Alonso Garcia-Ruiz¹, Albert Pons-Escoda², Francesco Grussu¹, Pablo Naval-Baudin², Camilo Monreal¹, Antonio Lopez-Rueda³, Laura Oleaga³, Carlos Majos², and Raquel Perez-Lopez^1
1Vall d'Hebron Institute of Oncology, Barcelona, Spain, ²Bellvitge University Hospital, L'Hospitalet de Llobregat, Spain, ³Hospital Clínic de Barcelona, Barcelona, Spain

The robustness and accuracy of current MR methods to differentiate brain tumours is limited. In this study we investigate the potential of dynamic susceptibility perfusion-weighted imaging (DSC-PWI) normalized time-intensity-curves (nTIC) to support lymphoma diagnosis by harnessing voxelwise and temporal information to train a convolutional neural network (CNN). This novel approach discriminated patients with lymphoma from glioblastoma and metastasis with an average accuracy of 0.94, using only a limited number of patients for training, outperforming standard DSC-PWI measurements. Furthermore, it provides voxel-by-voxel lymphoma probability maps to further help visual diagnosis of neuroradiologists.

Martin Kozár^1,2, Sarah Al-Bachari³, Geoff J. M. Parker^4,5, Laura M. Parkes^1,2, and Ben R. Dickie^2,6
1Division of Neuroscience and Experimental Psychology, The University of Manchester, Manchester, United Kingdom, ²Geoffrey Jefferson Brain Research Centre, The University of Manchester, Manchester, United Kingdom, ³Faculty of Health and Medicine, The University of Lancaster, Lancaster, United Kingdom, ⁴Centre for Medical Image Computing, Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, ⁵Bioxydyn Limited, Manchester, United Kingdom, ⁶Division of Informatics, Imaging and Data Science, The University of Manchester, Manchester, United Kingdom

The Patlak model is commonly applied to brain DCE-MRI to quantify subtle blood-brain barrier permeability. This is based on existing evidence showing the Patlak model provides the optimal fit compared to other candidate models, including the extended Tofts model (Heye, Thrippleton et al. 2016). In this study, we compared five different tracer kinetic models in data with higher temporal resolution and determined that models which assume finite interstitial volume (v_e) (e.g. extended Tofts) consistently outperform models that assume infinite v_e (e.g. Patlak), particularly in gray matter.

Elles Elschot^1,2, Walter Backes^1,2,3, Joost de Jong^1,2, Gerhard Drenthen^1,2, Sau May Wong¹, Julie Staals^3,4, Robert van Oostenbrugge^2,3,4, Rob Rouhl⁴, and Jacobus Jansen^1,2,5
1Radiology and Nuclear Medicine, Maastricht University Medical Center +, Maastricht, Netherlands, ²School for Mental Health and Neuroscience, Maastricht University, Maastricht, Netherlands, ³CARIM School for Cardiovascular Diseases, Maastricht University, Maastricht, Netherlands, ⁴Neurology, Maastricht University Medical Center +, Maastricht, Netherlands, ⁵Electrical Engineering, Eindhoven University of Technology, Eindhoven, Netherlands

This study investigated to what extent blood-brain barrier (BBB) leakage (1) can be measured with DSC-MRI and (2) is influenced by perfusion. In vivo DCE (golden-standard) and DSC data of patients with cerebral small vessel disease (cSVD) and elderly controls were used, as well as simulations of signal curves. DSC-MRI, in contrast to DCE-MRI, is not sensitive enough to measure subtle leakage in cSVD. Further research is required to better disentangle perfusion effects from leakage, and therefore correction methods should be used with caution for measuring subtle leakage.
 

Radovan Jirik¹, Michal Bartos², Ondrej Macicek¹, and Zenon Starcuk, jr.^1
1Institute of Scientific Instruments, Czech Academy of Sciences, Brno, Czech Republic, ²Institute of Information Theory and Automation, Czech Academy of Sciences, Praha, Czech Republic

This work is focused on estimation of an arterial input function (AIF) in DCE-MRI. We propose a new concept of blind deconvolution to use as much available information as possible. In contrast to multi-channel blind deconvolution based on processing of selected tissue signals, our all-channel blind deconvolution uses tissue signals of all voxels within the studied tissue to simultaneously estimate the AIF and perfusion maps. In addition, edge-preserving spatial regularization is included in the perfusion-map estimation scheme. The method is illustrated on DCE-MRI recordings of tumor-bearing mice.

Ondrej Macicek¹, Radovan Jirik¹, Peter Latta², and Zenon Starcuk jr.^1
1Institute of Scientific Instruments of the Czech Academy of Sciences, Brno, Czech Republic, ²Central European Institute of Technology, Masaryk University, Brno, Czech Republic

We propose to combine DCE and DSC-MRI in a simultaneous acquisition and processing scheme to provide more information for estimation of the arterial input function using multi-channel blind deconvolution. On simulated and flow-phantom data, we show that DCE-DSC blind deconvolution is superior to standard DCE blind deconvolution, especially for low signal-to-noise-ratio conditions.

Chih-Hsien Tseng¹, Jaap Jaspers², Alejandra Mendez Romero², Piotr Wielopolski³, Matthias van Osch ⁴, and Frans Vos^1,3
1Imaging Physics, Delft University of Technology, Delft, Netherlands, ²Radiation Oncology, Erasmus Medical Center, Rotterdam, Netherlands, ³Radiology, Erasmus Medical Center, Rotterdam, Netherlands, ⁴Radiology, Leiden University Medical Center, Leiden, Netherlands

We aimed to compare the AIF determined from DCE-MRI with the AIF from DSC-MRI in back-to-back imaging, using the same dose of Gadobutrol. The DCE-driven AIFs showed the same ‘pattern’ as the DSC-driven AIFs, but had smaller variance of the shape and peak value. A linear relation was found between the contrast agent concentration from DCE and ΔR₂*, but the linear coefficients showed large variation across regions and subjects. Our findings show that the DCE-driven AIF has the potential to serve as an alternative for the DSC-driven AIFs for quantitative perfusion assessment.

Zaki Ahmed^1,2 and Ives R. Levesque^2,3
1Department of Radiology, Mayo Clinic, Rochester, MN, United States, ²Medical Physics Unit, McGill University, Montreal, QC, Canada, ³Research Institute of the McGill University Health Centre, Montreal, QC, Canada

Quantitative DCE-MRI analysis requires fitting a model to the acquired enhancing time-course data. These models involve a convolution operation, which has a variety of possible implementations. The goal of this work was to evaluate three different implementations of the convolution operation: (i) Fourier, (ii) Summation, (iii) Iterative. The accuracy and execution time of each implementation was compared on a virtual phantom. It was found that the iterative technique was the fastest while also having the best overall accuracy.

Owen Alun White¹, Evanthia Kousi¹, Clarrissa Sanders², and Maria Angelica Schmidt^1
1Radioisotope & Imaging Physics, The Royal Marsden NHS Foundation Trust, London, United Kingdom, ²St George's University Hospital NHS Foundation Trust, London, United Kingdom

Here we propose a novel method to assess the effect of different k-space sampling patterns on contrast agent uptake curves (CAUCs), which does not require specialist equipment. These effects on CAUCs are often overlooked, and must be addressed in the context of the many different national and international recommendations, and the prominence given to abbreviated DCE protocols. We validated the method to assess different k-space sampling patterns and used it to simulate the effect on contrast-agent uptake curves. We show that conventional protocols acquiring each frame in approximately 60s have more robust CAUC classification than abbreviated protocols.

Muhammad Asaduddin¹, Hong Gee Roh², Hyun Jeong Kim³, Eung Yeop Kim⁴, and Sung-Hong Park^1
1Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of, ²Department of Radiology, Konkuk University Medical Center, Seoul, Korea, Republic of, ³Department of Radiology, Daejeon St. Mary's Hospital, The catholic University of Korea, Daejeon, Korea, Republic of, ⁴Department of Radiology, Gil Medical Center, Gachon University College of Medicine, Incheon, Korea, Republic of

Perfusion maps and dynamic angiograms are both important for stroke/tumor treatment but commonly acquired in separate scans and thus may require additional injection of contrast agent for best results. In this work, we present a deep learning method to acquire perfusion maps from contrast-enhanced MRA data. Our results showed that an architecture of multiple decoders and an enhanced encoder produced perfusion maps that were visually and quantitatively similar to the standard DSC MRI-based perfusion maps. This approach enables us to acquire accurate perfusion maps and angiogram using a single contrast agent injection, reducing costs and risks while improving patient comfort.