Competition Winner Entry: Advances in Ultra-High Field MRI and their Application in Neuroimaging

MCI Neuroscience has been running a competition open to PhDs and Postdocs, encouraging them to write articles for Neurotech today. Winners receive wonderful prizes, as well as a featured spot on our blog. We are always happy to welcome new submissions.

The competition winner is Andrew Crofts, from the University of Leicester. He is a PhD student researching neurovascular coupling in ageing and disease using novel functional MRI methods. This article is a brilliant indication of the bright future Andrew has ahead of him, and we look forward to seeing more of him. 


Development of the first 0.1 Tesla MRI scanners was undoubtedly a revolution in medical science, an imaging technology that used neither ionising radiation nor injectable tracers. Those early scanners were limited by their weak magnetic fields, relatively low tech hardware and small number of sequences, giving low resolution and far fewer applications than scanners today. But even in a 0.1 Tesla scanner, those early pioneers could see the potential. As stronger magnets were used, technology and techniques were developed to exploit several properties of living tissue affected by magnetic fields, from the water diffusion properties of tumours to the resonance frequencies of different metabolites to changes between oxyhaemoglobin and deoxyhaemoglobin. Today, 3T scanners are common in hospitals and researchers use field strengths between 7 and 11T. As the field advances to use scanners of 14 or 20 Tesla, more techniques can be realised and refined, but more problems will have to be overcome.



Ten years ago, the best available resolution of fMRI was 2mm3, and while improvements have been made in that time, submillimetre resolution is not yet common in human fMRI. 1-2mm resolution is of course a good resolution for many applications, otherwise fMRI would not have been nearly as widespread in research and as a diagnostic tool for the past twenty years. However, available resolution limits the ability to localise regions of activity. When applied to brain mapping, either for research or for planning brain surgery, this resolution limits the identification of a functional region. Activity tends to be defined simply by region (e.g. prefrontal cortex) and the complexities of different cortical layers and individual circuits cannot yet be mapped. Improving the resolution to micrometres opens up numerous applications. As individual cortical layers are defined by the predominant neuron type, improving localisation of regions with enhanced or depressed activity can improve drug discovery, and potentially correct drug prescription if high field fMRI becomes commonplace in a clinical setting. As fMRI relies on blood oxygenation changes (the blood oxygen level dependent or BOLD response), increase in resolution to the submillimetre level allows for differences in response in individual capillary beds. In a preclinical setting, the ability to non-invasively image the microvasculature at this level can give new insights into neurovascular coupling, highlighting the capillaries’ responses to pharmacological agents, changes at different stages of hypertension or atherosclerosis or response to amyloid or prions.



Diffusion weighted imaging builds a structural image based on displacement of extracellular and intracellular water. Free water typically diffuses at a rate of 1µm/ms, but this rate is reduced in tissue as solid structures limit the molecules’ movement. Variation in diffusion-driven displacement of water provides contrast, allowing high resolution structural images to be built. Techniques such as tractography rely on this method in order to image white matter tracts, damage to white matter from stroke or injury, and determine structural connectivity, by calculating voxel information on which direction the diffusion is fastest (diffusion tensor imaging). Diffusion-weighted imaging can also be used to observe cytotoxic oedema in stroke lesions, the first application that gave the technique clinical use, and to determine cancer progression. These insights into tissue architecture open up several other angles of study that can be investigated by further improving dMRI techniques and technology. Cell density, membrane permeability, cellular composition, and the recently proposed possibility of structured water bound to membranes, can all affect water diffusability, and so can potentially be imaged through improvements in dMRI. High field strengths can allow for diffusion-weighted imaging at the level of individual neurons, opening up the possibility of localising degeneration to specific synapses or axons, measuring diffusion of intracellular water and its effect on cell structure and function, and modulating the amount of water-membrane interactions affecting the dMRI signal to image specific aspects of tissue structure. Improved field strengths also allow for better SNRs in diffusion-weighted fMRI, a new technique which measures cell swelling during neuronal activation, as water enters the neuron. This technique detects neuronal activity, directly, unlike blood oxygenation level dependent fMRI, without the time lag found in the neurovascular response, allowing neuronal and vascular events to be separated and activity to be better pinpointed.

Improved field strength also improves sensitivity to other molecules, such as n-acetyl aspartate, creatine, lactate and neurotransmitters. The ability to measure an apparent diffusion coefficient (ADC) for neurotransmitters gives obvious benefits in studying many neurological and neurodegenerative conditions where neurotransmitter production or uptake is disrupted, from depression to schizophrenia to Alzheimer’s disease. What remains to be seen is whether higher field strengths could allow ADCs of these molecules to be imaged, or whether ADC in an individual voxel must be measured through diffusion-weighted spectroscopy.



MR spectroscopy, rather than building an image, instead targets a small volume of tissue and detects concentrations of various metabolites, with the main detectable compounds being n-acetyl aspartate, choline, lactate, glutamate/glutamine, creatine, phosphocreatine and glycine. However, compared to water protons, the sensitivity of scanners to these metabolites is very low (hence why MR Spectroscopic Imaging, which does build an image from these signals, is only recently beginning to see use), and current scanners are limited to these few metabolites. Beyond 7T, experiments suggest that the relaxation properties of different metabolites change at different rates, giving a more dispersed spectrum. Combined with the increase in sensitivity from higher field strength, previously undetectable metabolites, including serine, methionine, folate and homocysteine will be detectable.

This could have applications in the study of mental disorders, potentially leading to a method of diagnosis led by detection of chemical changes in the brain rather than psychiatric symptoms. For example, previous studies in schizophrenia have focused on the glutamate/glutamine system, showing disease related changes in glutamate, GABA, glutamine, glutathione, serine and NAAG. GABA currently requires specialised techniques which filter out other metabolites, however this will not be necessary as field strength increases, potentially allowing MRS of the glutamate and GABA system to become a diagnostic tool in patients with psychiatric disorders, with MRS of serine becoming available as a research tool. Ultrahigh fields also open up the possibility of 31P spectroscopy, which has poor SNR on current scanners, allowing resolution of AMP, ADP and ATP, which may allow for direct quantification of energy metabolism.



As field strength increases, nuclei with negligible magnetic effects in current scanners become viable targets for imaging, including Na and Cl, which have SNRs a thousand times lower than that of protons. This opens up several neuronal processes to imaging, the most obvious being EPSPs and IPSPs, but also macromolecule interactions that effect the ions’ electric field. Sodium was the first of these ions to be investigated, the obvious choice as it has the highest concentration and is the main driver of activity, and attempts have been ongoing for 30 years. It wasn’t until 9.4T machines were available that any progress was made, due to the low sensitivity and short relaxation times, allowing quantification of tissue sodium and chloride ions recently. Potassium imaging has been attempted at 7T and 9.4T, however sensitivity was lower than sodium or chloride, and there is doubt that reliable results can even be obtained at 20T.

Imaging of these ions would each have distinct benefits. The recent study quantifying intracellular sodium was used to map cell volume fraction in the brain, supporting research showing that brain volume loss in normal ageing can be attributed to reduction in cell size while cell number and cell density remain constant, but this may vary depending on health, potentially providing another measure of disease progression. Intracellular vs. extracellular chloride was used at 7T to detect glioblastoma multiforme by the increase in Cl- in glioblastoma cells. If this became feasible for medical use it could drastically improve diagnosis, as currently glioblastoma can’t reliably be diagnosed noninvasively. Using longer imaging times, changes in ion homeostasis can be measured to infer average membrane potential in a region, giving indication of excitotoxicity in neurodegenerative disease.



There is an old physics joke, in which a farmer’s chickens won’t lay any eggs (or cows won’t produce milk, depending on who tells it), so he asks a physicist for help. The physicist makes some calculations, and tells the farmer “I have a solution, but it only works for spherical chickens in a vacuum.” The point being, theory assumes perfect conditions never seen in reality. Anyone who has refined a sequence on a phantom to get a high SNR and minimal distortion, yet been unable to get a good image on an animal can agree to that. The major problem with increasing field strengths is that small inhomogeneities in the magnetic field are amplified. Comparing single shot EPI at 3T with an equivalent sequence at 9.4T, for example, shows a much greater level of distortion in the image at the higher strength. This is compensated for by improved shimming hardware to compensate for field variation, and segmented sequences to break down data acquisition. However, whether this is effective at 14T and higher remains to be seen. Recent years have seen large advances in shim coils, including multi-channel shim arrays, in which each coil is driven individually to better target regions of field instability. Further advances are in development, including coils to perform real-time shimming in response to field variations during scanning, and integrated shim and RF receive coils, to save power and improve detection, and with each new refinement in technology, realising the potential of ultra-high field MRI gets closer.



While I have focused here on ultra-high field MRI, it would be remiss of me not to mention advances in the opposite – ultra-low field MRI. It seems like a step backwards on the surface. Talk of a 0.1T MRI scanner calls to mind Sir Peter Mansfield’s work in the 1970s developing the first EPI sequence rather than something that could be implemented in today’s research. However there are several groups developing ultra-low field scanners and methods that will benefit modern medical research. MEG technology, which measures the weak magnetic fields generated by neuronal activity using superconducting quantum interference devices (SQUIDs) and superconducting receivers, can be adapted to MRI systems to give good resolution at ultra-low fields. The low field strength means minimal distortion and noise due to field inhomogeneities, and has the obvious application of making MRI based diagnosis available to patients with metal implants. As the technology is based on systems used for MEG, it also allows for hybrid MRI/MEG systems to be developed, allowing structural imaging and functional MEG recordings to be done in a single session, making it easier to register MEG recordings to a structural image and making acquisition more efficient. The fact that applications are being found for ultra-low field MRI also leads to an important observation about high field MRI – that even if the potential for ultra-high field MRI is realised, lower field strengths are still a useful tool and should not be dismissed entirely in favour of high field strengths.



Duyn, JH (2012) The future of ultra-high field MRI and fMRI for study of the human brain, Neuroimage 62; 1241-1248

Uğurbil, K (2012) The road to functional imaging and ultrahigh fields, Neuroimage 62; 726-735


The image used for this article thumbnail can be found here. The article is also excellent further reading for those interested in learning more about this topic.

Trusted by

Micro Control Instruments Ltd.

Innovation Stables, Unit 12 | Newplace Park | Framfield | East Sussex | TN22 5RH | United Kingdom
Tel: +44 (0) 1825 890 858

ⒸMCI Neuroscience 2018