Electronic poster

Thursday 13:30-15:30 Computer 123

13:30 5067. New Calculation Method of Pixel Shift Map on PSF Mapping Technique: A Study on 7T MRI

Se-Hong Oh¹, Jun-Young Chung¹, Myung-Ho In², Maxim Zaitsev³, Oliver Speck², Young-Bo Kim¹, Zang-Hee Cho<sup>1

1</sup>Neuroscience Research Institute, Gachon University of Medicine and Science, Incheon, Korea, Republic of; ²Department of Biomedical Magnetic Resonance, Institute for Experimental Physics, Otto-von-Guericke University Magdeburg, Magdeburg, Germany; ³Department of Radiologic Research, Medical Physics, University Hospital of Freiburg, Freiburg, Germany

Echo-planar imaging (EPI) is one of the fastest and most widely used MRI pulse sequences in the field of MRI. Compared to conventional imaging sequence, EPI is more prone to a variety of artifacts. A prominent EPI artifact is geometric distortion due to strong magnetic field inhomogeneity and susceptibility. Previous PSF mapping method, which was implemented by Zaitsev et al. used GE (Gradient Echo) StdDev (standard deviation) image as a base and produced a ¡°mask¡± to extrapolate pixel shift map. Flow artifact as well as setting of the parameters (i.e. threshold value) can affect the result of mask. And the extrapolated shift map which resulting shift maps with extrapolation eventually have error. Consequently corrected images will also have errors induced by mask errors and flow artifacts. So we propose new mask calculation method based on using a 2D PSF data based, not based on the GE StdDev image as previously used. This method is capable of making automatic mask calculation procedure, along with the advantage of eliminating flow induced ghost artifact all together.

14:00 5068. Accelerated Point Spread Function Mapping Using Compressed Sensing for EPI Geometric Distortion Correction

Iulius Dragonu¹, Juergen Hennig¹, Maxim Zaitsev<sup>1

1</sup>Diagnostic Radiology, University Hospital Freiburg, Freiburg, Baden Wuerttemberg, Germany

Single-shot echo-planar imaging (EPI) is a fast technique allowing the acquisition of an image following a single RF excitation. The high temporal resolution of EPI makes it the method of choice for applications such as fMRI or diffusion tensor imaging. However, EPI is prone to geometric and intensity distortions in the presence of magnetic field inhomogeneities. Several correction techniques have been proposed in the past based on field map acquisitions or point spread function (PSF) acquisitions. Parallel imaging techniques were employed for accelerating the PSF data acquisition. In this work we demonstrate that compressed sensing (CS) reconstruction can be used for acquiring the PSF data set with high acceleration factors for accurate geometric distortion corrections.

14:30 5069. Distortion Correction for Echo Planar MR Imaging Using the Point Spread Function (PSF) Map with Bregman Iteration

Yu Cai¹, Weili Lin, Qingwei Liu, Craig Hamilton², Hongyu An

¹Department of Radiology, University of North Carolina-Chapel Hill, Chapel Hill, NC, United States; ²Wake Forest University

Point Spread Function (PSF) mapping techniques have shown promise for geometric distortion correction in Echo Planar Imaging (EPI)(1), where the distortion information is mapped by applying additional phase encoding gradients with a constant time (PSF encoding). Cai et al(2) introduced the inverse solution of the PSF map with the Tikhonov regularization method for EPI distortion correction. The smoothness penalty in the Tikhonov regularization causes it sensitive to the aliasing artifact in its reconstructed image and fine textile structure blurring. Here we apply the total variation (TV) regularization with Bregman iteration method(3) to the PSF map in which the penalty term is adaptively updated based on the Bregman distance, which is immune to the above effects. The proposed approach compared with the Tikhonov regularization methods were evaluated at 3.0T with human subjects while at 9.4T with rats.

15:00 5070. Improved PSF-Based EPI Distortion Correction in Human Imaging at 7 Tesla

Myung-Ho In¹, Jun-Young Jung², Se-Hong Oh², Maxim Zaitsev³, Zang-Hee Cho², Oliver Speck<sup>1

1</sup>Department Biomedical Magnetic Resonance, Institute for Experimental Physics, Otto-von-Guericke University Magdeburg, Magdeburg, Germany; ²Neuroscience Research Institute, Gachon University of Medicine and Science, Inchoen, Korea, Republic of; ³Department of Radiologic Research, Medical Physics, University Hospital Freiburg, Freiburg, Germany

This method proposes an improved method with which distortions in EPI images in the ultra high field MRI such as 7.0 Tesla can be correct automatically and with high fidelity. The correction is a modification and extension of the point spread function (PSF) method previously developed. In addition to more precise mapping and correction of blurring, the method removes flow induced artifacts which can cause errors in the shift map derived from the PSF. The advantages of the proposed method for the correction of geometric distortions in EPI are demonstrated in human brain in vivo at 7.0 Tesla.

Topics in Image Analysis

Hall B Monday 14:00-16:00 Computer 124

14:00 5071. Automated Analysis of ACR Phantom Data as an Adjunct to a Regular MR Quality Assurance Program

Lawrence P. Panych^1,2, Lisa Bussolari¹, Robert V. Mulkern, <sup>23

1</sup>Radiology, Brigham and Women's Hospital, Boston, MA, United States; ²Radiology, Harvard Medical School, Boston, MA, United States; ³Radiology, Children's Hospital, Boston, MA, United States

A Matlab-based package for automatic analysis of phantom images was developed. The package analyzes images of the American College of Radiology (ACR) phantom, performing measurements similar to those required as part of the ACR accreditation program along with other useful measures. Analysis of the data from five 1.5T MR systems acquired during weekly QA scans was performed. Such data can be help to identify potential system problems, such as lower than usual SNR, and serve as an adjunct to a regular program of quality assurance.

14:30 5072. Weisskoff Stability Metrics Dependence on K-Space Trajectory

E. Brian Welch^1,2, Ad Moerland³, Elizabeth A. Moore³, J. Christopher Gatenby¹, John C. Gore<sup>1

1</sup>Vanderbilt University Institute of Imaging Science, Vanderbilt University Medical Center, Nashville, TN, United States; ²MR Clinical Science, Philips Healthcare, Highland Heights, OH, United States; ³MR Clinical Science, Philips Healthcare, Best, Netherlands

Measurement of an MR scanner’s signal stability is important for clinical and research sites acquiring data known to be adversely affected by system instabilities such as functional MRI. In particular, the Weisskoff plot and its associated radius of decorrelation (RDC) metric are often used to compare systems. The RDC is known to be influenced by the noise level of the data. However, it has not been widely reported that the observed RDC is also affected by k-space trajectory. Here we present stability measurements from a single scanner collected using four distinct k-space acquisition trajectories: Cartesian, multivane (propeller), spiral and radial.

15:00 5073. Array Coil Signal-To-Noise Ratio Measurement: A Comparison of Methods

Elizabeth M. Tunnicliffe¹, Martin J. Graves<sup>1

1</sup>Department of Medical Physics and Clinical Engineering, Cambridge University Hospitals, Cambridge, CB2 0QQ, United Kingdom

This phantom study compares six signal-to-noise ratio (SNR) measurement methods in two eight-channel head arrays. While the methods are equivalent for well de-coupled elements, this is not the case once noise correlations exist, which can indicate that an array is failing. In the absence of longitudinal SNR data, a comparison between the background region standard deviation SNR method and the noise power SNR method can provide evidence of array failure. In this situation the noise power method provides the most accurate estimate of the underlying SNR in a single image acquisition.

15:30 5074. Protocol for Regular Quality Control of MRI Scanners in a Clinical Setting

Joost Kuijer¹, Erwin Kist², Mark Hofman<sup>1

1</sup>Physics and Medical Technology, VU University Medical Center, Amsterdam, Netherlands; ²Radiology, VU University Medical Center, Amsterdam, Netherlands

A protocol for MRI Quality Control (QC) testing within 15 minutes was developed in a clinical setting including phantom positioning and image evaluation. The set-up consisted of a scan protocol using the ACR phantom, based on the ACR 2004 QC Manual. The images were sent from the scanner to a central server using a DICOM transfer, automatically analysed and compared to predefined action limits, based on clinical relevance and short-term reproducibility. Parameters included SNR, ghosting, image homogeneity and 3D geometric accuracy. Web-based reporting allowed direct feedback at the scanner, while trend plots provide insight in long-term stability.

Tuesday 13:30-15:30 Computer 124

13:30 5075. On Accelerated Dynamic Contrast-Enhanced Lung Perfusion Using K-T BLAST

Jia-Shuo Hsu¹, Shang-Yueh Tsai², Yi-Ru Lin³, Hsiao-Wen Chung<sup>1

1</sup>Institute of Biomedical Electronics and BioInformatics, National Taiwan University, Taipei, Taiwan; ²Dept. of Electrical Engineering, Chang-Gung University, Taoyuan, Taiwan; ³Dept. of Electronic Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan

k-t BLAST accelerates dynamic contrast-enhanced lung imaging with only limited penalty in RMS error. Yet its restrictions including initial-overshooting and temporal-smoothing cast uncertainties on perfusion quantifications and corresponding studies. This work shows that while those restrictions influence intensity estimation on patient images as predicted, little impact is inflicted on the corresponding perfusion quantification, suggesting feasibility of accelerated lung images in clinical studies.

14:00 5076. Nonrigid Registration Based Segmentation for the Analysis of Real-Time Cardiac Flow Images.

Freddy Odille¹, Jennifer Steeden¹, Vivek Muthurangu², David Atkinson<sup>1

1</sup>Centre for Medical Image Computing, University College London, London, United Kingdom; ²Centre for Cardiovascular MR, UCL Institute of Child Health, London, United Kingdom

Cardiac flow measurements can be obtained from real-time phase contrast MRI. Due to the compromised spatial resolution and signal-to-noise ratio, automatic segmentation of great vessels is challenging. Here, we propose to use nonrigid registration of the time series of magnitude images (148 frames) to propagate the segmentation performed manually in a reference frame. The registration, based on optical flow, includes smoothness constraints in both space and time, and is computationally very efficient. Flow measurements generated by manual and registration-based segmentations, as well as stroke volumes, were compared in data from 10 volunteers (rest and physical exercise), and showed good agreement.

14:30 5077. Center Point Trajectory Model for Cardiac Wall Motion Abnormality Assessment Compared with Echocardiography Strain

Ting Song^1,2, Alexander I. Bustamante³, Jeffrey A. Stainsby⁴, Maureen N. Hood^2,5, Vincent B. Ho<sup>2,5

1</sup>GE Healthcare Applied Science Laboratory, Bethesda, MD, United States; ²Radiology, Uniformed Services University of the Health Sciences, Bethesda, MD, United States; ³Cardiology, National Navy Medical Center, Bethesda, MD, United States; ⁴GE Healthcare Applied Science Laboratory, Toronto, ON, Canada; ⁵Radiology, National Navy Medical Center, Bethesda, MD, United States

A quantitative method, Central Point Trajectory (CPT), of assessing myocardial wall motion is evaluated in this paper. This center point trajectory method is compared and validated against echocardiography systolic peak strain maps and myocardial delayed enhancement images, which shows strong correlation in terms of detection of abnormal regions with reduced wall motion.

15:00 5078. Postprocessing Tool for 3D Strain Quantification from 3D Tagging Data

Marco Piccirelli¹, Roger Luechinger¹, Peter Boesiger<sup>1

1</sup>Institute for biomedical Engineering, University & ETH Zurich, Zurich, Switzerland

Tagging acquisition of the orbit during eye movement have proven to give new inside into the mechanical properties of orbital tissues, and to be a valuable tool for investigating ocular diseases etiologies. 3D tagging acquisition have been shown to be feasible. We present here a model-free method enabling to quantify out of 3D tagging data the inhomogeneous deformation along extraocular muscles. This tool able 3D strain quantification and is adaptable to other tissues, like the heart.

Wednesday 13:30-15:30 Computer 124

13:30 5079. Quantitative Analysis of Projection Breast Density Changes at Different Compression Angles Based on 3D MRI

Tzu-Ching Shih^1,2, Jeon-Hor Chen^2,3, D Chang³, K Nie³, M Lin³, O Nalcioglu³, Min-Ying Lydia Su<sup>3

1</sup>Department of Biomedical Imaging and Radiological Science, China Medical University, Taichung, Taiwan; ²Department of Radiology, China Medical University Hospital, Taichung, Taiwan; ³Tu & Yuen Center for Functional Onco-Imaging, University of California, Irvine, CA, United States

This study aims to demonstrate the effect of compression angle on the projection breast density at different compression ratios based on the patient-specific three-dimensional MR images. The fibroglandular tissue and tatty tissue were described by 3,488 and 11,803 tetrahedral elements. Within 50% to 70% compression ratio, the variation of the measured projection breast density was approximately 7%. In contrast, the variation of the projection breast density was nearly 11% for MOL view compression. This study provides a novel computer simulation approach to simulate the large deformation of breast compression. Compression angle of deviation may affect the measured projection breast density.

14:00 5080. Interactive Intensity Thresholding Based Breast Density Assessment in Sequential MR Examinations

Sadie Nicola Reed¹, Gokhan Ertas¹, Martin O. Leach<sup>1

1</sup>Cancer Research UK and EPSRC Cancer Imaging Centre, Institute of Cancer Research, Sutton, Surrey, United Kingdom

Breast density has been shown to be a strong risk factor for breast cancer. MR imaging allows direct volume estimation of the fibroglandular breast tissues providing an accurate breast density assessment. In this study, we investigate the value of interactive intensity thresholding in the assessment of breast density from sequential MR examinations. The results have shown a good consistency between the left and right breasts and a high level of reproducibility for sequential patient visits. The ability of the technique to highlight variations from normal breast development in sequential images could give valuable information in assessment of breast cancer risk.

14:30 5081. A Maximum Likelihood Method for Partial Volume Segmentation of Magnitude Breast MR Data

Melanie Freed^1,2, Christian Graff¹, Maria I. Altbach³, Jacco A. de Zwart⁴, Jeff H. Duyn⁴, Aldo Badano<sup>1

1</sup>CDRH/OSEL/DIAM, FDA, Silver Spring, MD, United States; ²Department of Bioengineering, University of Maryland, College Park, MD, United States; ³Department of Radiology, University of Arizona, Tucson, AZ, United States; ⁴NINDS/LFMI/Advanced MRI Section, National Institutes of Health, Bethesda, MD, United States

We apply maximum likelihood estimation techniques to magnitude MR images as a method for partial volume segmentation. The method is validated on noisy inversion recovery and saturation recovery images of a simulated MR breast phantom created from human CT data and then applied to inversion recovery images of a physical breast phantom. The segmentation algorithm is able to successfully separate tissue types in both simulated and phantom MR images.

15:00 5082. DIVA+QUADRANT: Novel Visualisation Software for DCE-MRI to Aid Breast Cancer Diagnosis and Neoadjuvant Chemotherapy Monitoring

Martin Lowry¹, David John Manton¹, Martin D. Pickles¹, Lindsay W. Turnbull<sup>1

1</sup>YCR Centre for MR Investigations, Hull-York Medical School, Hull, East Yorkshire, United Kingdom

The clinical utility of dynamic contrast-enhanced MRI (DCE-MRI) is well established, but the analysis of data by radiologists can be time-consuming. Novel visualisation software, called DIVA+QUADRANT, has been developed which quickly and clearly indicates those regions within a tumour which display the highest contrast agent enhancement (uptake) rate and the greatest degree of contrast agent wash out (signal decay); both well-established indicators of malignancy following the BIRADS-MRI lexicon. The software can also be used to monitor response to chemotherapy as it can map out areas where enhancement and washout rates have decreased, i.e. areas where vasculature shutdown is occurring.

Thursday 13:30-15:30 Computer 124

13:30 5083. Local Rigid and Volume Preserving Deformable Registration Method with Applications to Liver MR Data

Atilla Peter Kiraly¹, Christophe Chefd'Hotel¹, Clifford R. Weiss², Ralph Strecker<sup>3

1</sup>Imaging and Visualization, Siemens Corporate Research, Princeton, NJ, United States; ²Department of Radiology, The Johns Hopkins Univeristy School of Medicine, Baltimore, MD, United States; ³MR Oncology, Siemens Healthcare, Erlangen, Germany

A novel approach to ensure local rigidity and volume preservation with existing registration methods is presented. A modification to the deformation field is performed before application to the moving image. It adds little additional runtime and can be quickly implemented. Results are shown on T1 MR liver data with simulated enhancing structures to demonstrate the volume preserving properties.

14:00 5084. Non-Rigid Motion Compensation in MR Prostate Perfusion Imaging

Gert Wollny¹, Isabel Casanova¹, Thomes Hambrock², Andres Santos¹, Maria Jesus Ledesma-Carbayo<sup>1

1</sup>BIT-DIE, ETSI Telecomunicación, UPM, Madrid, Spain; ²Department of Radiology, Radboud University Nijmegen Medical Centre, Nijmegen, Netherlands

Dynamic Contrast enhance MRI has been established as accurate method in detection and localization of prostate cancer. Time series of three-dimensional datasets of the prostate are acquired and used to obtain per-voxel signal-intensity vs. time curves. These are then used to differentiate cancerous from non-cancerous tissue. However, rectal peristalsis and patient movement may result in spatial-mismatching of the serial datasets and, therefore, incorrect enhancement curves. In this work, we present a method based on image registration to compensate for these movements, and validate the method by comparing pre-and postregistration intensity time curves to manually obtained ones.

14:30 5085. Validation of Deformable Registration of Prostate MRI with and Without Endorectal Coil for IMRT Planning

Marnix Christiaan Maas¹, Corijn Kamerling¹, Simon van Kranen¹, Sara Muller², Jelle Teertstra², Floris Pos¹, Christoph Schneider¹, Jan Jakob Sonke¹, Marcel van Herk<sup>1

1</sup>Radiotherapy, NKI-AVL, Amsterdam, Netherlands; ²Radiology, NKI-AVL, Amsterdam, Netherlands

Using an endorectal coil (ERC) greatly improves the quality of prostate MR images, but results in displacements and deformations of the organ and its surrounding tissues, causing systematic errors in intensity modulated radiation therapy (IMRT) planning. We have implemented an image based method for the deformable registration of endorectal to non-endorectal MR images. Here we present a validation of this method using markers placed on corresponding anatomical structures in pairs of fixed and deformed images. The registration method was found to be feasible, and our results suggest that sufficient accuracy for use in radiotherapy planning is attainable.

15:00 5086. Automatic Segmentation of the Prostate in MR Images Using a Prior Knowledge of Shape, Geometry and Gradient Information

Yujin Jang¹, Helen Hong¹, Hak Jong Lee², Sung Il Hwang<sup>2

1</sup>Division of Multimedia Engineering, Seoul Women's University, Seoul, Korea, Republic of; ²Department of Radiology, Seoul National University Hospital of Bundang, Seongnam-si, Korea, Republic of

To segment the prostate in MR images with a poor tissue contrast and shape variation, we propose a reliable and reproducible segmentation method using a prior knowledge of shape, geometry and gradient information. The prostate surface is generated by 3D active shape model using adaptive density profile and multiresolution technique. To prevent holes from occurring by the convergence of the surface shape on the local optima, the hole is eliminated by 3D shape correction using geometry information. In the apex of the prostate which has a large anatomical variation, the boundary is refined by 2D contour correction using gradient information.

Image Analysis

Hall B Monday 14:00-16:00 Computer 125

14:00 5087. Automated Assessment of Ghost Artifacts in MRI

Sotirios A. Tsaftaris^1,2, Xiangzhi Zhou², Rohan Dharmakumar<sup>2

1</sup>Electrical Engineering and Computer Science, Northwestern University, Evanston, IL, United States; ²Radiology, Northwestern University, Chicago, IL, United States

Flow artifacts in MR images can appear as image ghosts within and outside the body cavity. Technical improvements to suppress these ghosts often rely on expert scoring or on semi-automated methods demanding tissue segmentation to evaluate the efficacy of the methods. These approaches can be labor/computation intensive, introduce observer bias, or error-prone if tissue segmentation is used. Herein we propose two fully automated image-processing methods relying on the statistical properties of background pixels to assess the presence of flow artifacts (appearing as image ghosts) without requiring segmentation. We demonstrate that the automated methods are as effective as image scoring approaches that rely on expert reviewers.

14:30 5088. Total Variation Denoising with Spatially Dependent Regularization

Florian Knoll¹, Yiqiu Dong², Christian Langkammer³, Michael Hintermüller^2,4, Rudolf Stollberger<sup>1

1</sup>Institute of Medical Engineering, Graz University of Technology, Graz, Austria; ²Institute of Mathematics and Scientific Computing, University of Graz, Graz, Austria; ³Department of Neurology, Medical University Graz, Graz, Austria; ⁴Department of Mathematics, Humboldt-University of Berlin, Berlin, Germany

The Total Variation regularization model is popular in MR research. In this model, a regularization parameter controls the trade-off between noise elimination, and preservation of image details. However, MR images are comprised of multiple details. This indicates that different amounts of regularization are desirable for regions with fine image details in order to obtain better restoration results. This work introduces spatially dependent regularization parameter selection for TV based image restoration. With this technique, the regularization parameter is adapted automatically based on the details in the images, which improves the reconstruction of details as well as providing an adequate smoothing for the homogeneous parts.

15:00 5089. Modeling Non-Central Chi Distributed Noise in T1-Weighted MR Images: Brain Tissue Segmentation Using Partial Volume Densities

Hugo Gerard Schnack¹, Rachel Brouwer¹, Hilleke Hulshoff Pol<sup>1

1</sup>Psychiatry, UMC Utrecht, Utrecht, Netherlands

A brain tissue segmentation algorithm is developed that includes a non-central Chi description of the MR scanner noise. It is applied to a set of MR images of 16 healthy human volunteers and found to produce significantly different tissue volume estimates, when compared to models incorporating Gaussian noise.

15:30 5090. Automatic Quality Assessment for Multi-Slice 2D FLAIR MR Imaging

Bénédicte Mortamet¹, Matt A. Bernstein², Clifford R. Jack², Jeffrey L. Gunter², Maria Shiung², Reto Meuli³, Jean-Philippe Thiran⁴, Gunnar Krueger<sup>1

1</sup>Advanced Clinical Imaging Technology, Siemens Healthcare Sector IM&WS S - CIBM, Lausanne, Switzerland; ²Mayo Clinic, Rochester, MI, United States; ³CHUV, Radiology, Lausanne, Switzerland; ⁴Signal Processing Laboratory (LTS5) Ecole Polytechnique Fédérale de Lausanne

The FLAIR contrast is increasingly used as part of routine protocol for brain MRI. It provides high sensitivity to a wide range of disease but is susceptible to patient motion. Resulting artifacts may obscure the pathology or mislead automated image analysis algorithms. We propose a method that automates quality classification of T2w 2D-FLAIR data. The validation based on 99 head scans confirms the robustness and reliability of the method. It could greatly improve clinical workflow as, in particular if integrated in online image reconstruction, it could provide immediate feedback to the MR technologist to repeat low-quality scans within the same session.