Final Report Summary - MEGMRI (Hybrid MEG-MRI imaging system)
In this project, we developed a hybrid imaging scanner that combines magnetoencephalography (MEG) and magnetic resonance imaging (MRI) technology and allows simultaneous structural (MRI) and functional (MEG) imaging of the human brain.
At the beginning of the project, we focused on sensor optimisation. We investigated three different sensor types: low temperature and high temperature superconducting quantum interference devices (SQUIDs) and mixed sensors based on giant magnetoresistance (GMR). Conventional low-temperature SQUIDs are suitable for MEG systems, but they do not tolerate magnetic field pulses generated during MRI. We developed field-tolerant low-temperature SQUIDs that are sufficient for MRI at ultra low fields (ULF-MRI). For high-temperature SQUIDs and mixed sensors, we improved their sensitivity and noise characteristics. All sensor types were pushed beyond the state of the art.
The second step was the design of the test systems. We developed three such systems, each with different sensors, geometry, coil system and electronics. After completing the test hardware for ULF-MRI we started to implement software and sequences for three-dimensional (3D) ULF-MRI. We implemented traditional MRI sequences such as spin-echo and gradient-echo. We have also designed an optimised prepolarised spin-echo sequence which produces the maximum contrast-to-noise ratio between brain white and gray matter. Another novel sequence invented and patented in this project was the polarisation encoding; the prepolarising field is varied in a way to gain additional information about the sample. The hardware, software and sequence development was successful and we demonstrated 3D MRI imaging of biological samples in each test system. These tests verified the hybrid imaging concept and gave us a lot of valuable information for designing the final prototype.
Based on the results with the test systems, we designed a full-scale MEG-MRI prototype. The final system includes a dewar from the commercial Elekta MEG system. The sensor array includes 72 low-temperature SQUIDs: 60 sensors in the occipital areas of the brain and 12 sensors in the temporal regions. The coil system includes the gradient, measurement, excitation and prepolarisation coils. For prepolarisation, a novel self-shielded superconconducting polarising coil mounted inside the dewar was developed. The SQUID and MRI coil control was achieved with the Elekta MEG electronics. Special, low-noise amplifiers were designed for the coils. To be able to perform MRI and MEG acquisitions with the same system, functions necessary for MRI were integrated in to the existing MEG acquisition software. These functions included MRI sequence generation and control for polarisation, gradient and excitation coils as well as synchronised resetting of the SQUID sensors. Virtual channels were added to the data stream to allow parallel and synchronised recording of the control signals driving the aforementioned coils. A separate image reconstruction module was programmed to read the recorded time-domain data files and to write DICOM-formatted image files. These files can readily be imported to the existing MEG-MRI integration software tool that allows visualising the 3D image as well as superimposing MEG source models on to it.
In the final part of the project, we extensively validated the final prototype by obtaining brain images and recording MEG from several subjects. The validations proved that we have accomplished our goal and built a unique hybrid MEG-MRI scanner.
Project context and objectives:
The project concept was to combine MEG and MRI by technology that will allow simultaneous MEG and ultra-low-field MRI (ULF-MRI) measurements. Our objective was to produce and validate a hybrid instrument able to perform simultaneous functional (MEG) and structural (ULF-MRI) imaging for obtaining undistorted structural brain images and time-resolved functional maps.
MEG is a non-invasive imaging technique that measures the magnetic fields that are produced by electrically activated neurons in the brain. MEG has a high temporal resolution. However, the MEG does not provide structural information. MRI is a non-invasive medical imaging technique used to visualise the structure and function of the body. It uses a magnetic field to align the magnetisation of hydrogen atoms in the body. Radio-frequency (RF)-pulses are then used to flip the magnetisation, causing the atoms to produce a rotating magnetic field detectable by the scanner. By manipulating the sample with additional magnetic fields one can gather enough information to construct an image of the body.
ULF-MRI is a new very promising alternative to conventional high-field MRI. When compared with high-field MRI, ULF-MRI will provide enhanced T1 image contrast, improve geometric accuracy, allow safer imaging for patients with pacemakers and other implants, pregnant women and infants, and reduce costs.
The hybrid imaging systems requires field-tolerant femtoTesla-range sensors. Low-temperature SQUIDs used in MEG systems do not tolerate magnetic field pulses generated during MRI. We developed field-tolerant low-temperature SQUIDs that are sufficient for MRI at ultralow fields while compatible with MEG-type sensor array. In parallel, a new type of magnetometer, called mixed sensor, based on the GMR sensor technology, has been developed within our consortium for low-field NMR. In this project, our goal was to push the sensor technology beyond the state-of-the art and fabricate sensors which are sensitive enough for MEG but can also tolerate magnetic fields generated by 3D ULF-MRI.
In addition to sensors, our goal was to develop hardware, software and 3D ULF-MRI sequences for hybrid systems. These require a development beyond the current state of the art as there was no hybrid device with a large sensor array and the whole head coverage. The developed hybrid imaging technology enables us to perform a precise alignment of MEG and ULF-MRI data. In the conventional applications of MEG the functional (MEG) and structural (MRI) imaging are performed separately. This will limit the localisation accuracy of MEG, as there is co-registration error between measurements done with two different scanners. Our hybrid system will eliminate this uncertainty related to MEG studies and will therefore contribute to improving techniques for the understanding of human behaviour and for developing new diagnostic methods and therapies. The simultaneous measurement will also simplify work flow as no high-field MRI will be necessary. In addition, the use of ULF-MRI instead of high-field MRI provides several benefits related to patient safety, enhanced contrast and reduced geometric distortions.
Project results:
Technology, concept and prototype for hybrid MEG-MRI scanner
MRI is a versatile non-invasive medical imaging technique widely used in research and clinical practice. MRI is based on the phenomenon of magnetic resonance - the property of certain materials to interact with RF electromagnetic waves when placed in a static magnetic field (usually denoted as B0 field). In a classical MRI experiment, the signal-to-noise ratio (SNR) increases with B0 strength. That's why ever since its emergence as a medical imaging method, MRI has been moving in the direction of increasing field strength. Current commercial MRI scanners use fields of several Teslas (for comparison, the magnetic field of Earth is of the order of 0.00005 Tesla).
An alternative approach known as ultra-low-field MRI (ULF-MRI) replaces a single B0 field with two separate fields, a (relatively) strong polarising field Bp and a much weaker imaging field Bi applied to the sample in succession. First suggested by Albert Macovsky and Steven Conolly in their seminal paper in 1993, ULF-MRI offers a number of advantages over the classical approach. It is more tolerant to the presence of metallic objects in the imaging volume, can better discriminate between certain types of tissues and allows open, bore-free geometry and silent operation of the imaging device.
One particularly important benefit of using a separate weak field for imaging is that it enables utilising extremely sensitive SQUID-based magnetometers for MR signal readout. SQUID sensor arrays are routinely used in MEG - recording of tiny magnetic fields generated by neuronal currents inside the brain. This makes MEG SQUID arrays a natural choice for ULF-MRI of the head. An additional benefit of this choice is that the same device can be used for both ULF-MRI and MEG, which opens possibilities for considerable cost savings and novel clinical applications.
In the course of the MEGMRI project, we have designed, constructed and tested a prototype device for combined MEG and ULF-MRI recordings. While similar hybrid MEG-MRI devices were constructed by other research groups (most notably by the Los Alamos group) before, they used only a small (less than 10) number of SQUID sensors, limiting their performance. Our device takes the concept of the hybrid MEG-MRI device one step further by employing an array of 72 sensors. Our prototype also features a number of other unique engineering solutions, such as a self-shielding superconducting polarising coil.
The measured magnetic signals are so weak that interference from ambient sources needs to be suppressed by placing the MEG-MRI device within a magnetically shielded room (MSR). Hence, the sensors can measure the subtle MEG signals. Many challenges have been overcome when designing the ULF-MRI functionality. The coil magnetising the imaged object produces a field of the same order as an everyday bar magnet. When the magnetisation has been achieved, another coil is responsible for the field in which the MR image is generated. This field is of the order of the Earth's magnetic field. The usage and interplay of these fields within the MSR has required careful considerations on sensor design, coil drive and structure, and image reconstruction.
Conventional MRI technology would not have been feasible in the ULF-MRI implementation. One key issue is noise in the images. Scaling down the fields in a conventional MRI scanner would have produced random noise images. On the other hand, with the new implementation, ULF-MRI images of the brain resolving structures down to a few millimetres. The MEG-MRI prototype has given valuable information on how to reduce the noise level. In addition, the MEG measurements with the prototype have shown results in brain activity measurements in accordance with current state-of-the-art MEG. One of the greatest advances achieved by the prototype in the conduct of MEG and MRI experiments is that both can be done in one session without moving the patient.
The MEG-MRI prototype is also a powerful tool for exploring new opportunities in ULF-MRI. While MEG research was started in the late 1960s, ULF-MRI has been explored only since the 1990s. Many things can be predicted in theory but the MEG-MRI prototype offers an experimental platform for verifying the predictions. For example, the magnetic fields generated by the coils need to be precisely tuned in conventional MRI, whereas in ULF-MRI this is not necessary. Furthermore, for technical reasons, the main coil current in conventional MRI has to be kept on constantly. The orders of magnitude weaker fields in ULF-MRI make new imaging schemes possible, such as how the MR images are encoded and later reconstructed in a computer. The MEG-MRI device has open structure when compared with a conventional MRI device, it weighs less and the ULF-MRI coils are much cheaper to construct. Great potential lies also in the diagnostic performance of the MEG-MRI prototype. It may be possible to resolve such tissue types from each other that do not generate contrast in conventional MRI. ULF-MRI opens the possibility to quantitative measurements without reference scans or targets. These opportunities may lead to unexpected but very useful application of the MEG-MRI prototype. Eventually, the MEG-MRI prototype can be considered safer than conventional MRI devices because the fields used are orders of magnitude lower and heat-inducing radiofrequency fields are not needed. The projectile danger is nearly negligible and interference with medical implants is less of a problem due to the lack of strong physical coupling mechanisms to the used magnetic fields. The imaging procedure is comfortable for the patient because no hearing protection is needed due to the silent operation of the MEG-MRI device.
We have validated the performance of our prototype MEG-MRI device in both MEG and ULF-MRI modalities. In a simple visual stimulation experiment, the prototype's MEG performance was very similar to that of the commercial MEG scanner. The MRI performance of the prototype, while being inferior to that of commercial high-field MRI scanners, clearly demonstrates the feasibility of in-vivo ULF-MRI imaging of the human brain. We were able to record the anatomical MR images of several healthy subjects and one stroke patient that clearly resolve the basic anatomical structures of the head.
Software and sequences
ULF-MRI software was developed as a Matlab toolbox and includes codes to generate images from MEGMRI devices, to check the quality of these images, and to compare them with standard high-field MRI images. Within this software, it is possible to co-register ULF-MR images with high field images to provide MEG source reconstruction a suitable volume conductor model even if low field images do not enclose the whole head.
In Aalto University, we designed and implemented new MRI sequences for ULF-MRI. We designed an optimised prepolarised spin-echo sequence which produces the maximum contrast-to-noise ratio between brain white and gray matter. Another novel sequence invented and patented in this project was the polarisation encoding; the prepolarising field is varied in a way to gain additional information about the sample.
The software integration for the final prototype was done at Aalto University. The sequence generation part was developed to be compatible with the Elekta MEG software used to control the final prototype: A Matlab interface is used to specify the desired sequence. The sequence file is designed to be compatible with the Elekta MEG system. The instructions from the file are read by the Elekta MEG system, which controls the MRI coils and SQUID sensors. The software allows nearly arbitrary MRI sequences. MRI signals from the sensors are acquired with the Elekta MEG software. The acquired data is processed with Matlab to produce images.
To be able to perform MRI and MEG acquisitions with the same system, functions necessary for MRI were integrated in to the existing MEG acquisition software. These included MRI sequence generation and control for polarisation, gradient and excitation coils as well as synchronised resetting of the SQUID sensors. Virtual channels were added to the data stream to allow parallel and synchronised recording of the control signals driving the aforementioned coils. A separate image reconstruction module was programmed to read the recorded time-domain data files and to write DICOM-formatted files. These files can readily be imported to the existing MEG-MRI integration software tool that allows visualising the 3D image as well as superimposing MEG source models on to it.
Low filed MRI system based on mixed sensors
Commissariat à l'energie atomique (CEA) has developed a 3D MRI test system working at 77 K for mixed sensors. The system is placed outside of a shielded room. The coil system is made of four main coils with three static gradients. The homogeneity achieved is 10 ppm and the main field is variable up to 10 mT. An additional set of three coil pairs create the dynamic gradients for the imaging control. A spectrometer has been developed. This spectrometer is ready and can control the frequencies, phase and amplitude of the RF frequency and the gradients. Echo planar sequences have been developed and 3D images can now be acquired. This test system has allowed demonstrating the possibility of performing 1 mm resolution images in a volume of 6 x 6 x 6 cubic centimetres. A full MRI setup for brain imaging working at 10 mT with a homogeneity volume of 20 x 20 x 20 cubic centimetres was built during the last part of the MEGMRI project.
Technology for field-tolerant low-critical temperature SQUIDS
Low critical temperature superconducting quantum interference devices (low-Tc SQUIDs) are ultrasensitive sensors of magnetic field capable of measuring magnetic fields down to the range of 1 - 10 fT, i.e. field levels of about one billionth of the Earth's magnetic field. Due to this, they have an established position in MEG systems as typical signals from neuronal activity are in the range of 1 pT or less. As MEG measurements are performed in magnetically shielded environments, the SQUIDs are not exposed to large external magnetic fields. However, in MRI the measurement sequences involve large magnetic fields. Even in ultra-low-field MRI the measurement field is typically in the order of tens of microtesla and the prepolarisation field in the order of tens of mT. In practice, the SQUID sensor needs to be able to measure small field variations down to its detection limit while the measurement field is on. Also the SQUID needs to recover from the impact caused by the prepolarisation field within a reasonable time (milliseconds in practice). This, in particular, is to be taken into account in sensor design as the prepolarisation field is typically a factor of 1000 billion times larger than the detection limit of the SQUID.
Practical SQUID sensors are based on superconducting thin-film technology. As itemised below, most challenges related to using the sensors with magnetic field sequences are due to certain undesired properties of superconducting thin films in under large magnetic fields. The most conventional solution to avoid thin-film exposure to magnetic fields in ULF-MRI is to place the SQUID sensor in a tight magnetic shield while the superconducting wire-wound pickup antenna is left outside of the shield. Furthermore, the pickup antenna is typically wound in axial geometry in such a way that it only detects the second-order gradient of the field thus enabling to large extent the separation of fields to be measured from other field components. In a typical MEG system, on the other hand, the sensor, including the SQUID and the pickup, are planar first-order gradiometers and magnetometers. The whole sensor can also be fabricated with thin-film technology. Such sensor is also convenient in MEG application consisting of hundreds of SQUIDs in a helmet configuration. It has certain benefits in industrial manufacturability.
In this project, our approach was to study the possibility of extending the applicability of MEG-type sensor to ULF-MRI while maintaining the properties important in MEG. In the course of doing so, we identified the main issues to be solved in sensor design. Also, a fabrication process enabling sub-micrometer features in SQUID designs was developed as narrow linewidths are expected to improve the field tolerance.
The most obvious problem is that superconducting thin films tend to trap flux. This means generation of vortices, the quanta of magnetic flux, within the films during the exposure to a large magnetic field. The vortices may remain trapped even when the external field is removed. The related magnetic flux may couple to the active elements of the SQUID circuit, the Josephson junctions, killing the magnetic field sensitivity of the device. In the project, we designed and fabricated a number of SQUID models and tested their tolerance against prepolarisation field pulsing. Although with some SQUID models the threshold field of about 10 mT was achieved, it was concluded that for ULF-MRI application some improvement is still preferable. Therefore, we developed partially shielded SQUID package. With this arrangement the field at the location of the SQUID was significantly attenuated but while the total flux penetrating the pickup loop was reduced only slightly (about 10 %) thus not compromising the sensitivity. Using the shielded module, sensors recover from field pulsing at least 50 mT.
Another problem we discovered was noise transients of SQUIDs that emerged after the prepolarisation pulse. The underlying reason was deduced to be random vortex motion in the structures of the pickup circuit after the magnetic field was removed. It was found that wide-line pickup circuits typically used in MEG were particularly bad in this respect generating magnetic field noise spectral densities up to about 1 pT/sqrt(Hz) in the time window needed for MRI measurement. As this is completely unacceptable for imaging purposes, we fabricated and tested pickup circuits with varying techniques and design parameters. It was found that narrow-line thin-film pickups and wire-wound pickups showed little or no excess noise in the frequency range of interest for ULF-MRI.
Further issues that needed to be taken into account were the magnetisation of the superconducting structures in the sensor potentially spoiling the homogeneity of the measurement field.
As a final step a sufficient number of sensor modules were fabricated and tested for the final prototype. The noise level corresponding to a single sensor in the final prototype was measured to be 4-5 fT/sqrt(Hz) in the frequency range used in imaging. Within imaging sequences, the sensors recovered from magnetic field transients under prepolarisation of 22 mT used in imaging in such a way that the noise was not elevated during MRI acquisition. The noise contribution of the SQUID sensor itself was estimated to be 2-4 fT/sqrt(Hz) depending on the channel as estimated by cross-correlation measurements between adjacent channels.
Technology for field-tolerant high-critical temperature SQUIDS
The Chalmers Tekniska Hoegskola developed high-critical temperature (high-Tc) dc SQUID sensors for integration into a hybrid test MEGMRI system. Through the course of this project, these sensors were pushed beyond the state-of-the-art in terms of combining ultra-high sensitivity (< 20 fT) and tolerance to strong pulsed magnetic fields (> 10 mT). The high-Tc SQUIDs were successfully implemented in small-scale MEG and ULF-MRI test systems built at Chalmers during the course of the project.
High-Tc SQUIDs can be operated at liquid nitrogen temperature (77 K) that significantly simplifies cooling requirements when compared with the liquid helium that is used for low-critical temperature SQUIDs. This may be advantageous, especially for MEG where the separation between the SQUID and scalp should be minimised. However, the magnetic field sensitivity of single-layer high-Tc SQUID magnetometers is typically a factor of 10 lower than equivalent low-Tc SQUIDs. A significant limitation of the single layer magnetometer design is the very large inductance mismatch between the pick-up loop and the SQUID loop that reduces the effective area of the sensor (that equivalently increases the magnetic field-to-flux transformation coefficient, and thus the sensitivity of the sensor). To improve the transformation coefficient and thus the magnetic field sensitivity of the sensors, flux transformers with a multi-turn input coil should be used. We developed both types of high-Tc SQUID sensors: directly coupled (dc) single layer SQUID-magnetometers and flip-chip SQUID magnetometers with multilayer superconducting flux transformers.
The dc-coupled SQUIDs were fabricated in a single layer of superconducting YBCO film on a bicrystal substrate. The dc-coupled SQUID magnetometers demonstrated a magnetic field sensitivity of about 35 fT/sqrt(Hz) with excellent low-frequency noise characteristics below 10 Hz, which is important for, e.g. MEG. In addition, these SQUIDs had flux dams incorporated in the pickup loop that suppressed the supercurrent induced by the large pre-polarisation pulses with magnitudes up to 1 mT.
The main challenge for the fabrication of multilayer structures in high-Tc superconducting materials is to obtain c-oriented film on the entire length of the top electrode. In order to provide proper conditions for the growth of c-oriented high-Tc film, very shallow edge slopes of the bottom electrode need to be produced. Another important issue is the smoothness of the bottom electrode because the presence of droplets and precipitates on the surface of the bottom electrode can lead to the formation of unwanted short-circuits between layers. In order to solve these problems, we developed a polishing process for fabrication of multilayer structures in high-Tc films. The polishing method proved to be extremely effective in reducing the step angle of the bottom layer, thereby drastically increasing the Tc and the Jc of the top electrode without degrading the properties of the bottom layer nor introducing unwanted shorts through the insulating layer. The polishing technology has been successfully applied for the fabrication of multilayer flux transformers on 10x10 mm SrTiO3 substrates. Measurements of the flip-chip magnetometers were performed using various types of SQUIDs fabricated on bicrystal substrates. The best gain was achieved for a single washer SQUID. The flip-chip magnetometers showed an increased level of low-frequency noise compared to a single SQUID. Further optimisation of fabrication technology and design will be needed in order to reduce the low-frequency noise.
We have successfully implemented the developed high-Tc SQUIDs in both MEG and ULF-MRI test systems, developed at Chalmers during the project. In these applications, we have used dc- coupled SQUIDs. Despite the fact that flip-chip SQUID magnetometers have demonstrated better sensitivity above 1 kHz, their low-frequency noise was still higher than in dc-coupled SQUIDs.
As a proof-of-principle, the recordings of spontaneous alpha- and mu-rhythms from a human brain were done using the dc-coupled single layer SQUID magnetometers. By measuring brain signals in two different states (eyes open / shut for alpha and hands relaxed / flexed for mu), one can distinguish the activated alpha- and mu-waves. The performed MEG recordings of spontaneous brain activity using our high-Tc SQUID magnetometers showed that despite higher noise-levels compared with their low-Tc counterparts, high-Tc SQUIDs can be used to detect and record physiologically relevant brain rhythms with comparable signal-to-noise ratios.
Mixed sensore for the detection of ultra-low fields
CEA has developed superconducting / magnetoresistive mixed sensors. These sensors are based on a bridge of micro sise giant magnetoresistive elements which sense the current circulating in a one-square-centimeter superconducting loop. The giant magnetoresistive elements based on spin electronics are presently the most sensitive micro size magnetic sensors and are implemented on all hard disks for data storage. The superconducting loops act as a flux-field concentrator with a gain of more than 1000. These sensors are competitive in terms of sensitivity to low Tc SQUIDs at frequencies above 10 kHz but present a 1/f noise which limits the sensitivity at 1 Hz to several hundred fT. During the MEGMRI project, several key improvements have been performed: the optimisation of flux transformers to couple mixed sensors to the external world, the development of mixed sensors on large wafers and the reduction of 1/f low frequency noise by using a modulation of the supercurrents circulating in the superconducting loop. These sensors are working at 77 K and may be used for ultra-sensitive magnetometers and RF detectors for low field MRI or antennas. They are more sensitive than tuned coils below 500 kHz. Their advantage compared with SQUIDs is the strong immunity against RF perturbations and strong RF pulsed allowing a working outside magnetic shielded rooms for low field MRI.
Potential impact:
Impact on clinical applications
The developed hybrid imaging technology has many potential clinical applications. In this project, we identified two major applications.
In patients with pharmacoresistant epilepsy, surgical resection of the epileptogenic cortex is useful. The help of 3D functional imaging of the brain by magnetoencephalography with a reference structural mapping is valuable in locating the epileptogenic area and minimising damage to irretrievable cortical areas during operation. With simultaneous recording of MEG and MRI, the reliability of the signal superimposition on the anatomy is increased; this may reduce the need for intraoperative recordings in the future. The intraoperative functional recordings prolong operations and increase the risk of complications.
The enhanced T1 image contrast of ULF-MRI may provide new applications in oncology diagnostic imaging. The associated MEG would provide simultaneous presurgical functional mapping of brain areas.
Impact on pediatric studies
The developed hybrid imaging system is completely non-invasive and it has an open design which is much more convenient for small children than closed MRI systems. In addition, the ULF-MRI part in the hybrid system has no strong magnetic field. The gradient coils do not create the disturbing noise associated with high-field MRI. The RF coils will deposit less power to the tissue than those used for high field MRI and at lower frequencies. Consequently, ULF-MRI and hybrid MEG-MRI systems are particularly suitable for pediatric recordings. This makes possible to image pediatric brain diseases which were not possible or practical with the present instruments.
Impacts on neuroscience
The combination of functional imaging by MEG and structural imaging by MRI solves one of the most difficult barriers to exact matching of MEG and MRI coordinate systems. The precise alignment of MEG and MRI data eliminates an uncertainty in MEG studies and contributes to improving techniques for the understanding of human behaviour. The MEG-MRI hybrid system has a direct impact on the understanding of the link between neuronal activity and behavioural performance because functional imaging with high dynamics and high resolution is the best tool for monitoring precisely the neuronal activity and correlates it with stimuli and / or brain operations.
Exploitation of hybrid MEG-MRI technology
Inside the consortium, the commercial exploitation for MEG-MRI hybrid systems is in the interest of Elekta which is the world leader of MEG systems. We have estimated that the cost for hybrid MEG-MRI scanner would be only 30 % higher than conventional MEG unit. Assuming that we can clinically demonstrate hybrid scanners superb imaging power over other technologies in some specific clinical application (e.g. imaging of brain tumour or epilepsy patients), then new markets will appear. In this application scenario, we have estimated that the size of the market would be 60 - 100 MEUR annually.
Before commercial exploitation, further improvement of the prototype is needed. This is in the interest of Aalto, which will continue research on hybrid MEG-MRI technology. If this is successful, then Elekta can make the final decision about the commercial exploitation.
Impact and exploitation on high-Tc SQUID technology
Since their discovery in 1986, much effort has been made to use high-Tc superconducting materials in place of their low-Tc counterparts in many applications, mainly due to the possibility of using cheaper and more flexible cooling technology. The complexity of high-Tc superconducting materials has been the main obstacle on the way to their successful implementation.
We have demonstrated that the simple and high-yield bicrystal grain boundary-based YBCO SQUIDs were sufficiently sensitive to detect spontaneous signals generated by the brain at frequencies below 20 Hz, yielding a SNR of approximately 10. The SNR is similar to that which is obtained with low-Tc SQUIDs, despite the fact that the noise level of high-Tc SQUIDs is roughly an order of magnitude higher. These results suggest that high-Tc technology may supplement or replace its conventional low-Tc counterpart in future MEG systems.
Further improvement of the sensitivity of high-Tc SQUID magnetometers requires reliable and high-quality multilayer technology. A new polishing method developed for fabrication of high-Tc multilayer structures improves surface-smoothness and step edges, thereby increasing yield. Combined with bicrystal high-Tc SQUID sensors, these transformers enabled at least a four-fold gain in sensitivity when compared to standard single-layer high-TC SQUID sensors of the same size. Such technology can be scaled to large wafers and is likely to increase magnetic field sensitivity of high-Tc SQUID such that they can truly begin to compete with their low-Tc counterparts.
In future applications, micro-cryocooling technology can replace the liquid nitrogen cryostats used to cool the sensors, thereby reducing the size of high-Tc SQUID systems and eliminating cryogen consumption. Furthermore, it may be possible to develop flexible MEG helmets using the micro-cooling technology. An array of high-Tc SQUID magnetometers in a flexible helmet will increase spatial resolution of MEG recordings and enable the study of new sources of brain activity, including higher-order and higher frequency sources. Initial work has been performed in collaboration with Kryoz Technologies BV (Enschede, the Netherlands) and preliminary results showed very high potential of the microcooling technology.
Future work will also focus on continuing development of the high-Tc SQUID MEG and ULF-MRI systems and their assessment for clinical applications. Specifically, the low- and high-frequency MEG signals are worthy of further investigation, both from a technical as well as a neurophysiological points of view. This work is performed in collaboration with the Sahlgrenska University Hospital and the University of Gothenburg.
Impact and exploitation of field-tolerant low-temperature SQUIDS
Robustness against external magnetic fields in SQUID technology is a benefit in most applications including but not limited to NMR and MRI based applications. For example, within the project we demonstrated SQUIDs capable of operating in Earth's magnetic field with no external shields. We also developed sub- µm SQUID fabrication technology with increased production capacity. The process has recently been adopted to commercial SQUID fabrication and it is and will be exploited in other ongoing and future development projects including e.g. development of SQUID-based readout for submillimetre wave and X-ray imaging systems for astronomy. A sub-mm imager can further be used e.g. as a sensitive security camera. Potential applications include also geophysics. As an example, hundreds of superconducting rock magnetometers have been installed worldwide to measure remnant magnetisation of geological samples. The number of annual installations is about ten systems consisting of typically three SQUID channels each. Recently, SQUIDs have been used in geophysical explorations to find more and other deposits of natural resources deep underground.
The field-tolerant SQUIDs developed in this project have applications as indicated above. Inside the consortium, the commercial exploitation of developed sensor technology is in the interest of the Aivon which is already producing sensors for conventional Elekta-based MEG systems. In addition VTT is continuing the research related to field-tolerant SQUIDs. Aivon and VTT have technology to commercialise the developed sensors as soon as there is demand for them. Currently, the commercial exploitation depends on the exploitation of developed imaging technologies. If these new technologies such as hybrid MEG-MRI or ULF-MRI are commercialised, then there will be big markets for sensors.
Impact and exploitation of spin electronics based magnetic sensors
Spin electronics based sensors are constantly improved and the MEGMRI project has given the opportunity of developing a new generation of very sensitive spin electronics based magnetic sensors. These new sensors present at room temperature the sensitivity of a few pT/sqrt(Hz). The applications of these sensors are magnetometers for earth and planetary field magnetic imaging for cultural heritage, space magnetometry and petrol applications, buried object detection for environment and security applications and medical applications, mainly magnetic detection based biochips. Mixed sensors technology at 77 K has applications for ultra-sensitive magnetometry mainly for research purpose and for medical applications and in particular ultra-low field MRI. CEA has a number of patents on spin electronics based very sensitive magnetic sensors. Mixed sensors production at the wafer scale by a European company is already in place. CEA is presently investigating a route for very high sensitive room temperature sensor volume production.
Impact and exploitation of ULF-MRI based on mixed sensors
ULF-MRI systems are a low cost and easily implementable scanners with acceptable MRI resolution for a number of clinical applications: strokes, cancer detection. Enhanced contrasts in tumours recently demonstrated in mT fields opens new possibilities for ULF-MRI. The impact of this technology, in particular based of mixed sensors may be huge. The threshold for the development of that technique at the large scale will be the demonstration of a full head image with a resolution better than 2 mm in a time less than 10 min. Interest of ULF-MRI is also for patients where high field MRI is not possible like patients with metallic implants or premature babies. A full head system has been built and a protocol for clinical tests has been requested. As soon as a clear demonstration of the capability of this system for brain imaging is demonstrated, CEA valorisation will put in place a route for exploiting this new approach of MRI.
Dissemination - Arranged events
The project consortium arranged several events related to the MEGMRI project. All of these events were open to external audience. The list of organised symposiums, workshops, seminars and training events are given below:
- The Biomag 2010, Satellite Symposium 'Ultra-low-field MRI and its combination with MEG' was organised on 28 March 2010, Dubrovnik, Croatia.
- Brainstorming Workshop - MEGMRI 'The dream you can realise through the hybrid MEGMRI system: identifying the "killer application" of the new-generation scanner', was organised on 16 November in Rome.
- 'Devices for brain research and diagnostics' (open seminar) was organised on 3 June 3 2009 in Gothenburg.
- A training course on 'Flux3D modelling' was arranged by Cedrat in November 2009.
- A training course on '3D-MRI techniques' was arranged by CEA in Helsinki, on 2 June 2010.
- A MEG course was arranged by UdA and Elekta in Chieti, Italy, 31 May 2011.
Dissemination - Publications and thesis
During the project, we have published our main scientific results in top journals and conferences. In total, we have published 24 peer-reviewed articles and several other publications. In addition, two PhD thesis and four MSc thesis were done in this project.
Website: http://www.megmri.net/
Contact:
Prof. Risto Ilmoniemi
Aalto University
Department of Biomedical Engineering and Computational Science (BECS)
P.O. Box 12200
FI-00076 Aalto
Tel.: +35-850-5562964
Email: risto.ilmoniemi@aalto.fi
Participants:
Aalto University - Finland coordinator
Prof. Risto Ilmoniemi (risto.ilmoniemi@aalto.fi via e-mail)
Aivon Oy- Finland
Dr Jari Penttilä (jari@aivon.fi via e-mail)
Cedrat Technologies SA- France
Mr Patrick Meneroud (patrick.meneroud@cedrat.com via e-mail)
Chalmers University of Technology - Sweden
Prof. Dag Winkler (dag.winkler@chalmers.se via e-mail)
Università degli Studi Gabriele d'Annunzio Chieti-Pescara - Italy
Prof. Gian Luca Romani (glromani@itab.unich.it via e-mail)
Commissariat à l'energie atomique - France
Prof. Claude Fermon (claude.fermon@cea.fr via e-mail)
Elekta AB - Sweden
Dr Antti Ahonen (antti.ahonen@elekta.com via e-mail)
Associazone Fatebenefratelli per la Ricerca - Italy
Prof. Paolo Maria Rossini (paolomaria.rossini@afar.it via e-mail)
Hospital District of Helsinki and Uusimaa (HUS) - Finland
Dr Jyrki Mäkelä (jyrki.makela@hus.fi via e-mail)
Imaging Technology Abruzzo s.r.l - Italy
Prof. Antonello Sotgiu (sotgiu@univaq.it via e-mail)
Physikalisch-Technische Bundesanstalt - Germany
Prof. Lutz Trahms (Lutz.trahms@ptb.de via e-mail)
Università degli Studi di Parma - Italy
Prof. Giacomo Rizzolatti: giacomo.rizzolatti@unipr.it
VTT Technical Research Centre of Finland - Finland
Dr Juha Hassel (juha.hassel@vtt.fi via e-mail)
At the beginning of the project, we focused on sensor optimisation. We investigated three different sensor types: low temperature and high temperature superconducting quantum interference devices (SQUIDs) and mixed sensors based on giant magnetoresistance (GMR). Conventional low-temperature SQUIDs are suitable for MEG systems, but they do not tolerate magnetic field pulses generated during MRI. We developed field-tolerant low-temperature SQUIDs that are sufficient for MRI at ultra low fields (ULF-MRI). For high-temperature SQUIDs and mixed sensors, we improved their sensitivity and noise characteristics. All sensor types were pushed beyond the state of the art.
The second step was the design of the test systems. We developed three such systems, each with different sensors, geometry, coil system and electronics. After completing the test hardware for ULF-MRI we started to implement software and sequences for three-dimensional (3D) ULF-MRI. We implemented traditional MRI sequences such as spin-echo and gradient-echo. We have also designed an optimised prepolarised spin-echo sequence which produces the maximum contrast-to-noise ratio between brain white and gray matter. Another novel sequence invented and patented in this project was the polarisation encoding; the prepolarising field is varied in a way to gain additional information about the sample. The hardware, software and sequence development was successful and we demonstrated 3D MRI imaging of biological samples in each test system. These tests verified the hybrid imaging concept and gave us a lot of valuable information for designing the final prototype.
Based on the results with the test systems, we designed a full-scale MEG-MRI prototype. The final system includes a dewar from the commercial Elekta MEG system. The sensor array includes 72 low-temperature SQUIDs: 60 sensors in the occipital areas of the brain and 12 sensors in the temporal regions. The coil system includes the gradient, measurement, excitation and prepolarisation coils. For prepolarisation, a novel self-shielded superconconducting polarising coil mounted inside the dewar was developed. The SQUID and MRI coil control was achieved with the Elekta MEG electronics. Special, low-noise amplifiers were designed for the coils. To be able to perform MRI and MEG acquisitions with the same system, functions necessary for MRI were integrated in to the existing MEG acquisition software. These functions included MRI sequence generation and control for polarisation, gradient and excitation coils as well as synchronised resetting of the SQUID sensors. Virtual channels were added to the data stream to allow parallel and synchronised recording of the control signals driving the aforementioned coils. A separate image reconstruction module was programmed to read the recorded time-domain data files and to write DICOM-formatted image files. These files can readily be imported to the existing MEG-MRI integration software tool that allows visualising the 3D image as well as superimposing MEG source models on to it.
In the final part of the project, we extensively validated the final prototype by obtaining brain images and recording MEG from several subjects. The validations proved that we have accomplished our goal and built a unique hybrid MEG-MRI scanner.
Project context and objectives:
The project concept was to combine MEG and MRI by technology that will allow simultaneous MEG and ultra-low-field MRI (ULF-MRI) measurements. Our objective was to produce and validate a hybrid instrument able to perform simultaneous functional (MEG) and structural (ULF-MRI) imaging for obtaining undistorted structural brain images and time-resolved functional maps.
MEG is a non-invasive imaging technique that measures the magnetic fields that are produced by electrically activated neurons in the brain. MEG has a high temporal resolution. However, the MEG does not provide structural information. MRI is a non-invasive medical imaging technique used to visualise the structure and function of the body. It uses a magnetic field to align the magnetisation of hydrogen atoms in the body. Radio-frequency (RF)-pulses are then used to flip the magnetisation, causing the atoms to produce a rotating magnetic field detectable by the scanner. By manipulating the sample with additional magnetic fields one can gather enough information to construct an image of the body.
ULF-MRI is a new very promising alternative to conventional high-field MRI. When compared with high-field MRI, ULF-MRI will provide enhanced T1 image contrast, improve geometric accuracy, allow safer imaging for patients with pacemakers and other implants, pregnant women and infants, and reduce costs.
The hybrid imaging systems requires field-tolerant femtoTesla-range sensors. Low-temperature SQUIDs used in MEG systems do not tolerate magnetic field pulses generated during MRI. We developed field-tolerant low-temperature SQUIDs that are sufficient for MRI at ultralow fields while compatible with MEG-type sensor array. In parallel, a new type of magnetometer, called mixed sensor, based on the GMR sensor technology, has been developed within our consortium for low-field NMR. In this project, our goal was to push the sensor technology beyond the state-of-the art and fabricate sensors which are sensitive enough for MEG but can also tolerate magnetic fields generated by 3D ULF-MRI.
In addition to sensors, our goal was to develop hardware, software and 3D ULF-MRI sequences for hybrid systems. These require a development beyond the current state of the art as there was no hybrid device with a large sensor array and the whole head coverage. The developed hybrid imaging technology enables us to perform a precise alignment of MEG and ULF-MRI data. In the conventional applications of MEG the functional (MEG) and structural (MRI) imaging are performed separately. This will limit the localisation accuracy of MEG, as there is co-registration error between measurements done with two different scanners. Our hybrid system will eliminate this uncertainty related to MEG studies and will therefore contribute to improving techniques for the understanding of human behaviour and for developing new diagnostic methods and therapies. The simultaneous measurement will also simplify work flow as no high-field MRI will be necessary. In addition, the use of ULF-MRI instead of high-field MRI provides several benefits related to patient safety, enhanced contrast and reduced geometric distortions.
Project results:
Technology, concept and prototype for hybrid MEG-MRI scanner
MRI is a versatile non-invasive medical imaging technique widely used in research and clinical practice. MRI is based on the phenomenon of magnetic resonance - the property of certain materials to interact with RF electromagnetic waves when placed in a static magnetic field (usually denoted as B0 field). In a classical MRI experiment, the signal-to-noise ratio (SNR) increases with B0 strength. That's why ever since its emergence as a medical imaging method, MRI has been moving in the direction of increasing field strength. Current commercial MRI scanners use fields of several Teslas (for comparison, the magnetic field of Earth is of the order of 0.00005 Tesla).
An alternative approach known as ultra-low-field MRI (ULF-MRI) replaces a single B0 field with two separate fields, a (relatively) strong polarising field Bp and a much weaker imaging field Bi applied to the sample in succession. First suggested by Albert Macovsky and Steven Conolly in their seminal paper in 1993, ULF-MRI offers a number of advantages over the classical approach. It is more tolerant to the presence of metallic objects in the imaging volume, can better discriminate between certain types of tissues and allows open, bore-free geometry and silent operation of the imaging device.
One particularly important benefit of using a separate weak field for imaging is that it enables utilising extremely sensitive SQUID-based magnetometers for MR signal readout. SQUID sensor arrays are routinely used in MEG - recording of tiny magnetic fields generated by neuronal currents inside the brain. This makes MEG SQUID arrays a natural choice for ULF-MRI of the head. An additional benefit of this choice is that the same device can be used for both ULF-MRI and MEG, which opens possibilities for considerable cost savings and novel clinical applications.
In the course of the MEGMRI project, we have designed, constructed and tested a prototype device for combined MEG and ULF-MRI recordings. While similar hybrid MEG-MRI devices were constructed by other research groups (most notably by the Los Alamos group) before, they used only a small (less than 10) number of SQUID sensors, limiting their performance. Our device takes the concept of the hybrid MEG-MRI device one step further by employing an array of 72 sensors. Our prototype also features a number of other unique engineering solutions, such as a self-shielding superconducting polarising coil.
The measured magnetic signals are so weak that interference from ambient sources needs to be suppressed by placing the MEG-MRI device within a magnetically shielded room (MSR). Hence, the sensors can measure the subtle MEG signals. Many challenges have been overcome when designing the ULF-MRI functionality. The coil magnetising the imaged object produces a field of the same order as an everyday bar magnet. When the magnetisation has been achieved, another coil is responsible for the field in which the MR image is generated. This field is of the order of the Earth's magnetic field. The usage and interplay of these fields within the MSR has required careful considerations on sensor design, coil drive and structure, and image reconstruction.
Conventional MRI technology would not have been feasible in the ULF-MRI implementation. One key issue is noise in the images. Scaling down the fields in a conventional MRI scanner would have produced random noise images. On the other hand, with the new implementation, ULF-MRI images of the brain resolving structures down to a few millimetres. The MEG-MRI prototype has given valuable information on how to reduce the noise level. In addition, the MEG measurements with the prototype have shown results in brain activity measurements in accordance with current state-of-the-art MEG. One of the greatest advances achieved by the prototype in the conduct of MEG and MRI experiments is that both can be done in one session without moving the patient.
The MEG-MRI prototype is also a powerful tool for exploring new opportunities in ULF-MRI. While MEG research was started in the late 1960s, ULF-MRI has been explored only since the 1990s. Many things can be predicted in theory but the MEG-MRI prototype offers an experimental platform for verifying the predictions. For example, the magnetic fields generated by the coils need to be precisely tuned in conventional MRI, whereas in ULF-MRI this is not necessary. Furthermore, for technical reasons, the main coil current in conventional MRI has to be kept on constantly. The orders of magnitude weaker fields in ULF-MRI make new imaging schemes possible, such as how the MR images are encoded and later reconstructed in a computer. The MEG-MRI device has open structure when compared with a conventional MRI device, it weighs less and the ULF-MRI coils are much cheaper to construct. Great potential lies also in the diagnostic performance of the MEG-MRI prototype. It may be possible to resolve such tissue types from each other that do not generate contrast in conventional MRI. ULF-MRI opens the possibility to quantitative measurements without reference scans or targets. These opportunities may lead to unexpected but very useful application of the MEG-MRI prototype. Eventually, the MEG-MRI prototype can be considered safer than conventional MRI devices because the fields used are orders of magnitude lower and heat-inducing radiofrequency fields are not needed. The projectile danger is nearly negligible and interference with medical implants is less of a problem due to the lack of strong physical coupling mechanisms to the used magnetic fields. The imaging procedure is comfortable for the patient because no hearing protection is needed due to the silent operation of the MEG-MRI device.
We have validated the performance of our prototype MEG-MRI device in both MEG and ULF-MRI modalities. In a simple visual stimulation experiment, the prototype's MEG performance was very similar to that of the commercial MEG scanner. The MRI performance of the prototype, while being inferior to that of commercial high-field MRI scanners, clearly demonstrates the feasibility of in-vivo ULF-MRI imaging of the human brain. We were able to record the anatomical MR images of several healthy subjects and one stroke patient that clearly resolve the basic anatomical structures of the head.
Software and sequences
ULF-MRI software was developed as a Matlab toolbox and includes codes to generate images from MEGMRI devices, to check the quality of these images, and to compare them with standard high-field MRI images. Within this software, it is possible to co-register ULF-MR images with high field images to provide MEG source reconstruction a suitable volume conductor model even if low field images do not enclose the whole head.
In Aalto University, we designed and implemented new MRI sequences for ULF-MRI. We designed an optimised prepolarised spin-echo sequence which produces the maximum contrast-to-noise ratio between brain white and gray matter. Another novel sequence invented and patented in this project was the polarisation encoding; the prepolarising field is varied in a way to gain additional information about the sample.
The software integration for the final prototype was done at Aalto University. The sequence generation part was developed to be compatible with the Elekta MEG software used to control the final prototype: A Matlab interface is used to specify the desired sequence. The sequence file is designed to be compatible with the Elekta MEG system. The instructions from the file are read by the Elekta MEG system, which controls the MRI coils and SQUID sensors. The software allows nearly arbitrary MRI sequences. MRI signals from the sensors are acquired with the Elekta MEG software. The acquired data is processed with Matlab to produce images.
To be able to perform MRI and MEG acquisitions with the same system, functions necessary for MRI were integrated in to the existing MEG acquisition software. These included MRI sequence generation and control for polarisation, gradient and excitation coils as well as synchronised resetting of the SQUID sensors. Virtual channels were added to the data stream to allow parallel and synchronised recording of the control signals driving the aforementioned coils. A separate image reconstruction module was programmed to read the recorded time-domain data files and to write DICOM-formatted files. These files can readily be imported to the existing MEG-MRI integration software tool that allows visualising the 3D image as well as superimposing MEG source models on to it.
Low filed MRI system based on mixed sensors
Commissariat à l'energie atomique (CEA) has developed a 3D MRI test system working at 77 K for mixed sensors. The system is placed outside of a shielded room. The coil system is made of four main coils with three static gradients. The homogeneity achieved is 10 ppm and the main field is variable up to 10 mT. An additional set of three coil pairs create the dynamic gradients for the imaging control. A spectrometer has been developed. This spectrometer is ready and can control the frequencies, phase and amplitude of the RF frequency and the gradients. Echo planar sequences have been developed and 3D images can now be acquired. This test system has allowed demonstrating the possibility of performing 1 mm resolution images in a volume of 6 x 6 x 6 cubic centimetres. A full MRI setup for brain imaging working at 10 mT with a homogeneity volume of 20 x 20 x 20 cubic centimetres was built during the last part of the MEGMRI project.
Technology for field-tolerant low-critical temperature SQUIDS
Low critical temperature superconducting quantum interference devices (low-Tc SQUIDs) are ultrasensitive sensors of magnetic field capable of measuring magnetic fields down to the range of 1 - 10 fT, i.e. field levels of about one billionth of the Earth's magnetic field. Due to this, they have an established position in MEG systems as typical signals from neuronal activity are in the range of 1 pT or less. As MEG measurements are performed in magnetically shielded environments, the SQUIDs are not exposed to large external magnetic fields. However, in MRI the measurement sequences involve large magnetic fields. Even in ultra-low-field MRI the measurement field is typically in the order of tens of microtesla and the prepolarisation field in the order of tens of mT. In practice, the SQUID sensor needs to be able to measure small field variations down to its detection limit while the measurement field is on. Also the SQUID needs to recover from the impact caused by the prepolarisation field within a reasonable time (milliseconds in practice). This, in particular, is to be taken into account in sensor design as the prepolarisation field is typically a factor of 1000 billion times larger than the detection limit of the SQUID.
Practical SQUID sensors are based on superconducting thin-film technology. As itemised below, most challenges related to using the sensors with magnetic field sequences are due to certain undesired properties of superconducting thin films in under large magnetic fields. The most conventional solution to avoid thin-film exposure to magnetic fields in ULF-MRI is to place the SQUID sensor in a tight magnetic shield while the superconducting wire-wound pickup antenna is left outside of the shield. Furthermore, the pickup antenna is typically wound in axial geometry in such a way that it only detects the second-order gradient of the field thus enabling to large extent the separation of fields to be measured from other field components. In a typical MEG system, on the other hand, the sensor, including the SQUID and the pickup, are planar first-order gradiometers and magnetometers. The whole sensor can also be fabricated with thin-film technology. Such sensor is also convenient in MEG application consisting of hundreds of SQUIDs in a helmet configuration. It has certain benefits in industrial manufacturability.
In this project, our approach was to study the possibility of extending the applicability of MEG-type sensor to ULF-MRI while maintaining the properties important in MEG. In the course of doing so, we identified the main issues to be solved in sensor design. Also, a fabrication process enabling sub-micrometer features in SQUID designs was developed as narrow linewidths are expected to improve the field tolerance.
The most obvious problem is that superconducting thin films tend to trap flux. This means generation of vortices, the quanta of magnetic flux, within the films during the exposure to a large magnetic field. The vortices may remain trapped even when the external field is removed. The related magnetic flux may couple to the active elements of the SQUID circuit, the Josephson junctions, killing the magnetic field sensitivity of the device. In the project, we designed and fabricated a number of SQUID models and tested their tolerance against prepolarisation field pulsing. Although with some SQUID models the threshold field of about 10 mT was achieved, it was concluded that for ULF-MRI application some improvement is still preferable. Therefore, we developed partially shielded SQUID package. With this arrangement the field at the location of the SQUID was significantly attenuated but while the total flux penetrating the pickup loop was reduced only slightly (about 10 %) thus not compromising the sensitivity. Using the shielded module, sensors recover from field pulsing at least 50 mT.
Another problem we discovered was noise transients of SQUIDs that emerged after the prepolarisation pulse. The underlying reason was deduced to be random vortex motion in the structures of the pickup circuit after the magnetic field was removed. It was found that wide-line pickup circuits typically used in MEG were particularly bad in this respect generating magnetic field noise spectral densities up to about 1 pT/sqrt(Hz) in the time window needed for MRI measurement. As this is completely unacceptable for imaging purposes, we fabricated and tested pickup circuits with varying techniques and design parameters. It was found that narrow-line thin-film pickups and wire-wound pickups showed little or no excess noise in the frequency range of interest for ULF-MRI.
Further issues that needed to be taken into account were the magnetisation of the superconducting structures in the sensor potentially spoiling the homogeneity of the measurement field.
As a final step a sufficient number of sensor modules were fabricated and tested for the final prototype. The noise level corresponding to a single sensor in the final prototype was measured to be 4-5 fT/sqrt(Hz) in the frequency range used in imaging. Within imaging sequences, the sensors recovered from magnetic field transients under prepolarisation of 22 mT used in imaging in such a way that the noise was not elevated during MRI acquisition. The noise contribution of the SQUID sensor itself was estimated to be 2-4 fT/sqrt(Hz) depending on the channel as estimated by cross-correlation measurements between adjacent channels.
Technology for field-tolerant high-critical temperature SQUIDS
The Chalmers Tekniska Hoegskola developed high-critical temperature (high-Tc) dc SQUID sensors for integration into a hybrid test MEGMRI system. Through the course of this project, these sensors were pushed beyond the state-of-the-art in terms of combining ultra-high sensitivity (< 20 fT) and tolerance to strong pulsed magnetic fields (> 10 mT). The high-Tc SQUIDs were successfully implemented in small-scale MEG and ULF-MRI test systems built at Chalmers during the course of the project.
High-Tc SQUIDs can be operated at liquid nitrogen temperature (77 K) that significantly simplifies cooling requirements when compared with the liquid helium that is used for low-critical temperature SQUIDs. This may be advantageous, especially for MEG where the separation between the SQUID and scalp should be minimised. However, the magnetic field sensitivity of single-layer high-Tc SQUID magnetometers is typically a factor of 10 lower than equivalent low-Tc SQUIDs. A significant limitation of the single layer magnetometer design is the very large inductance mismatch between the pick-up loop and the SQUID loop that reduces the effective area of the sensor (that equivalently increases the magnetic field-to-flux transformation coefficient, and thus the sensitivity of the sensor). To improve the transformation coefficient and thus the magnetic field sensitivity of the sensors, flux transformers with a multi-turn input coil should be used. We developed both types of high-Tc SQUID sensors: directly coupled (dc) single layer SQUID-magnetometers and flip-chip SQUID magnetometers with multilayer superconducting flux transformers.
The dc-coupled SQUIDs were fabricated in a single layer of superconducting YBCO film on a bicrystal substrate. The dc-coupled SQUID magnetometers demonstrated a magnetic field sensitivity of about 35 fT/sqrt(Hz) with excellent low-frequency noise characteristics below 10 Hz, which is important for, e.g. MEG. In addition, these SQUIDs had flux dams incorporated in the pickup loop that suppressed the supercurrent induced by the large pre-polarisation pulses with magnitudes up to 1 mT.
The main challenge for the fabrication of multilayer structures in high-Tc superconducting materials is to obtain c-oriented film on the entire length of the top electrode. In order to provide proper conditions for the growth of c-oriented high-Tc film, very shallow edge slopes of the bottom electrode need to be produced. Another important issue is the smoothness of the bottom electrode because the presence of droplets and precipitates on the surface of the bottom electrode can lead to the formation of unwanted short-circuits between layers. In order to solve these problems, we developed a polishing process for fabrication of multilayer structures in high-Tc films. The polishing method proved to be extremely effective in reducing the step angle of the bottom layer, thereby drastically increasing the Tc and the Jc of the top electrode without degrading the properties of the bottom layer nor introducing unwanted shorts through the insulating layer. The polishing technology has been successfully applied for the fabrication of multilayer flux transformers on 10x10 mm SrTiO3 substrates. Measurements of the flip-chip magnetometers were performed using various types of SQUIDs fabricated on bicrystal substrates. The best gain was achieved for a single washer SQUID. The flip-chip magnetometers showed an increased level of low-frequency noise compared to a single SQUID. Further optimisation of fabrication technology and design will be needed in order to reduce the low-frequency noise.
We have successfully implemented the developed high-Tc SQUIDs in both MEG and ULF-MRI test systems, developed at Chalmers during the project. In these applications, we have used dc- coupled SQUIDs. Despite the fact that flip-chip SQUID magnetometers have demonstrated better sensitivity above 1 kHz, their low-frequency noise was still higher than in dc-coupled SQUIDs.
As a proof-of-principle, the recordings of spontaneous alpha- and mu-rhythms from a human brain were done using the dc-coupled single layer SQUID magnetometers. By measuring brain signals in two different states (eyes open / shut for alpha and hands relaxed / flexed for mu), one can distinguish the activated alpha- and mu-waves. The performed MEG recordings of spontaneous brain activity using our high-Tc SQUID magnetometers showed that despite higher noise-levels compared with their low-Tc counterparts, high-Tc SQUIDs can be used to detect and record physiologically relevant brain rhythms with comparable signal-to-noise ratios.
Mixed sensore for the detection of ultra-low fields
CEA has developed superconducting / magnetoresistive mixed sensors. These sensors are based on a bridge of micro sise giant magnetoresistive elements which sense the current circulating in a one-square-centimeter superconducting loop. The giant magnetoresistive elements based on spin electronics are presently the most sensitive micro size magnetic sensors and are implemented on all hard disks for data storage. The superconducting loops act as a flux-field concentrator with a gain of more than 1000. These sensors are competitive in terms of sensitivity to low Tc SQUIDs at frequencies above 10 kHz but present a 1/f noise which limits the sensitivity at 1 Hz to several hundred fT. During the MEGMRI project, several key improvements have been performed: the optimisation of flux transformers to couple mixed sensors to the external world, the development of mixed sensors on large wafers and the reduction of 1/f low frequency noise by using a modulation of the supercurrents circulating in the superconducting loop. These sensors are working at 77 K and may be used for ultra-sensitive magnetometers and RF detectors for low field MRI or antennas. They are more sensitive than tuned coils below 500 kHz. Their advantage compared with SQUIDs is the strong immunity against RF perturbations and strong RF pulsed allowing a working outside magnetic shielded rooms for low field MRI.
Potential impact:
Impact on clinical applications
The developed hybrid imaging technology has many potential clinical applications. In this project, we identified two major applications.
In patients with pharmacoresistant epilepsy, surgical resection of the epileptogenic cortex is useful. The help of 3D functional imaging of the brain by magnetoencephalography with a reference structural mapping is valuable in locating the epileptogenic area and minimising damage to irretrievable cortical areas during operation. With simultaneous recording of MEG and MRI, the reliability of the signal superimposition on the anatomy is increased; this may reduce the need for intraoperative recordings in the future. The intraoperative functional recordings prolong operations and increase the risk of complications.
The enhanced T1 image contrast of ULF-MRI may provide new applications in oncology diagnostic imaging. The associated MEG would provide simultaneous presurgical functional mapping of brain areas.
Impact on pediatric studies
The developed hybrid imaging system is completely non-invasive and it has an open design which is much more convenient for small children than closed MRI systems. In addition, the ULF-MRI part in the hybrid system has no strong magnetic field. The gradient coils do not create the disturbing noise associated with high-field MRI. The RF coils will deposit less power to the tissue than those used for high field MRI and at lower frequencies. Consequently, ULF-MRI and hybrid MEG-MRI systems are particularly suitable for pediatric recordings. This makes possible to image pediatric brain diseases which were not possible or practical with the present instruments.
Impacts on neuroscience
The combination of functional imaging by MEG and structural imaging by MRI solves one of the most difficult barriers to exact matching of MEG and MRI coordinate systems. The precise alignment of MEG and MRI data eliminates an uncertainty in MEG studies and contributes to improving techniques for the understanding of human behaviour. The MEG-MRI hybrid system has a direct impact on the understanding of the link between neuronal activity and behavioural performance because functional imaging with high dynamics and high resolution is the best tool for monitoring precisely the neuronal activity and correlates it with stimuli and / or brain operations.
Exploitation of hybrid MEG-MRI technology
Inside the consortium, the commercial exploitation for MEG-MRI hybrid systems is in the interest of Elekta which is the world leader of MEG systems. We have estimated that the cost for hybrid MEG-MRI scanner would be only 30 % higher than conventional MEG unit. Assuming that we can clinically demonstrate hybrid scanners superb imaging power over other technologies in some specific clinical application (e.g. imaging of brain tumour or epilepsy patients), then new markets will appear. In this application scenario, we have estimated that the size of the market would be 60 - 100 MEUR annually.
Before commercial exploitation, further improvement of the prototype is needed. This is in the interest of Aalto, which will continue research on hybrid MEG-MRI technology. If this is successful, then Elekta can make the final decision about the commercial exploitation.
Impact and exploitation on high-Tc SQUID technology
Since their discovery in 1986, much effort has been made to use high-Tc superconducting materials in place of their low-Tc counterparts in many applications, mainly due to the possibility of using cheaper and more flexible cooling technology. The complexity of high-Tc superconducting materials has been the main obstacle on the way to their successful implementation.
We have demonstrated that the simple and high-yield bicrystal grain boundary-based YBCO SQUIDs were sufficiently sensitive to detect spontaneous signals generated by the brain at frequencies below 20 Hz, yielding a SNR of approximately 10. The SNR is similar to that which is obtained with low-Tc SQUIDs, despite the fact that the noise level of high-Tc SQUIDs is roughly an order of magnitude higher. These results suggest that high-Tc technology may supplement or replace its conventional low-Tc counterpart in future MEG systems.
Further improvement of the sensitivity of high-Tc SQUID magnetometers requires reliable and high-quality multilayer technology. A new polishing method developed for fabrication of high-Tc multilayer structures improves surface-smoothness and step edges, thereby increasing yield. Combined with bicrystal high-Tc SQUID sensors, these transformers enabled at least a four-fold gain in sensitivity when compared to standard single-layer high-TC SQUID sensors of the same size. Such technology can be scaled to large wafers and is likely to increase magnetic field sensitivity of high-Tc SQUID such that they can truly begin to compete with their low-Tc counterparts.
In future applications, micro-cryocooling technology can replace the liquid nitrogen cryostats used to cool the sensors, thereby reducing the size of high-Tc SQUID systems and eliminating cryogen consumption. Furthermore, it may be possible to develop flexible MEG helmets using the micro-cooling technology. An array of high-Tc SQUID magnetometers in a flexible helmet will increase spatial resolution of MEG recordings and enable the study of new sources of brain activity, including higher-order and higher frequency sources. Initial work has been performed in collaboration with Kryoz Technologies BV (Enschede, the Netherlands) and preliminary results showed very high potential of the microcooling technology.
Future work will also focus on continuing development of the high-Tc SQUID MEG and ULF-MRI systems and their assessment for clinical applications. Specifically, the low- and high-frequency MEG signals are worthy of further investigation, both from a technical as well as a neurophysiological points of view. This work is performed in collaboration with the Sahlgrenska University Hospital and the University of Gothenburg.
Impact and exploitation of field-tolerant low-temperature SQUIDS
Robustness against external magnetic fields in SQUID technology is a benefit in most applications including but not limited to NMR and MRI based applications. For example, within the project we demonstrated SQUIDs capable of operating in Earth's magnetic field with no external shields. We also developed sub- µm SQUID fabrication technology with increased production capacity. The process has recently been adopted to commercial SQUID fabrication and it is and will be exploited in other ongoing and future development projects including e.g. development of SQUID-based readout for submillimetre wave and X-ray imaging systems for astronomy. A sub-mm imager can further be used e.g. as a sensitive security camera. Potential applications include also geophysics. As an example, hundreds of superconducting rock magnetometers have been installed worldwide to measure remnant magnetisation of geological samples. The number of annual installations is about ten systems consisting of typically three SQUID channels each. Recently, SQUIDs have been used in geophysical explorations to find more and other deposits of natural resources deep underground.
The field-tolerant SQUIDs developed in this project have applications as indicated above. Inside the consortium, the commercial exploitation of developed sensor technology is in the interest of the Aivon which is already producing sensors for conventional Elekta-based MEG systems. In addition VTT is continuing the research related to field-tolerant SQUIDs. Aivon and VTT have technology to commercialise the developed sensors as soon as there is demand for them. Currently, the commercial exploitation depends on the exploitation of developed imaging technologies. If these new technologies such as hybrid MEG-MRI or ULF-MRI are commercialised, then there will be big markets for sensors.
Impact and exploitation of spin electronics based magnetic sensors
Spin electronics based sensors are constantly improved and the MEGMRI project has given the opportunity of developing a new generation of very sensitive spin electronics based magnetic sensors. These new sensors present at room temperature the sensitivity of a few pT/sqrt(Hz). The applications of these sensors are magnetometers for earth and planetary field magnetic imaging for cultural heritage, space magnetometry and petrol applications, buried object detection for environment and security applications and medical applications, mainly magnetic detection based biochips. Mixed sensors technology at 77 K has applications for ultra-sensitive magnetometry mainly for research purpose and for medical applications and in particular ultra-low field MRI. CEA has a number of patents on spin electronics based very sensitive magnetic sensors. Mixed sensors production at the wafer scale by a European company is already in place. CEA is presently investigating a route for very high sensitive room temperature sensor volume production.
Impact and exploitation of ULF-MRI based on mixed sensors
ULF-MRI systems are a low cost and easily implementable scanners with acceptable MRI resolution for a number of clinical applications: strokes, cancer detection. Enhanced contrasts in tumours recently demonstrated in mT fields opens new possibilities for ULF-MRI. The impact of this technology, in particular based of mixed sensors may be huge. The threshold for the development of that technique at the large scale will be the demonstration of a full head image with a resolution better than 2 mm in a time less than 10 min. Interest of ULF-MRI is also for patients where high field MRI is not possible like patients with metallic implants or premature babies. A full head system has been built and a protocol for clinical tests has been requested. As soon as a clear demonstration of the capability of this system for brain imaging is demonstrated, CEA valorisation will put in place a route for exploiting this new approach of MRI.
Dissemination - Arranged events
The project consortium arranged several events related to the MEGMRI project. All of these events were open to external audience. The list of organised symposiums, workshops, seminars and training events are given below:
- The Biomag 2010, Satellite Symposium 'Ultra-low-field MRI and its combination with MEG' was organised on 28 March 2010, Dubrovnik, Croatia.
- Brainstorming Workshop - MEGMRI 'The dream you can realise through the hybrid MEGMRI system: identifying the "killer application" of the new-generation scanner', was organised on 16 November in Rome.
- 'Devices for brain research and diagnostics' (open seminar) was organised on 3 June 3 2009 in Gothenburg.
- A training course on 'Flux3D modelling' was arranged by Cedrat in November 2009.
- A training course on '3D-MRI techniques' was arranged by CEA in Helsinki, on 2 June 2010.
- A MEG course was arranged by UdA and Elekta in Chieti, Italy, 31 May 2011.
Dissemination - Publications and thesis
During the project, we have published our main scientific results in top journals and conferences. In total, we have published 24 peer-reviewed articles and several other publications. In addition, two PhD thesis and four MSc thesis were done in this project.
Website: http://www.megmri.net/
Contact:
Prof. Risto Ilmoniemi
Aalto University
Department of Biomedical Engineering and Computational Science (BECS)
P.O. Box 12200
FI-00076 Aalto
Tel.: +35-850-5562964
Email: risto.ilmoniemi@aalto.fi
Participants:
Aalto University - Finland coordinator
Prof. Risto Ilmoniemi (risto.ilmoniemi@aalto.fi via e-mail)
Aivon Oy- Finland
Dr Jari Penttilä (jari@aivon.fi via e-mail)
Cedrat Technologies SA- France
Mr Patrick Meneroud (patrick.meneroud@cedrat.com via e-mail)
Chalmers University of Technology - Sweden
Prof. Dag Winkler (dag.winkler@chalmers.se via e-mail)
Università degli Studi Gabriele d'Annunzio Chieti-Pescara - Italy
Prof. Gian Luca Romani (glromani@itab.unich.it via e-mail)
Commissariat à l'energie atomique - France
Prof. Claude Fermon (claude.fermon@cea.fr via e-mail)
Elekta AB - Sweden
Dr Antti Ahonen (antti.ahonen@elekta.com via e-mail)
Associazone Fatebenefratelli per la Ricerca - Italy
Prof. Paolo Maria Rossini (paolomaria.rossini@afar.it via e-mail)
Hospital District of Helsinki and Uusimaa (HUS) - Finland
Dr Jyrki Mäkelä (jyrki.makela@hus.fi via e-mail)
Imaging Technology Abruzzo s.r.l - Italy
Prof. Antonello Sotgiu (sotgiu@univaq.it via e-mail)
Physikalisch-Technische Bundesanstalt - Germany
Prof. Lutz Trahms (Lutz.trahms@ptb.de via e-mail)
Università degli Studi di Parma - Italy
Prof. Giacomo Rizzolatti: giacomo.rizzolatti@unipr.it
VTT Technical Research Centre of Finland - Finland
Dr Juha Hassel (juha.hassel@vtt.fi via e-mail)