Novel MRI diffusion method for characterization of malignant brain tumor microstructure and assessment of early treatment response
Using diffusion MRI to study demyelination in cortex and deep gray matter in animal model of multiple sclerosis
We have applied a biophysical model of diffusion to study dendrite density and diffusion in cortex and deep gray matter in an animal model of multiple sclerosis. Our study shows the potential of diffusion MRI to detect subtle changes in myelin content in gray matter, thereby improving out understanding of the disease.
Conventional MRI techniques have a very low sensitivity to gray matter demyelination. However, novel MRI techniques show potential to detect myelin content in gray matter. We have investigated whether diffusion-weighted MRI that is based on biophysical models of tissue microstructure can detect demyelination in deep gray matter and cortex in the curpizone mouse model of multiple sclerosis. Twenty female were divided randomly into 4 groups of 5 mice each. Three of the groups were exposed to cuprizone rich diet by adding cuprizone to milled mouse chow and then sacrificed after 3 and 5 weeks of cuprizone exposure, and 4 weeks after ending the 5-week cuprizone exposure. The control mice were sacrificed at the end of the experimental period. Mice were euthanized by asphyxiation, the brains were then extracted and stored at 4 degrees C in formalin. Two days before scanning, the brains were rinsed in saline to remove formalin and then secured inside a plastic tube that was filled with oil (perfluoropolyether) which improves image quality and is not visible on the MR images. Scanning was performed on a 7T horizontal-bore magnet using a mouse head quadrature volume resonator. Diffusion images (resolution 98µm x 98µm, slice thickness 0.75 mm) were acquired using a Stejskal-Tanner spin-echo diffusion preparation with b-factors ranging linearly from 880 to 14080 s/mm2. Total scan time was 3 hours 18 min. Images were analyzed in MATLAB using a dendrite density model of gray matter which assumes that the MR diffusion signal originates from two components: (i) the dendrites and axons, modeled as long cylinders with one diffusion coefficient parallel to the cylindrical axis (D_L), and (ii) an isotropic mono-exponential diffusion component describing water diffusion within extracellular space (D_ext). The results of the analysis were parametric maps of neurite density, D_L and D_ext. Region-of-interest analysis (ROI) was performed in the right cortex (CX) and in deep gray matter (DGM) within a slice which was the same for all animals. Mean values within the ROIs were recorded and used to compute the population mean and standard error. Wilcoxon rank sum test was used to compute significant differences between groups. There was a visible reduction in neurite density in DGM and CX after 5 weeks of exposure, and remyelination 4 weeks after terminating the exposure. We detected significant differences in the values of neurite density and D_L between baseline and at 3 and 5 weeks of cuprizone exposure in both DGM and CX. After 3 weeks of exposure, neurite density decreased by 16% and D_L increased by app 20%. There were no significant differences in values between 3 and 5 weeks of exposure. Both, the neurite density and D_L returned to the baseline values after remyelination. D_ext in DGM and CX did not change significantly as a result of cuprizone exposure and had mean values between 0.27 and 0.35 µm2/ms for all experimental groups. Our measurements showed degree of demyelination in the cuprizone mouse model of multiple sclerosis correlated well with the neurite density and intra-axonal diffusion parameter of the dendrite density model. These parameters could therefore serve as markers of demyelination and improve the diagnosis and treatment monitoring of multiple sclerosis.
Using advanced diffusion MRI techniques and biophysical models of tissue microstructure to detect pathology in the brain
The goal of project one was to use a biophysical model of diffusion in tissue to detect demyelination in deep gray matter in animal models of multiple sclerosis using DW-MRI. The goal of project two was to develop an imaging version of the modulated gradient spin-echo technique (MGSE) and to test its potential on a metastatic ex-vivo mouse brain.
This year I have focused on 2 projects. First, in collaboration with Dr. Stig Wergeland (Norwegian Multiple Sclerosis Competence Center), Dr. Erlend Hodneland (Christian Michelsen Research, Bergen, Norway), Dr. Renate Gruner and Vanja Flatberg (Department of Physics and Technology, University of Bergen) I continued working on quantifying the loss of myelination in mice models of multiple sclerosis using a diffusion model of neurite density that was developed by Dr. Sune Jespersen at Aarhus University. Second, I spent 5 months abroad, in the NMR group lead by Dr. Igor Serša at the Institute Jožef Stefan (IJS) in Ljubljana, Slovenia, developing novel diffusion-weighted pulse sequences that would be able to improve the contrast between normal and cancerous tissue. Project 1: Multiple sclerosis is a chronic inflammatory demyelinating disease of the central nervous system, characterized by demyelinated lesions in white as well as gray neuronal matter. The conventional MRI techniques have a very low sensitivity to gray matter demyelination, however, in the last decade, novel MRI techniques such as magnetization transfer imaging (MTI) showed the potential to detect myelin content in gray matter. In 2014, I initiated a project aimed at detecting demyelination in deep gray matter (DGM) in mice exposed to cuprizone using diffusion-weighted (DW) MRI. The goal was to determine weather any of the parameters of the Jespersen’s diffusion model, such as neurite density and intra- and extra-axonal diffusion, could serve as markers of gray matter demyelination. The experiments were performed at the end of 2014, beginning of 2015. Our preliminary results, which I reported on in 2014, showed a clear loss of neurite density and an increase in longitudinal intra-axonal diffusion in white matter tracks. More importantly, a detailed analysis that we performed in the second half of 2015 showed the same trends in the cortex and DGM. Moreover, the results of DW-MRI correlated well with those obtained with MTI technique as well as with histology. We are currently writing a manuscript to report on our findings. Project 2: The modulated-gradient spin-echo (MGSE) pulse sequence that was introduced by Dr. Janez Stepisnik at ISJ encodes diffusive spin motion with a train of p RF pulses that are applied during a constant field gradient. Unlike the standard diffusion preparation that uses a pair of pulsed gradients separated by one p pulse, the MGSE technique is able to probe diffusive motion at much higher temporal frequencies. It is possible to reach frequencies in the range of a couple of kHz (below 1 ms diffusion times). At such short diffusion times, the MRI signal becomes sensitive to structural changes occurring at the level of 1 um. If successful, this technique would have big diagnostic potential in differentiating normal from pathological tissue. At the IJS in Ljubljana I implemented the MGSE pulse sequence on a vertical, 400 MHz micro-imaging system. In order to develop an imaging version of the technique, a few challenges had to be solved. For instance, applying the RF pulses during a constant gradient pulse produces slice selection, so the imaging slice had to be adjusted to fit within the slice produced by the diffusion gradients. I then tested the performance of the pulse sequence on water and various biological samples, such as muscle, and tumor tissue. The preliminary results look promising, so I plan to continue this work in 2016.
Applications of novel diffusion-weighted MRI methods and biophysical models of tissue microstructure
The main goal of the project was to use a biophysical model of neurite density to detect myelination losses in deep gray matter in animal models of multiple sclerosis using diffusion-weighted magnetic resonance imaging.
Diffusion-weighted (DW) MRI provides a noninvasive tool to probe tissue microstructure. However, a precise correspondence between physiological or biophysical properties and the obtained diffusion parameters remain uncertain due to lack of specificity. The many immerging biophysical models of diffusion have to be applied first in preclinical setting, to allow for a comparison between the in-vivo and the ex-vivo imaging methods. Once the models are validated, they hold a great potential for characterizing tissue microstructure in-vivo, identifying pathology and evaluating early response to treatment. One of such models was developed by S. N. Jespersen et al.  and is based on the assumption that water diffusion can be described in terms of two non-exchanging components. One diffusion component is associated with diffusion in cylindrically symmetric structures, such as dendrites and axons (ie, neurites), with exchange of water being negligible on the time scale of the diffusion imaging experiment. The net signal from this component is a sum of the signal from all neurites, weighted by a probability density function specifying the number of neurites in every direction. The second component of the diffusion signal accounts for diffusion in extracellular space, cell bodies and glia cells, and is thus best described by as a hindered isotropic diffusion with an effective diffusion constant. One of the strengths of this model over the traditional DTI is that it better describes diffusion in anisotropic structures, such as dendrites in gray matter, which appear isotropic on DTI due to a uniform orientation distribution. Jespersen et al. validated their model by histology and tested its capabilities to determine neurite loss in a rat model of chronic stress . My main research focus this past year has been to apply the neurite density model to quantify loss of myelination in mice models of multiple sclerosis. In particular, we want to study changes in deep gray matter and relate them to changes observed with the magnetization transfer ratio (MTR) imaging. Although this research does not directly relate to our main project goal of studying tumor microstructure, the imaging techniques that we are implementing on our system can be applied to a wide variety of problems. Since I have collaborated with researchers (Dr. Stig Wergeland) at the Norwegian Multiple Sclerosis Competence Center in the past , it seemed natural to build up on our previous work and to use the neurite density model to quantify demyelination in different brain regions. I first implemented the relevant diffusion protocol on the preclinical MRI system. We then scanned 6 ex-vivo mice brains from each of the following 4 groups: controls, 3-weeks and 5-weeks of cuprizone exposure, and 2 weeks after termination of curprizone exposure. I have modified the post-processing script from S. N. Jespersen to read and analyze our data. Preliminary results show visible neurite density losses in corpus callosum and other white-matter tracks after 5 weeks of cuprizone exposure, and an increase in neurite density after termination of exposure. However, a detailed statistical analysis will have to be performed to quantify possible differences in deep gray matter. 1. Jespersen SN, et al. Neuroimage, 2007, 34(4):1473–86; Jespersen SN, et al. Neuroimage, 2010, 49(1):205–16. 2. Vestergaard-Poulsen P, et al. PLoS ONE. 2011, 6(7):e20653. 3. Fjær S, et al. PLoS ONE, 2013, 8(12):e84162.
Novel MR diffusion methods for imaging of brain tumours
The main goal of the project is to implement and test novel diffusion MRI methods and tissue microstructure models for better characterization, diagnosis and treatment response of malignant brain tumors.
Diffusion-weighted MRI provides a noninvasive tool to probe tissue microstructure and fuels the hope that the MRI scanner may eventually serve as an in-vivo microscope of the brain and body. However, a more precise correspondence between physiological or biophysical properties and the obtained diffusion parameters remain uncertain due to lack of specificity. The many immerging biophysical models of diffusion have to be applied first in preclinical setting, as these allow for a comparison between the in-vivo and the ex-vivo imaging methods. Once the models are validated, they hold a great potential for characterizing tissue microstructure, identifying pathology and evaluating early response to treatment.
The focus in 2013 was to test the capabilities of the newly upgraded preclinical MRI system, especially in relation to diffusion imaging. I was able to obtain high-quality ADC (apparent diffusion coefficient) and DTI (diffusion tensor imaging) using conventional diffusion protocols. These protocols are now available to other scientists working in the preclinical imaging field in Bergen. My next goal is to streamline diffusion-data processing, either by developing data-processing software in-house, in collaboration with experts within image visualization group, or by utilizing noncommercial (“for academic use”) software, or both.
In parallel to this work, I used my new knowledge of pulse-sequence and methods scripting on the preclinical Bruker system to begin implementing the OGSE (Oscillating Gradient Spin Echo) DW MRI method. As reported before, this task proved to be more complex than initially thought, because the pulse sequence cannot be implemented as a simple modification of the commercially available diffusion method. Despite this setback, I have made significant progress and hope to have the OGSE method working by the summer of 2014. In addition, my knowledge of pulse-programming language now enables me to participate in research projects that require special imaging protocols not implemented on our scanner.
In 2013 I continued my collaboration with the cancer group at the Institute of Biomedicine at the University of Bergen, studying the effect of anti-VEGF cancer treatment on tumor stroma. I coordinated and assisted with MRI scanning and analyzed the majority of MR data. The manuscript on this work is in preparation.
I have also initiated two international collaborations, one with a G. Sanguinetti, currently at the INRIA Sophia Antipolis in France, and one with S. Jespersen at Aarhus University in Denmark. Both collaborations focus on testing how well different biophysical models of tissue microstructure describe diffusion-weighted MR signal. Unlike signal models of diffusion, the biophysical models attempt to relate diffusion signal to tissue structure, such as cell and fiber density, compartment sizes, membrane permeability, and various cytoarchitectural attributes. G. Sanguinetti is testing a modified version of the co-called AxCaliber imaging method which infers the distribution of axon diameters within each white matter voxel. My contribution to the project is to collect the relevant MR images in order to validate the model. The collaboration with S. Jespersen involves using his biophysical diffusion model to measure dendrite density, a parameter of great importance in many brain disorders.
Novel MR diffusion method for imaging of brain tumors
The main goal of the project is to implement and test a novel diffusion-weighted MRI method (called oscillating gradient spin-echo or OGSE technique) for better characterization, diagnosis and treatment response of malignant brain tumors.
Diffusion-weighted (DW) MRI is commonly used to depict variations in the rate of water diffusion within cancer tissue. Self-diffusion of water molecules is restricted by the presence of barriers a molecule encounters during an imaging experiment, such as cell membranes and subcellular organelles. Since the apparent diffusion coefficient (ADC) depends on the microstructural alterations in the tissue, changes in ADC may be used to evaluate pathophysiological changes, including the response of tumors to treatment.
Conventional diffusion-weighted imaging techniques are mainly sensitive to changes in tissue cellularity. On the contrary, the emerging techniques using OGSE (oscillating gradient spin echo) at high gradient frequency are capable of probing variation in tissue structure occurring on spatial scales much smaller than a single cell. OGSE DW-MRI can therefore provide a more sensitive probe of tissue microstructure, differentiating between different stages and types of cancer. In addition, the technique can be potentially much more sensitive to detect early response to anti-cancer treatment.
In 2012, I continued to focus on implementing the OGSE DW MRI method on the preclinical 7T MRI system at the University of Bergen. I was learning the programming language of Paravision, which is used to design and control the execution of MR pulse sequences on the scanner. I started to modify the existing diffusion-weighted protocol with the goal of replacing the standard Stejskal-Tanner gradients with oscillating gradient waveform. However, since some of the existing Paravision libraries are not accessible to the users, I had to change the approach and started to program the OGSE pulse sequence from the basics. This is a bigger task then expected.
In line with the proposed plan, I tested the conventional PGSE (pulsed-gradient spin echo) DW MRI method on animals with brain tumors. I also participated in a research project that addressed the effect of antiangiogenic cancer treatment on the brain stroma. As part of this project I optimized the conventional diffusion pulse sequences on the preclinical scanner. However, from these measurements it became obvious that the scanner hardware was suboptimal for the collection of DW-EPI (diffusion-weighted echo-planar images), which required a better field homogeneity than the one we were able to achieve with the old hardware. I therefore spent a significant amount of effort (through my 50% engineering position at the Molecular Imaging Center) to raise the money for a substantial scanner upgrade. The result of these efforts was a 5.5 million NOK upgrade of the scanner (in Dec 2012), which included a new shim and gradient coil, and a stronger gradient amplifier that is needed for the implementation of OGSE DW MRI method. I am in the process of testing the new hardware and optimizing the conventional diffusion protocols (diffusion-weighted as well as diffusion tensor imaging). Preliminary results on rats with brain tumors are very promising and indicate that the cancer-research groups at the University of Bergen will be able to use the conventional PGSE diffusion MRI methods as standard protocols in the study of brain tumors. This is an important milestone that had to be achieved before implementation of more advanced diffusion protocols, such as OGSE diffusion method.
I also took part in a course organized by the European Society of Magnetic Resonance in Medicine and Biology (ESMRMB): “Diffusion, what it means and how to measure it”. The course not only presented advanced theory of water diffusion in biological tissue, but also novel ways of interpreting diffusion data, a topic that was very relevant for my research. I was also able to strengthen my connections to the international scientific community that is studying diffusion and interpretation of diffusion data in biological tissue.
Finally, I started to collaborate with Dr. John Georg Seland from the Department of Chemistry, University of Bergen. We are investigating model systems (composed of beads of various sizes) in which we can separate the effects of restricted diffusion from effects caused by multi-component diffusion. This would enable us to determine the applicability of various diffusion models to study tissue microstructure.
Novel MR diffusion method for imaging of brain tumors
The goal of the proposed project is to implement and test a novel MRI method (called oscillating gradient spin-echo or OGSE technique) for better characterization, diagnosis and treatment response of malignant brain tumors.
Non-invasive imaging of brain tumors is an essential clinical tool in evaluation of tumor grade, in surgical planning, and in assessment of treatment response. To be of use, the imaging method has to be able to differentiate between the highly vascular mass and the infiltrative cancer cell pool in tumors. One of the most commonly used clinical MR methods for diagnosing brain tumors is based on imaging perfusion in the brain with the help of intravenous contrast agents. The increased vascularization of the angiogenic brain tumors and the leakiness of the tumor vessels increase the concentration of the contrast agent, and consequently, the MRI signal intensity in the tumor tissue as compared to the normal brain tissue increases. However, this MRI method is not able to detect the infiltrative tumor cell pool which is non-enhancing on perfusion MRI. To that purpose, T2-weighted images in combination with fluid suppression (T2-weighted FLAIR) are commonly used. The infiltrative tumors show as hyperintense regions on T2-weighted images, while fluid suppression is used to differentiate them from intracerebral edema. The lack of single reliable method for brain tumor assessment makes the diagnosis harder and necessitates the use of biopsies.
In the recent year, diffusion weighted MRI (DW MRI) has been gaining recognition in the clinical MRI setting. DW MRI is based on the diffusion of water protons which in turn depends on the geometry of the medium. Cellular membranes, organelles and other tissue structures present barriers to diffusion, thus reducing diffusion coefficient compared to the unrestricted diffusion in CBF. Since tumor microstructure is very different from the normal tissue, DW MRI shows great potential for tumor imaging. However, research showed that currently used DW MRI methods which are based on pulsed gradient spin echo techniques lack the sensitivity necessary to detect, for instance, early stages of glioblastomas, possibly because the mass density of the tumor is similar to that of the normal tissue.
The OGSE DW MRI method, on the other hand, is able to generate MR contrast based on the difference in tissue structure on a cellular and sub-cellular level. In addition to being a sensitive technique to detect early stages of cancer and response to treatment, it is also a good candidate for differentiating between infiltrative and angiogenic tumors, which will be the focus of this project.
I have started the process of implementing the OGSE DW MRI method on the preclinical 7T MRI system at the University of Bergen. The OGSE method employs a set of oscillating diffusion gradients instead of commonly used rectangular gradients and these have to be programmed into the sequence and tested on phantoms. Furthermore, I am working on optimizing the conventional DW MRI method in order to eventually compare the capabilities of the two methods. Finally, I am exploring several software packages in order to determine the best analysis tool for processing of DW MR images.
Novel MRI method for characterization of glioblastomas.
The main objective of the project is to develop new MRI methodology for better characterization and diagnosis of glioblastomas, as well as for evaluating response to anti-angiogenic therapy. In particular, we would like to explore whether this new MRI technique has the potential to detect invasive, as opposed to angiogenic, tumor cell pools.
We propose to implement a novel MR diffusion method called oscillating-gradient spin-echo (OGSE) technique on a preclinical small-animal 7 Tesla MR scanner (located at Vivarium, Department of Biomedicine, University of Bergen). We hypothesize that the method will enable us to characterize tumor microstructure on a subcellular level, giving us therefore the possibility to differentiate between invasive and angiogenic tumor cell pools, and to better visualize response to anti-angiogenic treatment.
Clarification: I have worked on this post doctoral project in a 50% position from Aug 1st 2010 till October 20th 2010 when I started my maternity leave. During this time I have:
a) Researched literature to obtain information on the current status of diffusion MRI of cancer in animal models and on diffusion models of tissue microstructure.
b) Attended a three-day course on Diffusion MRI in Oxford, England, which was sponsored by the European Society for Magnetic Resonance in Medicine and Biology, and which hosted carefully selected experts in the field of diffusion MRI.
c) I began to test and optimize the currently available diffusion method (PGSE) to image brains of healthy animals as a starting point. One of the objectives of this project is to make a detailed comparison of the MR imaging contrast obtained with the conventional (PGSE) and the novel (OGSE) pulse sequences in animal models of gliolastomas, so the optimization of the PGSE technique is essential to the project objectives.
d) Had several communication exchanges with my collaborators in Slovenia and Department of Chemistry in Bergen regarding theoretical models of diffusion MR and the characterization of microstructure using diffusion MRI.
e) Communicated with the manufacturer (Bruker Biospin, Germany) of the 7T small animal MRI system at the Molecular Imaging Center in Bergen regarding the implementation of the novel diffusion method (OGSE) on our system.
f) Submitted and got approved an application to FOTS (No 2340, Testing of MRI protocols) for testing MRI protocols on healthy animals.
Loss or Mislocalization of Aquaporin-4 Affects Diffusion Properties and Intermediary Metabolism in Gray Matter of Mice.
Neurochem Res 2016 Dec 30. Epub 2016 des 30
Increased microvascular permeability in mice lacking Epac1 (Rapgef3).
Acta Physiol (Oxf) 2016 Apr 20. Epub 2016 apr 20
Multimodal Imaging of Orthotopic Mouse Model of Endometrial Carcinoma.
PLoS One 2015;10(8):e0135220. Epub 2015 aug 7
MRI of the central nervous system in rats following heliox saturation decompression.
Undersea Hyperb Med 2015 Jan-Feb;42(1):57-64.
Multimodal imaging of gliomas in the context of evolving cellular and molecular therapies.
Adv Drug Deliv Rev 2014 Sep 30;76():98-115. Epub 2014 jul 28
Image-based assessment of microvascular function and structure in collagen XV- and XVIII-deficient mice.
J Physiol 2014 Jan 15;592(Pt 2):325-36. Epub 2013 nov 11
Deep gray matter demyelination detected by magnetization transfer ratio in the cuprizone model.
PLoS One 2013;8(12):e84162. Epub 2013 des 30
Nitroreductase, a near-infrared reporter platform for in vivo time-domain optical imaging of metastatic cancer.
Cancer Res 2013 Feb 15;73(4):1276-86. Epub 2012 des 10
Using Single-Channel Blind Deconvolution to Choose the Most Realistic Pharmacokinetic Model in DCE MR Imaging
Appl Magn Reson 2015