Juan Pablo Meneses^1,2,3, Cristobal Arrieta^1,2, Gabriel della Maggiora^1,2, Pablo Irarrazaval^1,2,3,4, Cristian Tejos^1,2,3, Marcelo Andia^1,2,5, Sergio Uribe^1,2,5, and Carlos Sing Long^1,2,4,6,7
1Biomedical Imaging Center, Pontificia Universidad Católica de Chile, Santiago, Chile, ²ANID – Millennium Science Initiative Program – Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, ³Electrical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, ⁴Institute for Biological and Medical Engineering, Pontificia Universidad Catolica de Chile, Santiago, Chile, ⁵Radiology Department, School of Medicine, Pontificia Universidad Catolica de Chile, Santiago, Chile, ⁶Institute for Mathematical & Computational Engineering, Pontificia Universidad Catolica de Chile, Santiago, Chile, ⁷ANID – Millennium Science Initiative Program – Millennium Nucleus for Discovery of Structure in Complex Data, Santiago, Chile

MRI water-fat separation is an ill-posed inverse problem usually addressed using complex numerical methods. This problem is related to an MR signal physical model that includes the effects of R₂^* signal decay ratio and main magnetic field inhomogeneities (Δf), which induce non-linearities that are related to the ill-posedness of the inverse problem. In this work, we propose an Optimal Transport driven Cycle-consistent Generative Adversarial Networks (OT-CycleGAN) framework, which is physics-based, and could use partially labeled training data to estimate the non-linear components (R₂^* and Δf), and posteriorly compute the water-fat images through a conventional least-squares approach.

Jayse M Weaver^1,2, Nathan T Roberts^2,3, Diego Hernando^1,2, and Scott B Reeder^1,2,4,5,6
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ²Radiology, University of Wisconsin-Madison, Madison, WI, United States, ³Electrical Engineering, University of Wisconsin-Madison, Madison, WI, United States, ⁴Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States, ⁵Medicine, University of Wisconsin-Madison, Madison, WI, United States, ⁶Emergency Medicine, University of Wisconsin-Madison, Madison, WI, United States

Chemical-shift-encoded MRI (CSE-MRI) is an established technique to estimate proton density fat-fraction (PDFF) as a quantitative imaging biomarker of hepatic steatosis. However, due in part to low flip angle acquisitions, CSE-MRI acquisitions often have low SNR leading to biased and noisy PDFF measurements. Unreliable PDFF measurements are particularly problematic considering clinically relevant PDFF thresholds may be as low as 3-5%. There are currently no methods to gauge the reliability of PDFF measurements. We propose a method for generating color-coded CSE-MRI estimation confidence maps, using a normalized residual metric to determine PDFF measurement reliability.

Mark Zamskiy¹, Kilian Weiss², Felix N. Harder¹, Stefan Ruschke¹, Marcus R. Makowski¹, Rickmer F. Braren¹, and Dimitrios C. Karampinos^1
1Department of Diagnostic and Interventional Radiology, School of Medicine, Technical University of Munich, Munich, Germany, ²Philips GmbH Market DACH, Hamburg, Germany

Volumetric liver T2-mapping is of interest in the characterization of focal lesions and diffuse disease but remains technically challenging due to respiratory motion. The present work introduces a novel T2-prepared (T2prep) radial stack-of-stars two-point Dixon gradient echo sequence (T2prep-SoS-Dixon-TFE) for achieving motion-robust 3D isotropic T2-mapping in navigator-gated scans using a B0/B1-insusceptible modified adiabatic BIR-4 RF-pulse for T2prep. T2-mapping is performed via dictionary matching to Bloch simulations for the Dixon-decomposed water signal. The proposed method enables motion-robust liver fat confounder-free volumetric T2 quantification in better agreement with MRS compared to GraSE and may improve the clinical applicability of free-breathing abdominal T2-mapping.

Mary Kate Manhard^1,2, Amol S. Pednekar^1,2, Hui Wang^1,2,3, Jean A. Tkach^1,2, and Jonathan R. Dillman^1,2
1Department of Radiology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States, ²Department of Radiology, University of Cincinnati College of Medicine, Cincinnati, OH, United States, ³Clinical Science, Philips Healthcare, Cincinnati, OH, United States

Parametric mapping aids in the non-invasive evaluation and longitudinal monitoring of diffuse chronic liver disease. Commercially available parametric mapping techniques require multiple scans over multiple breath-holds. A multi-inversion spin and gradient echo (MI-SAGE) method is presented to simultaneously estimate T₁, T₂, and T₂^* at six slice locations in the liver in a single 16s breath-hold. In five volunteers, the MI-SAGE method produced parametric values comparable to those obtained using modified Lock Locker (T₁), multiple gradient and spin echo (T₂), and multiple gradient echo (T₂^*) sequences. MI-SAGE offers the potential to significantly shorten the duration of the quantitative MR liver exams.

Ty A Cashen¹, Sangtae Ahn², Uri Wollner³, Graeme McKinnon⁴, Isabelle Heukensfeldt Jansen², Rafi Brada³, Dan Rettmann⁵, Xucheng Zhu⁶, and Ersin Bayram^7
1GE Healthcare, Madison, WI, United States, ²GE Research, Niskayuna, NY, United States, ³GE Research, Herzliya, Israel, ⁴GE Healthcare, Waukesha, WI, United States, ⁵GE Healthcare, Rochester, MN, United States, ⁶GE Healthcare, Menlo Park, CA, United States, ⁷GE Healthcare, Houston, TX, United States

3D T1-weighted gradient echo imaging is a key component of the MRI assessment of the abdomen, particularly for the identification and characterization of liver tumors, however, significant acceleration is necessary to consistently mitigate respiratory motion artifact. A variable density Poisson disc undersampled acquisition with a densely connected iterative deep convolutional neural network reconstruction was developed to provide next-generation acceleration up to a factor of 10. On retrospectively undersampled data, the technique outperformed compressed sensing reconstruction in terms of normalized mean-squared error and structural similarity; with a prospectively undersampled scan, the technique maintained image quality in terms of artifact and contrast.

David Stockton¹, Mihaela Rata^1,2, Francesca Castagnoli¹, Joshua Shur¹, Georgina Hopkinson¹, Alison Macdonald¹, Marcel Dominik Nickel³, Stephan Kannengiesser³, Dow-Mu Koh^1,2, and Jessica M Winfield^1,2
1MRI Unit, The Royal Marsden NHS Foundation Trust, London, United Kingdom, ²Division of Radiotherapy and Imaging, The Institute of Cancer Research, London, United Kingdom, ³MR Application Predevelopment, Siemens Healthcare GmbH, Erlangen, Germany

An advanced reconstruction method, which employs a neural network trained on high resolution data to conduct interpolation, combined with iterative denoising, was applied in contrast-enhanced liver imaging in patients receiving a hepatocyte-specific contrast agent. The advanced reconstruction provides improved image quality, with better contrast-to-noise ratio, vessel edge sharpness, and lesion edge sharpness compared with the standard reconstruction using zero-filling interpolation without advanced reconstruction. The improvement in image quality was also observed in faster acquisitions (lower phase resolution), thus enabling shorter breath-hold acquisitions. 

Dragana Savic¹, Ferenc E. Mózes¹, Peregrine G. Green¹, Matthew K. Burrage¹, Gabrielle Allen², Timothy James², Leanne Hodson¹, Stefan Neubauer¹, Michael Pavlides¹, and Ladislav Valkovič^1
1University of Oxford, Oxford, United Kingdom, ²Oxford University Hospitals, Oxford, United Kingdom

This study investigated the in vivo effects on acetylcarnitine levels in the liver after an injection of L-carnitine. L-carnitine facilitates transport of fatty acids into the mitochondria. We show that a single injection of L-carnitine modulated acetylcarnitine levels in the liver and that it modulated several blood markers in the healthy volunteer. Further studies are needed to understand the plasma metabolic changes observed with L-carnitine supplementation.

Lieke van den Wildenberg¹, Ayhan Gursan¹, Leonard W.F. Seelen¹, Tijl A. van der Velden¹, Mark Gosselink¹, Martijn Froeling¹, Wybe J.M. van der Kemp¹, Dennis Klomp¹, and Jeanine Prompers^1
1Imaging and Oncology, UMC Utrecht, Utrecht, Netherlands

Alterations in ³¹P metabolite concentrations in the liver can be a consequence of liver disease and therefore 3D ³¹P MRSI can help in diagnosis and following progression and treatment response. However, commonly used ³¹P surface coils have a limited penetration depth and do not provide full coverage of the liver. We used an integrated ³¹P whole-body transmit coil with a local ³¹P receive array for ³¹P MRSI of the liver at a 7T MRI scanner. Here we demonstrate that ³¹P signals can be measured throughout the liver and ³¹P metabolite levels can be quantified with good to excellent test-retest reliability.

Stephen Bawden^1,2, Bernard Lanz³, Peter E Thelwall⁴, Guruprasad P Aithal², and Penny Gowland^1
1Sir Peter Mansfield Imaging Centre, School of Physics and Astronomy, University of Nottingham, Nottingham, United Kingdom, ²NIHR Nottingham Biomedical Research Centre, Nottingham University hospitals NHS trust and the University of Nottingham, University of Nottingham, Nottingham, United Kingdom, ³Sir Peter Mansfield Imaging Centre, School of Medicine, University of Nottingham, Nottingham, United Kingdom, ⁴Translational and Clinical Research Institute, Newcastle University, Newcastle, United Kingdom

Oxidative stress plays a central role in the development of both acute and chronic liver injury, with glutathione being the primary anti-oxidant. This study develops a method of measuring glycine to glutathione flux in vivo using 13C MRS and a novel labelling strategy.  Quantification of [2-13C] glycine and [2-13C] glutathione concentrations following glycine ingestion show that the protocol used in this study does successfully enrich the glutathione pool and confirm the assumptions of the metabolic model proposed.  This provides a powerful methodology to investigate oxidative stress in patient populations and in response to drug intervention.

Yaewon Kim¹, Cornelius von Morze², Jeremy W. Gordon¹, Peder E. Z. Larson¹, and Michael A. Ohliger^1
1Department of Radiology and Biomedical Imaging, University of California, San Francisco, CA, United States, ²Mallinckrodt Institute of Radiology, Washington University, St. Louis, MO, United States

 Hyperpolarized ¹³C MRS is a powerful emerging technique for assessing metabolic liver diseases, including diabetes and fatty liver. Chemical shift separation allows monitoring of many metabolites at once. Here, we explore the possibility of co-polarizing [1-¹³C]pyruvate and [2-¹³C]dihydroxyacetone to simultaneously assess two important metabolic pathways that are altered in liver disease. Co-polarization was successful, with preserved T₁ relaxation times. Although co-polarization led to modest decreases in polarization levels, all expected metabolites were observed when injected into a healthy rat. By examining two pathways at the same time, co-polarization can be an important tool for assessing liver disease.

 

Abi Spicer¹, Susan Francis¹, Penny Anne Gowland¹, and Stephen Bawden^1,2
1Physics, University of Nottingham, Nottingham, United Kingdom, ²NIHR Nottingham Biomedical Research Centre, Nottingham University hospitals NHS trust, Nottingham, United Kingdom

13C Magnetic Resonance Spectroscopy (MRS) in the liver provides the only non-invasive method of studying metabolites in vivo, with standard protocols using multiple averages during a free breathing period. Using a respiratory belt to monitor the breathing cycle and in-house scripts, spectra were binned into inhale and exhale groups categorised as spectra within 2 standard deviations of the average minima/maxima. A systematic reduction in B0 during exhale compared with inhale but minimal impact on average B1 values during an inhale versus exhale. However, no significant changes in Phase angle, area under the curve or FWHM between the data sets.

Matthew Gibbons¹, Edgar Castellanos Diaz¹, Suneil K Koliwad², Peter W Hunt², Jean-Marc Schwarz^2,3, Kathleen Mulligan^2,3, Robert H Lustig⁴, Alejandro Gugliucci³, Diana L Alba², Ayca Erkin-Cakmak⁴, and Susan M Noworolski^1
1Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States, ²Department of Medicine, University of California San Francisco, San Francisco, CA, United States, ³College of Osteopathic Medicine, Touro University - California, Vallejo, CA, United States, ⁴Department of Pediatrics, University of California San Francisco, San Francisco, CA, United States

The objective of this study was to generate an automatic liver segmentation method. Two methods were compared. The first, M1, was a Convolutional Neural Network (CNN) trained on proton density fat fraction (PDFF) maps. The second, M2, was the CNN trained on multiparametric MRI (mpMRI) images combined with an error detection protocol. The distributions for Dice similarity coefficient (DSC), volume, and PDFF were improved for M2 versus M1. The DSC mean increased from 0.91 to 0.96. The M2 method was effective in detecting and correcting poor segmentations while significantly reducing processing time as compared to manual segmentation.

Annemarie Knill^1,2, Antonio Candito¹, Jessica Winfield^1,2, James Larkin^1,2, Samra Turajlic^2,3, Dow Mu Koh^1,2, Christina Messiou^1,2, and Matthew Blackledge^1
1The Institute of Cancer Research, London, United Kingdom, ²The Royal Marsden NHS Foundation Trust, London, United Kingdom, ³The Francis Crick Institute, London, United Kingdom

We present a proof-of-concept study to assess whether deformable registration followed by tissue classification using machine learning (ML) is an effective method for the delineation of liver metastases in whole-body diffusion-weighted imaging (WB-DWI). Deformable atlas-based registration achieves good quality delineation of the liver (Dice coefficient > 70%) and out of three ML models random forest achieved the best F-1 measure for segmenting disease within the liver.