ISMRM 24th Annual Meeting & Exhibition • 07-13 May 2016 • Singapore

Scientific Session: Automating & Speeding Algorithms

Monday, May 9, 2016
Summit 1
16:30 - 18:30
Moderators: James Pekar, Brian Rutt

Fully Automated Data Management and Quality Assurance in Very Large Prospective Cohort MR Imaging Studies – the MR Imaging Study within the German National Cohort
Jochen G. Hirsch1, Alexander Köhn1, Daniel C. Hoinkiss1, Jonas Peter1, Andreas Thomsen1, Matthias Günther1,2, and the German National Cohort MRI Study Investigators3
1Fraunhofer MEVIS, Bremen, Germany, 2University Bremen, Bremen, Germany, 3NAKO MR Imaging Core, Munich, Germany
We present a fully automated data management workflow and quality assurance, which is set up for large, multicentric cohort studies including whole-body MR imaging. The workflow includes a modality worklist, exam-synchronous DICOM transfer to centralized storage, quality control of MR acquisition, various image-based quality measures, web-based radiological image review for incidental findings, visual quality scores, as well as long-term archiving. This workflow, implemented in the MRI Study of the German National Cohort, enables to acquire and process more than 30 whole-body MRI scans per day, available for IF reading within 4 hours. Deviations, outliers, technical failures are pointed out on-the-fly.

Automated slice positioning for 2D MRA in bolus tracking of DCE-MRI
Takao Goto1 and Mirai Araki1
1MR Engineering, GE Healthcare, Hino-shi, Japan
Accurate placement of a 2D plane across the aorta while examining scout images is a complex task and  makes the operator's workflow difficult in bolus tracking of DCE-MRI. We present a new method for automated slice positioning for 2D MRA used to monitor bolus arrival. The 2D plane was planned by aorta detection using both Hough Forests and AdaBoost classifiers following the classification of axial images. A dataset with 40 patients was tested, and 35 cases depicted the cross section of the aorta clearly. This automation will help the operator and decrease the total study time.

Automatic Pipeline for Regional Brain Analyses in Demyelinated Mice
Emilie Poirion1, Daniel Garcia Lorenzo1, Isaac Adanyeguh1, Marie-Stéphane Aigrot1, Alexandra Petiet2, and Bruno Stankoff1,3
1Brain and Spine Institute, INSERM U1127/CNRS UMR 7225, Sorbonnes Universités, UPMC, CHU Pitié-Salpêtrière, 47 Bd de l'hôpital, 75013 Paris, Paris, France, 2Brain and Spine Institute, Center for Neuroimaging Research (CENIR), CHU Pitié-Salpêtrière, 47 Bd de l'hôpital, 75013 Paris, Paris, France, 3AP-HP, Saint Antoine Hospital, Department of Neurology, 184 Bd du Faubourg Saint Antoine, 75012 Paris, Paris, France
Experimental studies in mouse models offer the opportunity to combine in-vivo longitudinal high-field MRI and histological analyses. However, automatic MRI tools for processing rodent data to avoid manual processing are lacking. We proposed an automatic pipeline to perform systematic analyses on large murine cohorts with longitudinal data. We first applied artifacts correction as bias correction to optimize the subsequent steps. We then registered masks of regions of interest (ROIs) for our analyses onto each subject from which we extracted the quantitative data. This pipeline provides a way of quickly analyzing ROI regardless of disease models or the MRI sequence.

Improving robustness in automated slice positioning for knee MR by combining landmark detection and image processing
Takamasa Sugiura1, Shuhei Nitta1, Taichiro Shiodera1, Yuko Hara1, Yasunori Taguchi1, Tomoyuki Takeguchi1, Takuya Fujimaki2, Kensuke Shinoda2, Hiroshi Takai2, and Ayako Ninomiya2
We propose an improved automatic slice positioning algorithm for knee MR which combines conventional machine-learning based landmark detection with advanced image processing techniques. Conventional slice positioning methods determine the diagnostic slice center and orientation by detecting anatomical landmarks in the scout image. However, computing slice positions from landmarks can be inadequate since landmarks vary across patients and can be cut-off from scout images. Here, we use not only landmark detection but also image processing based contour detection of the femoral condyle and angle estimation of the femur and tibia to enable slice positioning for a wider range of scout images.

3D magnetic resonance fingerprinting with a clustered spatiotemporal dictionary
Pedro A. Gómez1,2, Guido Buonincontri3, Miguel Molina-Romero1,2, Cagdas Ulas1,2, Jonatahn I. Sperl2, Marion I. Menzel2, and Bjoern H. Menze1
1Technische Universität München, Garching, Germany, 2GE Global Research, Garching, Germany, 3Istituto Nazionale di Fisica Nucleare, Pisa, Italy
We present a method for creating a spatiotemporal dictionary for magnetic resonance fingerprinting (MRF). Our technique is based on the clustering of multi-parametric spatial kernels from training data and the posterior simulation of a temporal fingerprint for each voxel in every cluster. We show that the parametric maps estimated with a clustered dictionary agree with maps estimated with a full dictionary, and are also robust to undersampling and shorter sequences, leading to increased efficiency in parameter mapping with MRF. 

Multi-dimensional phase unwrapping: a new and efficient linear algebraic formulation using weighted least-squares - Permission Withheld
Laurent Lamalle1,2, Georgios Gousios3, and Matthieu Urvoy3
1Inserm US 17 & CNRS UMS 3552, Grenoble, France, 2Université Joseph Fourier & CHU de Grenoble, UMS IRMaGe, Grenoble, France, 3SFR RMN Biomédicale et Neurosciences, Université Joseph Fourier, Grenoble, France
Phase information of MR images can provide quantitative access to various physical properties of the examined sample, such as local $$$B_0$$$ values, magnetic susceptibility or flow. Phase is a continuous information whose estimation typically requires unwrapping. In this study, we propose a novel phase estimation algorithm which: (1) relies on a numeric scheme that is robust to phase jumps, and (2) is optimized for execution on modern parallel processors.

Fast liver FOV localization for improved liver-MRI workflow - Permission Withheld
Arathi Sreekumari1, K S Shriram1, Uday Patil1, Ersin Bayram2, Dattesh Shanbhag1, and Rakesh Mullick1
1GE Global Research, Bangalore, India, 2GE Healthcare, Houston, TX, United States
In this work we are focusing on automating the scan coverage and FOV for liver MRI acquisitions. We demonstrate that using fast scout images, we can achieve very good localization of liver FOV, irrespective of anatomy differences and hand-up / hands-down positioning.

Simultaneous measurement of short and long T2* components using hybrid encoding
Hyungseok Jang1,2, Curtis N Wiens1, and Alan B McMillan1
1Department of Radiology, University of Wisconsin, Madison, WI, United States, 2Department of Electrical and Computer Engineering, University of Wisconsin, Madison, WI, United States
In this study, we propose a highly time efficient quantitative imaging scheme where short and long T2* components can be simultaneously estimated. This method is based on a multi-echo UTE hybrid encoding scheme, where the central SPI region is oversampled to allow measurement of short T2* across a wide range of TEs. The UTE acquisition is immediately followed by a gradient echo train to measure long T2*. We show the proposed method can obtain an extensive number of images (e.g., approximately 750 images) within a single acquisition and reasonable scan time.

Automatic Classification of Brain Connectivity Matrices - a toolbox for supporting neuropsychiatric diagnosis
Ricardo Jorge Maximiano1, Tiago Constantino1,2,3, André Santos-Ribeiro1,4, and Hugo Alexandre Ferreira1
1Institute of Biophysics and Biomedical Engineering, Faculty of Sciences of the University of Lisbon, Lisbon, Portugal, 2Spitalzentrum Biel, Bienne, Switzerland, 3Lisbon School of Health Technology - ESTeSL, Lisbon, Portugal, 4Centre for Neuropsychopharmacology, Imperial College London, London, United Kingdom
In this work, a user-friendly toolbox that aims to classify automatically brain connectivity matrices is described. To test this tool, we used the Parkinson’s Progression Markers Initiative (PPMI) data which includes structural and functional Magnetic Resonance Imaging data of healthy subjects, patients with “scans without evidence for dopaminergic deficit” (SWEDD) and patients diagnosed with Parkinson’s Disease (PD). Using default parameters, this tool was able to achieve a maximum accuracy of 85.4% in classifying the 3 groups of subjects by selecting features that were related to the rostral middle frontal gyrus and splenium, which are in agreement with PD literature.

Rapid Two-Step QSM Without A Priori Information
Christian Kames1,2, Vanessa Wiggermann1,3,4, and Alexander Rauscher1,4
1UBC MRI Research Centre, University of British Columbia, Vancouver, BC, Canada, 2Department of Engineering Physics, University of British Columbia, Vancouver, BC, Canada, 3Department of Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, 4Department of Pediatrics, University of British Columbia, Vancouver, BC, Canada
Current state-of-the-art QSM reconstruction algorithms are plagued by the trade-off between reconstruction speed and quality. We propose a novel two-step dipole inversion algorithm 20x faster than MEDI and HEIDI, while producing qualitatively appealing images with a root-mean-square error less than MEDI’s and HEIDI’s when compared to COSMOS. The proposed method works by first reconstructing the well-conditioned k-space region through the use of a Krylov subspace solver, followed by a total variation minimization to fill in the ill-conditioned k-space region.

The International Society for Magnetic Resonance in Medicine is accredited by the Accreditation Council for
Continuing Medical Education to provide continuing medical education for physicians.