Rencheng Zheng¹, Qidong Wang², Ziying Feng³, Chengyan Wang³, and He Wang^1,3
1Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China, ²Radiology department, The first affiliated hospital, College of medicine, Zhejiang University, Hangzhou, China, ³Human Phenome Institute, Fudan University, Shanghai, China

The objective of this study is to perform automatic multi DCE phases liver tumor detection using deep learning based segmentation and radiomics guided candidate filtering. The proposed model consists mainly of two stages, primary segmentation based on a U-net architecture neural network in stage1, and suspected tumor discrimination mechanism using multi DCE phases radiomics features including shape features, texture features, time dimension features and location information in stage 2. The proposed two-stage model exhibits superior performance in HCC tumor segmentation with a mean Dice score of 0.7928 in test set.

Wu Zhou¹, Wanwei Jian¹, and Guangyi Wang^2
1School of Medical Information Engineering, Guangzhou University of Chinese Medicine, Guangzhou, China, ²Department of Radiology, Guangdong General Hospital, Guangzhou, China

Contrast agent has several limitations in clinical practice, and the diagnostic performance of non-enhanced MR for lesion characterization should be thoroughly exploited. Inspired by the work of cross-modal learning framework，we propose a deeply supervised cross-modal transfer learning method to remarkably improve the malignancy characterization of HCC in non-enhanced MR, in which the cross-modal relationship between the non-enhanced modal and contrast-enhanced modal is explicitly learned and subsequently transferred to another CNN model for improving the characterization performance of non-enhanced MR. The visualization method Grad-CAM is also applied to verify the effectiveness of the proposed cross-modal transfer learning model.

Stefanie Hectors^1,2,3, Paul Kennedy^1,2, Kuang-Han Huang^1,4, Hayit Greenspan⁵, Scott Friedman⁶, and Bachir Taouli^1,2
1BioMedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ²Department of Diagnostic, Molecular and Interventional Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ³Department of Radiology, Weill Cornell Medicine, New York, NY, United States, ⁴Prealize Health, Palo Alto, CA, United States, ⁵Medical Imaging Processing Lab, Tel Aviv University, Tel Aviv, Israel, ⁶Division of Liver Diseases, Icahn School of Medicine at Mount Sinai, New York, NY, United States

In this study we developed a fully automated deep learning algorithm based on gadoxetic acid-enhanced MRI for noninvasive prediction of liver fibrosis. We found good-to-excellent performance of the algorithm in an independent test set (AUC 0.77 – 0.91), which was equivalent to the diagnostic performance of MR elastography (AUC 0.86 – 0.92, p-values between methods >0.134). The developed algorithm may potentially allow for noninvasive liver fibrosis assessment, without the need for invasive biopsies. 

Jie Chen^1,2, Jun Chen¹, Jeremiah Heilman¹, Kevin Glaser¹, Richard Ehman¹, and Meng Yin^1
1Mayo Clinic, Rochester, MN, United States, ²West China Hospital, Sichuan University, Chengdu, China

Simultaneous stiffness assessment of multiple organs in patients with systemic diseases may be beneficial. Here we used four passive MRE drivers to assess the feasibility and repeatability of simultaneous stiffness measurements of the liver, spleen and kidneys using 60-Hz vibrations. No significant bias was observed in the stiffness measurements of the liver and kidneys with 1 versus multiple drivers. Splenic stiffness was lower with four drivers. With four drivers, it is possible to obtain simultaneous stiffness measurements of the liver, spleen and kidneys with repeatability coefficients similar to those using a single driver while also overcoming attenuation in the liver.

Thierry Lefebvre^1,2,3, Léonie Petitclerc^1,2,4, Giada Sebastiani⁵, Jeanne-Marie Giard^2,6, Marie-Pierre Sylvestre^2,7, Bich N. Nguyen⁸, Guillaume Gilbert^1,9, Guy Cloutier^1,10,11, and An Tang^1,2,10
1Department of Radiology, Radio-Oncology and Nuclear Medicine, Université de Montréal, Montreal, QC, Canada, ²Centre de recherche du Centre hospitalier de l'Université de Montréal (CRCHUM), Montréal, QC, Canada, ³Medical Physics Unit, McGill University, Montréal, QC, Canada, ⁴C.J. Gorter Center for High Field MRI, Department of Radiology, Leiden University Medical Center (LUMC), Leiden, Netherlands, ⁵Department of Medicine, Division of Gastroenterology and Hepatology, McGill University Health Centre (MUHC), Montréal, QC, Canada, ⁶Department of Medicine, Division of Hepatology and Liver Transplantation, Université de Montréal, Montréal, QC, Canada, ⁷Department of Social and Preventive Medicine, École de santé publique de l’Université de Montréal (ESPUM), Montréal, QC, Canada, ⁸Service of Pathology, Centre hospitalier de l'Université de Montréal (CHUM), Montréal, QC, Canada, ⁹MR Clinical Science, Philips Healthcare Canada, Markham, ON, Canada, ¹⁰Institute of Biomedical Engineering, Université de Montréal, Montréal, QC, Canada, ¹¹Laboratory of Biorheology and Medical Ultrasonics (LBUM), Centre de recherche du Centre hospitalier de l'Université de Montréal (CRCHUM), Montréal, QC, Canada

MR elastography for staging liver fibrosis assesses the right liver lobe and requires external hardware. MRI cine-tagging evaluates cardiac-induced strain and shows promise for assessing fibrosis in the left lobe without additional hardware. Shear modulus measured by MRE provided higher AUCs than that of strain measured on tagged images for distinguishing stages F0 vs. ≥F1 (0.87 vs. 0.81, P=0.083) and ≤F3 vs. F4 (0.91 vs. 0.87, P=0.043). Hence, MRE provided a diagnostic accuracy similar or higher than that of MRI cine-tagging for staging of liver fibrosis. Strain could be evaluated on screening abdominal MRI to assess the left liver.

Jingbiao Chen^1,2, Jin Wang³, Jiahui Li¹, Jie Chen¹, Xin Lu¹, Hiroaki Yashiro⁴, Jenifer Siegelman⁴, Christopher T Winkelmann⁴, Richard L Ehman¹, Vijay H Shah², and Meng Yin^1
1Radiology, Mayo Clinic, Rochester, MN, United States, ²Gastroenterology and Hepatology, Mayo Clinic, Rochester, MN, United States, ³the Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China, ⁴Research and Development, Takeda Pharmaceuticals International Co., Cambridge, MA, United States

Liver biopsy remains the gold standard for staging liver fibrosis. However, its invasive nature makes it unacceptable for long-term disease dynamic monitoring. In addition, current histopathological scoring systems for staging liver fibrosis are not quantitative. Also, the inflammatory response of increased interstitial fluid volume is precursory and occult with histological analysis. It is critical to address the overlooked fluid-associated inflammatory response and semi-quantitative fibrosis grading. Here, we use a novel technique, magnetic resonance elastography, combined with ALT, as a noninvasive quantitative method to quantify and monitor hepatic water and fibrosis content in a mouse model with varying disease progression/regression.

Ferenc E. Mozes¹, Emmanuel A. Selvaraj¹, Michael Pavlides^1,2, Matthew D. Robson^1,3, and Elizabeth M. Tunnicliffe^1
1OCMR, RDM Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom, ²Translational Gastroenterology Unit, University of Oxford, Oxford, United Kingdom, ³Perspectum Diagnostics Ltd., Oxford, United Kingdom

Non-alcoholic fatty liver disease is on the rise and liver biopsies used to diagnose it need to be replaced by non-invasive methods such as T₁ mapping. Guidance for MRI scans allow for free water consumption before scans, increasing measurement variability. We therefore aimed to assess the effect of hydration on liver T₁ by acquiring serial shMOLLI T₁ maps after participants drank 1 litre of isotonic water. As water passed through the stomach and small intestines it first reached the portal circulation and later the systemic circulation, causing an increase in liver T₁ followed by an increase in spinal muscle T₁.

Zhitao Li¹, John Pauly², and Shreyas Vasanawala^1
1Department of Radiology, Stanford University, Stanford, CA, United States, ²Electrical Engineering, Stanford University, Stanford, CA, United States

A radial dual-echo IR-SPGR technique combined with a principal component based iterative reconstruction algorithm are demonstrated for fat and water separated rapid high-resolution abdominal multi-parameter mapping. This method can yield high-quality fat-signal-fraction map, B₀ field map, composite T₁ map, water component T₁ map as well as a R₂^* map from a scan as fast as 3seconds/slice. The in-plane resolution of the resulting parameter maps is 1.25mm and the through-plane resolution is 5.00mm. With a selective inversion pulse, multiple slices can be acquired within a breath-hold.

Adam Michael Bush¹, Christopher Michael Sandino², Shreya Ramachandran³, Nicholas Dwork⁴, Frank Ong¹, Marcus Alley¹, and Shreyas Vasanawala^1
1Radiology, Stanford University, Palo Alto, CA, United States, ²Electrical Engineering, Stanford University, Palo Alto, CA, United States, ³Electrical Engineering, University of California Berkeley, Berkeley, CA, United States, ⁴Radiology, University of California San Francisco, San Francisco, CA, United States

 

 

In this work we introduce a novel method for T2* determination using rosette k-space sampling and locally low-rank reconstruction.  This approach produced comparable T2* quantitation with higher spatial resolution, fewer motion artifacts and lessened variability without the use of gating. This approach offers a child-friendly, rapid, free-breathing, comprehensive assessment of liver and cardiac iron.

Eddy Solomon¹, Thomas Vahle², Jan Paska¹, Kai Tobias Block¹, Daniel K. Sodickson¹, Fernando Boada¹, and Hersh Chandarana^1
1Radiology, New York University School of Medicine, New York, NY, United States, ²Siemens Healthcare GmbH, Erlangen, Germany

The sensitivity of MRI sequences to motion impairs their reliability and diagnostic utility for examining the chest and abdomen. Established motion-compensation techniques are not accurate enough, come at the cost of patient comfort, and are limited by the MR imaging parameters. Here, we demonstrate a novel approach that detects respiratory signal from the amplitude modulation of a transmitted RF reference signal, termed ‘pilot-tone’ (PT). We show how the use of this simple RF transmitter, with its small dimensions, high sampling rate, and low interference with the MR acquisition, can produce motion corrected-images under free-breathing conditions.