Perfusion & Permeability: Contrast Agent Methods

Wednesday 14 May 2014

Mikael F. Forsgren^1,2, Jose L. Ulloa³, and Paul D. Hockings^4,5
1Wolfram MathCore AB, Linköping, Sweden, and Center for Medical Image Science and Visualization(CMIV), Linköping University, Linköping, Sweden,²Dept. of Medical and Health Sciences, Linköping University, Dept. of Radiation Physics, UHL County Council of Östergötland, Linköping, Sweden,³Bioxydyn Ltd., Manchester, United Kingdom, ⁴Drug Safety and Metabolism, AstraZeneca AB, Mölndal, Sweden, ⁵MedTech West, Chalmers University of Technology, Gothenburg, Sweden

Dynamic contrast-enhanced MRI seems promising for non-invasive quantification of liver function. Recently a model based method for the quantification was published, and herein we tested if the number of parameters could be reduced by linearizing the efflux into the bile (MRP2), and still maintain sufficient separation between normal and reduced liver function. This model reduction was tested on rats treated with a chemokine receptor antagonist that reduces liver function. We found that the uptake of contrast was unaffected by the model reduction of the efflux, and that the reduced model was significantly able to separate between normal and reduced function.

E.G.W. ter Voert¹, L. Heijmen², C.J.A. Punt³, J.H.W. de Wilt⁴, H.W.M. van Laarhoven^2,3, and A. Heerschap^1
1Radiology, Radboud University Medical Centre, Nijmegen, Gelderland, Netherlands, ²Medical Oncology, Radboud University Medical Centre, Nijmegen, Gelderland, Netherlands, ³Medical Oncology, Academic Medical Center, University of Amsterdam, Amsterdam, Noord-Holland, Netherlands, ⁴Surgery, Radboud University Medical Centre, Nijmegen, Gelderland, Netherlands

Reproducibility of dynamic contrast enhanced MRI (DCE-MRI) in tumors in the liver is hampered by respiratory and cardiac motion. The aim of this study was to implement and apply a structural similarity algorithm (SSIM) in the post-processing of high-temporal-resolution 2D DCE-MRI data to correct for movement artifacts. The implemented SSIM algorithm compares the images from the dynamic series with reference images and based on the returned index values it rejects motion corrupted time points. The application of SSIM was tested in 15 patients and it substantially improved the reproducibility of the DCE-MRI pharmacokinetic modeling (Tofts model) parameter K^trans in the liver.

Jin Zhang¹, Melanie Freed¹, Kerryanne Winters¹, and Sungheon Kim^1
1Radiology, New York University, New York, New York, United States

The influence of T₂^* on lesion signal intensity and arterial input function at high field can induce systemic errors in pharmacokinetic parameter estimation. When the arterial input function is estimated from reference tissue, the T₂^* effect on parameter estimation may be relatively smaller than when it is directly measured. In this study, we investigate the influence of a Gd-based contrast agent on the T₂^* of tumor at 7T and its effect on kinetic model analysis. Our preliminary results show no significant difference between pharmacokinetic model parameters from T₂^*-corrected and non-corrected data when a reference tissue arterial input function is used.

Elias Kellner¹, Irina Mader², and Valerij G Kiselev^1
1Medical Physics, University Medical Cener Freiburg, Freiburg, Germany, ²Section of Neuroradiology, Neurocenter, University Medical Center Freiburg, Freiburg, Germany

In this work, we examine the feasibility of measuring quantitative phase-based arterial input functions and venous output functions in an additional slice at the carotid arteries and examine the optimal location for the measurement.

Jen Moroz¹, Piotr Kozlowski^2,3, and Stefan A Reinsberg^1
1Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, ²Radiology, UBC, University of British Columbia, BC, Canada, ³UBC MRI Research Centre, University of British Columbia, BC, Canada

The arterial input function is an important input parameter for modelling DCE-MRI data. However, it is difficult to measure accurately in mice. A projection-based method can significantly improve the temporal resolution of the AIF, but may be biased due to local tissue enhancement. This simulation study extends the Cartesian projection-based method to a radial approach: Each projection serves as a single time-point measure of the AIF, but is also used to reconstruct a radial image to assess background tissue enhancement. The results show that an AIF, measured from single projections, can be corrected from a time-series of radial images to remove the bias caused from local tissue enhancement.

Ellis Beld¹, Frank F.J. Simonis¹, Johannes G. Korporaal¹, Uulke A. van der Heide², and Cornelis A.T. van den Berg^1
1Radiotherapy, UMC Utrecht, Utrecht, Netherlands, ²Radiotherapy, Netherlands Cancer Institute - Antoni van Leeuwenhoek Hospital (NKI-AVL), Amsterdam, Netherlands

Phase-sensitive MR procedures are affected by spatially and temporally varying B0 inhomogeneities. A higher order background phase correction, to compensate for these inhomogeneities, was used for phase-based measurement of the arterial input function (AIF) in dynamic contrast-enhanced (DCE) MRI. The background phase was calculated inwards from the subcutaneous fat layer, using a near-harmonic 2D reconstruction. Besides measurement of the AIF, the background corrected phase data enables detection of contrast induced phase modulation in tissues.

Patrik Brynolfsson¹ and Anders Garpebring^1,2
1Radiation Sciences, Umeå University, Umeå, Sweden, ²CJ Gorter Center for high field MRI, Leiden University Medical Center, Leiden, Zuid-Holland, Netherlands

A novel method for contrast agent (CA) quantification is proposed. Phase changes linearly with CA concentration, but is hard to use due to the ill-posed inversion from phase-change to CA concentration. We regularize the inversion with magnitude data in a statistical approach and evaluate the result on simulated DCE-MRI exams. For tissues with high CA concentration the novel method performed better than using magnitude information only, for low CA concentrations the methods were equivalent. The next step is to test the method on in-vivo data where many challenges still remain, such as phase-drift.

Julien Bouvier¹, Nicolas Coquery¹, Sylvie Grand², Thomas Perret¹, David Chechin³, Irene Tropres¹, Alexandre Krainik¹, and Emmanuel L Barbier^1
1U836, INSERM, Grenoble, France, France, ²Department of neuroradiology and MRI, CHU de Grenoble, France, France, ³Philips Healthcare, Suresnes, France, France

In clinical monitoring of brain tumors, Perfusion Weighted Imaging (PWI) contributes to tumor grading and to assess the response to treatment. Beyond tumor perfusion, tumor hypoxia determines the response of various therapeutic approaches including radiotherapy. All these parameters may be mapped with MRI. However, the integration of several MRI maps is difficult. This wealth of information is however difficult to interpret. Moreover, there are tight physiological links between these parameters. It should thus be possible to define clusters of pixels with similar physiological characteristics. In this study, multiparametric MRI data collected on tumor patient were analyzed with a model-based cluster approach.

Benjamin Lemasson^1,2, Nicolas Pannetier^3,4, Régine Farion^1,2, Emmanuel Barbier^1,2, Norbert Schuff^3,4, Michael Moseley⁵, Greg Zaharchuk⁵, and Thomas Christen^5
1U836, Iserm, Grenoble, France, ²Grenoble Institut des Neurosciences, Université Joseph Fourier, Grenoble, France, ³Va medical center, Centre for neurodegenerative des eases, San Francisco, CA, United States, ⁴University of California San Francisco, Department of Radiology and Biomedical Imaging, San Francisco, CA, United States, ⁵Stanford University, Department of Radiology, Stanford, CA, United States

In this study, we tested the ‘vascular fingerprinting’ approach in 8 rats bearing brain tumors. This recent method compares in vivo MR signal time evolutions to a dictionary of curves obtained with numerical simulations and creates maps of microvascular characteristics. We obtained good correlations with more conventional MR methods: steady-state susceptibility contrast imaging for blood volume mapping, Vessel Size Imaging and multiparametric quantitative BOLD imaging for blood oxygen saturation measurements. In two rats, high spatial resolution maps were obtained and compared to pimonidazole staining, a histological marker of tissue hypoxia.

Ashley M Stokes^1,2, Natenael Semmineh^1,3, and C. Chad Quarles^1,2
1Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States, ²Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, United States, ³Physics and Astronomy, Vanderbilt University, Nashville, TN, United States

Contrast agent (CA) leakage in tumors can lead to incorrect hemodynamic measures in DSC-MRI; therefore, DSC R₂^* signals must be leakage corrected to provide accurate CBF and CBV. We propose removing T₂^* leakage effects in T₁-insensitive dual-echo R₂^* by estimating the extravascular CA relaxivity using the transverse relaxivity at tracer equilibrium (TRATE). In rat glioma, we compared the uncorrected CBV and CBF to corrected values using the Weisskoff and TRATE corrections; these were further compared to MION reference values. The Weisskoff and TRATE corrections led to CBVs that were closer to the MION values, but both underestimated CBF.