The mouse was inserted into the x drive/receive coil cylinder using an animal support. The bore diameter is 32 mm. The selection field is generated by both the permanent magnets and the coil pair in the z direction. The drive field coils can move the FFP in all three spatial directions. For signal reception, each spatial component of the magnetisation is detected by a respective receive coil. In the x direction, the drive field coil is also used for signal reception.

The scanner has an effective bore size of 32mm. A pair of permanent magnets and a pair of coils produce the selection field gradient. The permanent magnets contribute 3 Tμ0−1 m−1 and the coils 2.5 Tμ0−1 m−1to the magnetic field gradient, respectively. The scanner uses three sets of drive field coils to enable 3D imaging. The drive field HD with amplitude of 18 mTμ0−1 in the vertical direction is produced by the selection field coils. The drive fields in the two orthogonal directions are produced by dedicated coils which are driven at the same amplitude. Three drive field frequencies are chosen to move the FFP along a 3D Lissajous trajectory. The frequencies for the three directions are 2.5 MHz/ 99 ≈ 25.25 kHz, 2.5 MHz/ 96 ≈ 26.04 kHz and 2.5 MHz/102 ≈ 24.51 kHz, respectively. The Lissajous trajectory has a repetition time of 21.5ms, corresponding to encoding 46.42 volumes per second, and covers a volume of about 20.4 Ã-12 Ã- 16.8mm3. The size of the gaps in the Lissajous pattern was chosen to match the desired resolution on the order of 1mm. Two saddle-type receive coil pairs are aligned approximately perpendicular to the bore. In the axial direction, the solenoid drive field coil is also used for receiving the signal. The series of in vivo experiments comprised scans on 18 mice using different concentrations of the nano-particle Resovist. Ten of the experiments were conducted with dosages low enough for human usage, approximately ranging between the standard dosage of 8 μmol (Fe) kg−1 Resovist used in MRI scans, and a dosage of 45 μmol (Fe) kg−1, which is slightly above the safe dosage of 40 μmol (Fe) kg−1 for human applications. (Weizenecker et al 2008).

Among the new improvements required for this level of imaging were a new receive amplifier concept for the reduction of noise and acquiring a system function for calibration.

The MPI results of the in vivo situation showed no drop in tracer performance. Different structures of the beating mouse heart were successfully identified with the spatial and temporal resolution provided by the data. There were images however that were unresolved fully, for example the left and right pulmonary arteries which have a diameter of approx 500μm, although these structures are considerably smaller than the smallest human coronary arteries normally treated. The results of this study can be used to predict the performance of a full human body MPI scanner. Scaling the system for human applications increases the patient noise contribution, so that SNR estimations have to consider the coil, amplifier and patient noise. The authors of this study estimated that with single-loop receive coils, due to the different scaling of noise contributions with size, amplifier noise would be dominant. With the demonstrated technology, the SNR in the human-size system would be at around 10% of the SNR shown here. However, other amplifier concepts (parametric amplifier, SQUID-based amplifier), cryogenic cooling of silicon J-FET amplifiers or modified tuning can lower the amplifier noise contribution to the level of the patient noise contribution.

The ratio between coil sensitivity and noise voltage at a given bandwidth must be compared in relation to the mouse scanner compared to the human scanner. With the current system, noise voltage as stated above is about 100 pV Hz−1/2 while receive coil sensitivity at the isocenter is about 150 μT A−1. In a patient-noise limited human-size scanner, as described in the previous simulation study (Weizenecker et al 2007), patient noise voltage is 1.8 pV Hz−1/2 (at 1 MHz) and coil sensitivities are 1.4 μT A−1 (single-loop rectangular receive coil (10Ã-10cm) at 10cm depth). If the same particle concentration is imaged at identical resolution using a comparable scanning sequence, the SNR scales proportional to this ratio, i.e., in the human-size system, the expected SNR would be 52% of the SNR found in the present system. Further room for improvement exists in the magnetic particles, encoding sequences and reconstruction algorithms, potentially summing up to a factor of more than 100 (Gleich and Weizenecker 2005; Weizenecker et al 2008). Selection field strength of 5.5 Tμ0−1m−1 can also be achieved over a large FOV; however, without expensive superconductors, only about 3 Tμ0−1m−1 might be feasible.

Resolution with this selection field strength would probably be slightly too low for the direct assessment of the diameters of relevant human coronary arteries. However, using the ability to quantify particle concentration (and therefore indirectly, blood volume) and using the dynamic information, stenosis should be detectable. On the other hand, as described above, technical improvements still offer the potential of substantially higher resolution.

Another conclusion from this study is that improved tracer materials will result in greater resolution and higher quality images. This is where the spectrometer developed by Biederer et al could prove very useful. Overall, the results show that the new imaging modality MPI is capable of in vivo imaging and therefore has the potential to become a clinically adopted imaging modality.


MPI has great potential for medical applications such as vascular or small intestine imaging, where fast dynamic information is required, and the targets are located relatively deep below the skin, the latter because the MPI signal is virtually un-attenuated by intervening tissue. Its sensitivity is improving, with its reported capability of imaging Resovist at concentrations as low as 40μmol(Fe) l−1, and with temporal and spatial resolutions comparable to established modalities: namely 21.5 ms at submillimetre resolution for a 3D field-of-view of 20 Ã- 12 Ã- 17mm3 (Weizenecker et al 2009). The technology, which uses the magnetic properties of iron-oxide nanoparticles injected into the bloodstream, has been used in a pre-clinical study to generate unprecedented real-time images of arterial blood flow and volumetric heart motion. This represents a major step forward in taking Magnetic Particle Imaging from a theoretical concept to an imaging tool to help improve diagnosis and therapy planning for many of the world's major diseases.