Michael Andreas Bieri¹, Christoph Barmet¹, Bertram Jakob Wilm¹, and Klaas Paul Pruessmann^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Zurich, Switzerland

FFT-based MR reconstruction algorithms get increasingly complex when incorporating various encoding terms such as static off-resonance, dynamic higher-order fields, coil sensitivity maps, non-Cartesian k-space sampling, etc. Therefore in this work a versatile algebraic reconstruction is employed that allows for incorporating various encoding mechanisms with only minor increase in complexity. A drawback is the enormous computing time even for small images. A 100x100 image needs minutes to be reconstructed on a current high end CPU. By implementing the algorithm on a graphics processing unit (GPU), speed-up of about 600x was achieved compared to a CPU. All result are shown at the example of dynamic higher order fields combined with static off-resonance and sensitivity encoding.

David S Smith^1,2, John C Gore^1,2, and Edward Brian Welch^1,2
1Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, United States, ²Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States

We show that compressed sensing MRI reconstruction using a Cartesian split Bregman solver can be dramatically accelerated using GPU computing. We find a factor of ~30 speedup for images of square dimension 512 and higher on a system based on an NVIDIA Tesla C2050 GPU card and Accelereye's Jacket GPU wrapper for MATLAB.

Nady M Obeid¹, Ian C Atkinson², Keith R Thulborn², and Wen-Mei W Hwu^1
1Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Champaign, IL, United States, ²Center for Magnetic Resonance Research, University of Illinois at Chicago, Chicago, IL, United States

In this work, we propose and elaborate on a method for accelerating gridding for the reconstruction of Non-Cartesian MRI using graphics processors (GPUs). Gridding interpolates the non-Cartesian input data onto a Cartesian grid as a preprocessing step to the FFT. We will explain the transformations needed to port the original CPU algorithm to the GPU, and how those transformation help improve performance. Our GPU-accelerated algorithm achieves up to 29X in runtime speedup for the gridding step, making fast reconstruction of non-Cartesian MR imaging data possible.

Seunghoon Nam^1,2, Tamer Ahmed Basha², Mehmet Akçakaya², Christian Stehning³, Warren J. Manning², Vahid Tarokh¹, and Reza Nezafat^2
1SEAS, Harvard University, Cambridge, MA, United States, ²Dept. of Medicine (Cardiovascular Division), Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, United States,³Philips Research, Hamburg, Germany

Three dimensional radial (kooshball) trajectory allows high isotropic spatial resolution, better depiction of cardiac anatomy and ease of image prescription in whole heart cardiac MRI. While kooshball trajectory can provide significant reduction in acquisition time compared with 3D Cartesian sampling trajectories, its application has been limited due to high computational burden associated with the reconstruction process. In this abstract, we implement and evaluate an iterative compressed sensing reconstruction method for kooshball trajectory on Graphics Processing Units (GPU). The GPU implementation shows 58× faster reconstruction time compared with a sequential C++ implementation.

Cheng Ouyang^1,2, Tobia Wech^1,3, and Li Pan^1,4
1Center for Applied Medical Imaging, Siemens Corporate Research, Baltimore, MD, United States, ²Bioengineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ³Institute of Radiology, University of Wuerzburg, Wuerzburg, Bavaria, Germany, ⁴Department of Radiology and Radiological Science, Johns Hopkins University, Baltimore, MD, United States

Compressed sensing (CS) has been proposed as a technique to enable acquisition and reconstruction of images that are sparse or compressible with little to no loss of image quality. However, the existing work in the CS literature has been focused on offline reconstruction and simulation, partially due to the concern of the time-consuming steps of the sparsifying transform and non-linear iterative reconstruction. In this work, we demonstrated the feasibility of an online implementation of compressed sensing to achieve real-time image reconstruction on a clinical scanner. The performance of the TVCMRI algorithm used in the online implementation was effective in producing reconstructed images close to ground truth with rapid reconstruction speed.

Mehmet Akcakaya *¹, Seunghoon Nam *^1,2, Tamer Basha¹, Vahid Tarokh², Warren J. Manning¹, and Reza Nezafat^1
1Dept. of Medicine (Cardiovascular Division), Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, United States, ²School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, United States

3D non-Cartesian sampling trajectories allow high isotropic spatial resolution, better depiction of cardiac anatomy and ease of image prescription in cardiac MRI. Furthermore, undersampling with these trajectories causes incoherent artifacts that may be removed using compressed sensing (CS). CS reconstruction is typically done using conjugate-gradient (CG) type algorithms, which require gridding and regridding to be performed at every iteration. In this abstract, we investigate an alternative method for CS reconstruction that only requires two gridding and one regridding operation in total irrespective of the number of iterations.

Anja Brau¹, Peng Lai¹, Srihari Narasimhan², Babu Narayanan³, and Vijaya Saradhi^2
1Global Applied Science Laboratory, GE Healthcare, Menlo Park, CA, United States, ²Computing & Decision Sciences Lab, GE Global Research, Bangalore, India, ³Medical Image Analysis Lab, GE Global Research, Bangalore, India

Computation of kernel weights as part of calibration in Compressed Sensing and Parallel Imaging algorithms like ESPIRiT and L1-SPIRiT is computationally expensive, especially for high channel count reconstructions. The weights are computed by obtaining a least squares fit for predicting target points in the calibration region using a set of source points in their neighborhood. The number of points in the neighborhood and their distance from the target point define quality of fit and computational complexity of the solver. We show that an optimally shaped neighborhood can give a significant improvement in the computational performance without sacrificing the image quality.

Francesco Santini¹, Sunil Patil^1,2, and Klaus Scheffler^1
1Radiological Physics, University of Basel Hospital, Basel, Basel, Switzerland, ²Center for Applied Medical Imaging, Siemens Corporation, Corporate Research, Baltimore, MD, United States

Image postprocessing and analysis is often an important part of the image reconstruction process, especially in quantitative MR imaging. This step is usually performed in a research environment by exporting images to a separate desktop computer and using commercial software packages. In this work, we present a simplified framework based on an open source scripting engine (Lua) that integrates in the scanner reconstruction pipeline, and allows simple and fast deployment of postprocessing and analysis programs.

Berengar W Lehr¹, Ferdinand Schweser¹, Andreas Deistung¹, Daniel Güllmar¹, and Jürgen R Reichenbach^1
1Medical Physics Group, Department of Diagnostic and Interventional Radiology I, Jena University Hospital, Jena, Germany

Prototype processing programs developed in research are mostly cumbersome to operate. Thus the way for a new processing method into clinical routine is long. The Medical Computation Server is an approach to integrate prototype processing programs into the existing clinical workflow. Providing a fast method to check if research promises are fulfilling clinical needs it indeed bridges the gap.

Florian Knoll¹, Manuel Freiberger¹, Kristian Bredies², and Rudolf Stollberger^1
1Institute of Medical Engineering, Graz University of Technology, Graz, Austria, ²Institute for Mathematics and Scientific Computing, University of Graz, Graz, Austria

Iterative image reconstruction of undersampled data from multiple coils has shown a huge potential for a wide range of applications during the last years. One of the main restrictions of these methods is the prolonged image reconstruction time. Parallelized implementations on graphics hardware were recently discovered as a feaseable method to significantly speedup image reconstruction. While programming graphics hardware was simplified significantly with the introduction of dedicated libraries for general purpose computing like CUDA or OpenCL, efficient implementation, especially concerning memory management, is still a challenging task. The goal of this work is to introduce an open source library designed for image reconstruction on GPUs. It is based on highly efficient implementations of numerical methods, but also includes code for iterative MR image reconstruction as well as a framework for finite element calculations and applications for Fluorescence Tomography. The results from this work illustrate the pronounced computational speedup with the GPU implementation. As the toolbox is an open-source project, all algorithms can be used either as a black box for the reconstruction of new data, or as a basis for own implementations for similar problems. The latter case is facilitated by the object-oriented and templated design which provides a well defined structure for extensions and allows for maximum code reusability. The intention behind the release of this open-source library is to alleviate the usage of the huge potential of graphics hardware in medical image reconstruction.

Elisa Placidi¹, Antonio Napolitano², Caroline L Hoad¹, Luca Marciani³, Klara C Garsed³, Robin C Spiller³, and Penny A Gowland^1
1SPMMRC, University of Nottingham, Nottingham, United Kingdom, ²Academic Radiology, QMC, Nottingham, United Kingdom, ³Nottingham Digestive Diseases Centre Biomedical Research Unit, Nottingham, United Kingdom

This abstract presents the development and the comparison of two automated operator-independent methods for the objective classification of the AC contents, based on previously optimised MRI techniques to study the gastrointestinal tract. A Gaussian fit of the histogram of the ROI and a texture analysis based on Gabor wavelet are compared: the first one results very time consuming while the second one is very quick and in accordance with the experimental results and it could be used in future studies for assessing the effectiveness of a range of agents designed to treat diarrhoeal diseases.

Li Dong¹, Hao Shen², Xiaojie Zhang¹, Wei Yu¹, Zhaoqi Zhang¹, Hua Guo³, Ren Wang¹, Dongxu Lu¹, and Chun Yuan^3,4
1Capital Medical University, Beijing Anzhen Hospital, Beijing, Beijing, China, People's Republic of, ²GE Healthcare, ³School of Medicine, Tsinghua University, ⁴University of Washington

We hypothesized that 3D black blood, a single sequence, can quantify carotid stenosis and plaque composition simultaneously. Bilateral carotid arteries from 9 symptomatic subjects underwent 3D SPGR and MRA scans. There was an excellent correlation of measuring stenosis between 3D SPGR and MRA (r = 0.95, p < 0.001). For those arteries with corresponding histology (n = 5), good agreements in identifying major plaque components were found. Therefore, 3D black blood images allow interactive reconstruction in arbitrary planes providing a more thorough visualization of luminal severity and plaque components.

Silvain Bériault¹, Fahd Al Subaie², Kelvin Mok³, Abbas F Sadikot², and G Bruce Pike^1
1McConnell Brain Imaging Centre, Montreal Neurological Institute, Montréal, Québec, Canada, ²Department of Neurology and Neurosurgery, Montreal Neurological Institute, ³Neuronavigation Unit, Montreal Neurological Institute

We propose an automated method for preoperative trajectory planning of deep brain stimulation image-guided neurosurgery. Our framework integrates multi-modal MRI analysis (T1w, SWI, TOF-MRA) to determine an optimal trajectory to the subthalamic nuclei while avoiding critical brain structures for prevention of hemorrhages, loss of function and other complications. Results show that our automatic method aggregates several surgical requirements into a meaningful trajectory ranking, providing neurosurgeons with an intuitive decision-support system intended to improve standard (manual) planning techniques offered by commercial neuronavigation platforms. Our method analyzes over 12,000 trajectories in less than 4 minutes.

Simone Bottan¹, Andri Fritz¹, Vartan Kurtcuoglu¹, Marianne Schmid Daners², Verena Knobloch³, Christian Langkammer^4,5, Nikolaus Krebs⁵, monika Gloor⁶, Eva Scheurer⁵, Klaus Scheffler⁴, Stefan Ropele⁴, Peter Boesiger³, Dimos Poulikakos¹, and Michaela Soellinger^4
1Laboratory of Thermodynamics in Emerging Technologies, Department of Mechanical and Process Engineering, ETH Zurich, Zurich, Zurich, Switzerland, ²Institute for Dynamic Systems and Control, ETH Zurich, Zurich, Switzerland, ³Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland, ⁴Department of Neurology, Medical University of Graz, Graz, Austria, ⁵Ludwig Boltzmann Institute for Clinical-Forensic Imaging, Graz, Austria, ⁶Department of Radiology, University Hospital Basel, Basel, Switzerland

Cerebrospinal fluid (CSF) shows pulsating dynamics, due to the transmission of the cardiac pulse pressure wave to the ventricular system. The exact description of this transmission is still debated in the scientific community. Several studies have shown choroid plexus (CP) to play an important role to establish CSF dynamics. In the presented study, we evaluated volumes of CP by post-mortem bSSFP MRI and compared them to in-vivo CSF flow measurements. Our results showed a very small volumetric deformation of CP to be necessary to induce in-vivo CSF strokes in the cerebral aqueduct.

Dongxiang Xu¹, Aaron Black², Yihua Liao², Timothy Carroll², Debiao Li², James Carr², Chun Yuan³, and Mary M McDermott^2
1Radiology, University of Washington, Seattle, WA, United States, ²Northwestern University, ³University of Washington

In a long term, large PAD clinical trial, the reproducibility of MRI based plaque burden evaluation is very important to the success of study. To evaluate the reliability of multi contrast MR imaging technique on femoral artery, two variability studies on a subset of trial data were conducted. The statistical analysis on quantitative plaque measurements demonstrated that reproducibility of MR imaging is satisfactory in PAD study. Using this technique can also help reduce the Inter-Reader variability in follow up plaque analysis. This means that current techniques used in MR imaging and plaque burden analysis can provide effective and reliable quantitative measurements for peripheral arterial disease evaluation.

Rupeng Li¹, J,B, Stephenson IV², Christopher Pawela³, Ji-Geng Yan², Andrew Nencka³, Anthony G Hudetz⁴, Hani Matloub², and James S Hyde^1
1Biophysics, Medical College of Wisconsin, Milwaukee, WI, United States, ²Plastic Surgery, Medical College of Wisconsin, ³Biophysics, Medical College of Wisconsin, ⁴Anesthesiology, Medical College of Wisconsin

We demonstrated that ICA can be used to show the somatosensory network in rat brain. This network is symmetrical in both hemispheres across the S1 area in healthy animals. Following brachial plexus avulsion injury, this network was disrupted. The somatosensory network on the injuried side switched from a synergic functional network to an anti-correlated network, which, we believe, is related to inhibitory neuronal activity.

Kathrin Ogris^1,2, Martin Urschler^1,3, Andreas Petrovic^1,4, Kathrin Yen^1,5, and Eva Scheurer^1,5
1Ludwig Boltzmann Institute for Clinical- Forensic Imaging, Graz, Austria, ²Department of Forensic Medicine, Medical University, Graz, Austria, ³Institute for Computer Graphics and Vision, University of Technology, Graz, Austria, ⁴Institute of Medical Engineering, University of Technology, Graz, Austria, ⁵Medical University, Graz, Austria

While the detection of hemorrhage in clinical medicine is focused on organs and body cavities forensic imaging requires reliable localization and analysis of hematomas in soft tissues such as the subcutaneous fatty tissue for the reconstruction of events. To evaluate different MRI sequences with respect to their ability of showing high contrast between hematoma and fatty tissue pork phantoms were examined after injection of small blood volumes, and the reproducibility of volume estimation by manual segmentation was assessed. A TSE T1T2w sequence yielded the best results for volume estimation of hematoma in fatty tissue. Reproducibility of the measurements was excellent.

Vitali Zagorodnov¹, and Kallam Hanimi Reddy^2
1Nanyang Technological University, Singapore, NA, Singapore, ²Nanyang Technological University

Despite steady improvements in automated brain structural measurement algorithms (FreeSurfer, VBM, CLASP) over recent years, segmentation errors still frequently occur and require tedious manual review and editing of the segmentation results. We propose a framework based on Support Vector Regression to automatically highlight the errors in cortical thickness measurements, obtained using FreeSurfer segmentation pipeline. Our approach exploits high correlations between regional cortical thicknesses , which allows “guessing” what the correct measurement should be for a specific region, based on the rest of the brain measurements.

Martin David King¹, Fernando Calamante², Chris A Clark¹, and David Gadian^1
1Institute of Child Health, University College London, London, United Kingdom, ²Brain Research Institute, Melbourne, Australia

A common feature of many MRI data analyses in an independent, voxel-by-voxel treatment. In many applications this is expected to be inefficient in its use of the data, and improved parameter estimates should be attainable by adopting a statistical model in which image voxels are modelled as belonging to a population with an underlying statistical distribution. Among the methods that are well documented in the statistics literature is Bayesian spatial random effects modelling, implemented using Markov chain Monte Carlo. In this work we use dynamic susceptibility contrast data to illustrate the strengths of the Bayesian random effects modelling approach.

Astrid Franz¹, Henriette Tschampa², Andreas Müller², Stefanie Remmele¹, Jochen Keupp¹, Jürgen Gieseke³, Hans Heinz Schild², and Petra Mürtz^2
1Philips Research, Hamburg, Germany, ²Department of Radiology, University Hospital Bonn, Bonn, Germany, ³Philips Healthcare, Hamburg, Germany

We present a nearly automatic segmentation and classification algorithm for human brain tumor tissue working on a combination of magnetic resonance T1 weighted contrast enhanced images and FLAIR images, based on a simple region growing technique. Algorithmic parameters are adapted automatically in the course of growing. The only required user interaction is a mouse click for providing the starting point. The algorithm is robust, i.e. independent of the given starting point within the tumor, and avoids leakage. We validated the algorithm on 20 test cases of human glioblastoma and meningioma. In 85% of the test cases we got satisfactory results.

Yalin Wang¹, Ashok Panigrahy^2,3, Rafael Ceschin², Songling Liu¹, Paul M Thompson⁴, and Natasha Leporé^3
1Computer Science and Engineering, Arizona State University, Tempe, AZ, United States, ²Radiology, Children's Hospital of Pittsburgh of UPMC, Pittsburgh, PA, United States, ³Radiology, Childrens Hospital Los Angeles, Los Angeles, CA, United States, ⁴Laboratory of Neuro Imaging, UCLA School of Medicine, Los Angeles, CA, United States

Changes in surface morphometry of the corpus callosum and lateral ventricles are likely sensitive indicators of diffuse white matter injury and of the interrelated subcortical grey matter injury in preterm neonate. Using brain structural magnetic resonance (MR) images, we propose a novel pipeline for regional group comparisons of the surface anatomy of subcortical structures in neonates. We study a total of 7 brain subcortical structures: the corpus collosum, lateral ventricle, thalamus, hippocampus, caudate nucleus, putamen and the 3rd ventricle. Our analysis will be applied to compare MR data of premature neonates to those of healthy term born neonatal controls.

Fahime Sheikhzadeh^1,2, and Roger Tam^2,3
1Biomedical Engineering Program, University of British Columbia, Vancouver, BC, Canada, ²MS/MRI Research Group, University of British Columbia, Vancouver, BC, Canada, ³Department of Radiology, University of British Columbia, Vancouver, BC, Canada

Most studies on multiple sclerosis (MS) lesions focus on correlation between lesion volume and clinical measures, such as the Expanded Disability Status Scale (EDSS). This preliminary study investigates the contribution of the spatial distribution of lesions to disability. The white matter lesions on the T2w and PDw MRIs of 24 MS patients were delineated and the 3D spatial distribution of the lesion voxels in each patient was computed using the variance of Euclidean distances from a fixed reference point. Correlation and linear regression analysis show that lesion distribution contributes independently to EDSS and is potentially more sensitive than lesion volume.

Saifeng Liu¹, Jaladhar Neelavalli², Jin Tang¹, and Ewart Mark Haacke^2,3
1School of Biomedical Engineering, McMaster University, Hamilton, Ontario, Canada, ²The Magnetic Resonance Imaging Institute for Biomedical Research, Detroit, Michigan, United States,³Academic Radiology, Wayne State University, Detroit, Michigan, United States

Susceptibility quantification for small objects was improved by considering the effective magnetic moment. We proved that the product of measured susceptibility and the apparent volume stays constantly at different echo times, and showed that susceptibility quantification for small objects is limited by error in volume estimation.

Noam Alperin¹, Snag Lee¹, Derek Monette¹, Ahmet Bagci¹, Birgit Ertl-Wagner², and Raymond Sekula^3
1University of Miami, Miami, FL, United States, ²University of Munich, Germany, ³Allegheny General Hospital, Pittsburgh

The diagnosis of Chiari Malformations type I (CMI) is based on tonsillar herniation of 5 mm or more below foreman magnum. However, this criterion is poorly correlated with symptoms severity or outcome. The size of the posterior cranial fossa (PCF) has been shown to be a more sensitive marker for CMI. Currently there is no reliable method for volumetric assessment of the PCF. An automated method for measurements of PCF and hindbrain tissue volumes has been developed and applied to CMI. Results demonstrate, for the first time, that in addition to smaller PCF, CMI is also associated with a smaller 4th ventricle.

Priya Goel¹, Sushmita Datta¹, Kurt H Bockhorst¹, Jovany C Navarro², Claudia S Robertson², and Ponnada A Narayana^1
1Diagnostic and Interventional Imaging, Medical School, The University of Texas Health Science Center at Houston, Houston, TX, United States, ²Neurosurgery, Baylor College of Medicine, Houston, TX, United States

In this study, we investigated the regional volume changes in brain caused by mild traumatic brain injury with hemorrhagic shock in rat models using tensor-based morphometry (TBM) that has methodological advantages over the commonly used voxel-based morphometric (VBM) techniques. Our analysis resulted in the detection of atrophy at the site of injury and some deep gray matter structures such as amygdala and suprachiasmatic nucleus which might perhaps explain the various post-traumatic disorders that the mTBI victims suffer.

Taehoon Shin¹, Bosco S Tjan², and Krishna S Nayak^3
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Psychology and Neuroscience, University of Southern California, Los Angeles, CA, United States, ³Electrical Engineering, University of Southern California, Los Angeles, CA, United States

First-pass cardiac perfusion MRI often suffers from artifactual sub-endocardial dark rim that lowers diagnostic specificity and limits its routine clinical use. Insufficient spatial resolution and motion during data acquisition have been identified as potential causes of dark rim artifact (DRA). This study uncovers a new potential source of perceived DRA that is rooted in human visual illusion. We demonstrate that DRA can be perceived solely due to visual illusion in the absence of any true signal dip, using numerical phantoms and in-vivo perfusion images.

Elmar Laistler¹, and Ewald Moser^1
1MR Center of Excellence, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria

We present a technique to obtain quantitative parameters describing the geometry of the cutaneous vessel tree. Data is obtained from MR microscopy of the human skin in vivo by transforming the segmented vessel data into a graph. The technique might help in diagnosis of vascular inflammation or malformation and can be used for treatment monitoring due to its non-invasiveness.

Shenghong Ju¹, and Xin-Gui Peng^1
1Department of Radiology, Zhongda Hospital, Southeast University, Nanjing, Jiangsu, China, People's Republic of

T1WI can be used to measure liver volume and observe fat distribution; IP-OP method gives a FC that is significantly lower than the actual value, especially in higher fat concentrations; CSI and 1H MRS are accurate in quantifying fat in both phantoms and liver in vivo. Given their excellent correlation and concordance with LL analysis, CSI and MRS may replace liver fat histological and chemical analysis in longitudinal studies.

Rado Andriantsimiavona¹, Martin Leach¹, Simon Doran¹, David Collins¹, Soeren Grimm², and Dow-Mu Koh^1
1Institute of Cancer Research UK, Sutton, Surrey, United Kingdom, ²Biotronics3D Ltd., London, United Kingdom

Tissue characterisation based on multiparametric MR imaging data can be harnessed to evaluate tumour heterogeneity and their response to conventional and novel therapies. Unfortunately, no framework and application enables these complex datasets to be displayed and correlated with each other. We have built a functional imaging framework to link analysis of multimodality and multiparametric data, explore and characterise tumoural tissues types. We introduced a 3-class fuzzy classification / ternary bit combination merging method to segment tumour and identify heterogeneous tumoural areas. We applied the method to clinical trial liver metastases dataset and automatically identified necrosis among other tumoural tissue subtypes.

Shing Chun Benny Lam¹, Jeremy F. Magland¹, Scott N. Hwang², and Felix W. Wehrli^1
1Radiology, University of Pennsylvania, Philadelphia, PA, United States, ²Radiology, Emory University, Atlanta, GA, United States

The structure model index (SMI) is a metric characterizing the structure type (plate versus rod) of a trabecular bone network. However, application of SMI to in vivo MRI has been limited by spatial resolution. Here, we explored two up-sampling methods, sinc-interpolation and subvoxel processing, to enable computation of SMI from in vivo MRI. Comparison with values derived from CTimages (regarded as ground truth) showed SMI values obtained from subvoxel processed images to be closer to ground truth than those derived from sinc-interpolated images in terms of correlation coefficients and root mean square differences. Finally, subvoxel processed images were less sensitive to variations in binarization threshold.

Artem Mikheev¹, Vivian S Lee¹, and Henry Rusinek^1
1Radiology, NYU School of Medicine, New York, NY, United States

Simple coregistration algorithms do not work well for DCE images because of the change in intensity due to the uptake and washout of contrast agent. We have implemented several coregistration algorithms and tested them in the setting of functional exams of the kidney. The variants differ in sequencing of coregistration steps and in the way the organ of interest is specified. Serial and full 4D modes were found to be more accurate than radial mode. There was no statistically significant advantage of full 4D over serial modes. The execution times ranged 16-25 sec for serial, 14-16 sec for radial, and 61-73 sec for full 4D. Targeted coregistration of time series data appears to be reliable and ready for clinical use.

Vincenzo Positano¹, Ilaria Bernardeschi¹, Virna Zampa², Martina Marinelli³, Maria Filomena Santarelli³, and Luigi Landini^1,4
1MRI Lab, Fondazione G.Monasterio CNR-Regione Toscana, Pisa, Italy, ²Department of Diagnostic and Interventional Radiology, University of Pisa, Pisa, Italy, ³Institute of Clinical Physiology, Pisa, Italy, ⁴Department of Information Engineering, University of Pisa, Pisa, Italy

The aim of the study is to develop an automatic image registration technique able to compensate kidney motion in renal perfusion studies. The developed method is able to compensate kidney motion with a precision not significantly different from inter-observer variability of human operators.

Stathis Hadjidemetriou¹, Wilfried Reichardt¹, Juergen Hennig¹, and Martin Buechert^2
1Department of Diagnostic Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany, ²MRDAC, University Medical Center Freiburg, Freiburg, Germany

Autosomal dominant polycystic kidney disease (ADPKD) is a genetic condition in which growing cysts disrupt the parenchyma and increase the total kidney volume. It eventually leads to kidney failure and no treatment for it is yet available. However, clinical treatment trials are performed monitored with MRI. It has been shown that imaging biomarkers extracted from MRI data can be used as surrogate markers for the progression and treatment of the condition. A method is presented for the automated extraction of the biomarkers from the images that can accelerate the trials and improve the objectivity of the results.

Kim Mouridsen¹, Anders Neumann², Lars Riisgaard Ribe³, Kristjana Yr Jonsdottir³, and Leif Østergaard^3
1Center for Functionally Integrative Neuroscience, Aarhus University Hospital, Aarhus University, Aarhus, Denmark, ²Aarhus University Hospital, ³Center for Functionally Integrative Neuroscience, Aarhus University Hospital, Aarhus University

The mismatch volume between perfusion weighted- and diffusion weighted imaging is an emerging target for thrombolytic therapy in clinical management and clinical trials. However, the critical delineation of the perfusion lesion boundary suffers extreme interobserver variability, due to diffuse gradients between normal and ischemic tissue. We propose estimating the perfusion lesion by finding a smooth contour, which minimizes variation around a patient-specific elevated MTT value inside the contour, and variation around a normal value outside the lesion, using variational calculus. We show that, in contrast to simple thresholding, this technique compares excellently with lesion outlines by four expert neuroradiologists

Jerry S. Cheung¹, Enfeng Wang^1,2, Xiaoying Wang³, and Phillip Zhe Sun^1
1Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, MGH and Harvard Medical School, Charlestown, MA 02129, United States, ²Department of Radiology, 3rd Affiliated Hospital, Zhengzhou University, China, People's Republic of, ³Neuroprotection Research Laboratory, Department of Radiology and Neurology, MGH and Harvard Medical School, Charlestown, MA 02129, United States

Iterative self-organizing data analysis technique algorithm (ISODATA) has been increasingly used to classify multi-parametric data for delineation of heterogeneous ischemic damage. Our study compared two different dimensionalities (1D vs. 2D) of ISODATA signature vector in ISODATA parameter space to segment evolving PWI/DWI mismatch. We showed that accurate delineation of PWI/DWI mismatch with 1D signature vector outperformed conventional 2D analysis, and offer a useful means to identify ischemic penumbra.

Geng Guangqiang¹, Lynne Bilston^1,2, Ralph Sinkus³, Roland Henry⁴, and Caroline Rae^1,5
1Neuroscience Research Australia, Sydney, NSW, Australia, ²Prince of Wales Clinical School, UNSW, Sydney, Australia, ³Centre de Recherches Biomédicales Bichat-Beaujon, Paris, France,⁴Departments of Radiology and Biomedical Imaging, Neurology, and Bioengineering Graduate Group, University of California, San Francisco, United States, ⁵UNSW, Syndey, Australia

Magnetic resonance elastography (MRE) measures the mechanical properties of variable biological tissues in vivo. Understanding of error propagation through the complex MRE reconstruction algorithms is necessary to determine the uncertainty of calculated elasticity values and the impact of different acquisition and modelling methods. We have quantified the accuracy and precision in simulated and real human brain MRE data through Monte Carlo simulations and wild bootstrap algorithm. The RMSE are low both with homoscedastic and heteroscedastic noise, which indicates high accuracy and precision of MRE reconstruction. Bootstrapping techniques are a promising method for evaluating reliability of MRE estimates of elasticity.

Christian G. Graff¹, and Kyle J. Myers^1
1Division of Imaging and Applied Mathematics, U. S. Food and Drug Administration, Silver Spring, MD, United States

We develop a metric for image quality assessment based on task performance using the Ideal (Bayesian) Observer SNR. The Ideal Observer is an upper bound for human performance and measures the ability to perform a task (in this case lesion detection) using statistical decision theory. Ideal Observer performance can be assessed from raw data or image reconstructions to measure the effectiveness of both the data acquisition and image reconstruction. This has advantages over other metrics such as SNR and CNR which are only weakly correlated with a readers ability to perform clinical tasks.

Don C Bigler¹, Megan Taylor Sutton¹, and Gregory J Moore^2
1Center for Emerging Neurotechnology and Imaging, Penn State Hershey Neuroscience Institute, Hershey, PA, United States, ²Radiology, Geisinger Medical Center, Danville, PA, United States

Localization of the anterior and posterior commissures is an important pre-processing step for many MR brain post-processing procedures. For large image datasets automatic localization is crucial. This work presents a method for automatic anterior and posterior commissure localization using artificial neural networks.

Maria Ida Iacono¹, Nikos Makris¹, Luca Mainardi², John Gale³, Andre Van der Kouwe¹, Azma Mareyam¹, Jonathan R Polimeni¹, Lawrence L Wald¹, Bruce Fischl¹, Emad N Eskandar⁴, and Giorgio Bonmassar^1
1Dept. of Radiology,MGH, A. A. Martinos Center for Biomedical Imaging, Charlestow, MA, United States, ²Bioengineering Department, Politecnico di Milano, ³Cleveland Clinic, Department of Neuroscience and Center for Neurological Restoration, Cleveland, Ohio, United States, ⁴Neurosurgery, Massachusetts General Hospital, Boston, MA, United States

Deep brain stimulation (DBS) of the globus pallidus internus (GPi) proves to significantly improve the motor symptoms of advanced PD. However, GPi is not clearly identifiable in pre-operative MRIs, used routinely in surgical DBS procedures. In this work we propose an atlas-based segmentation procedure based on a ultra-high resolution atlas to outline the GPi on the pre-operative low resolution images of PD patients. Our preliminary results show that a surface non-linear registration procedure based on multiple nuclei improves the accuracy in discriminating the target of interest.

Jean-Sebastien Raynaud¹, Antony Lee¹, Caroline Robic¹, Eric Giaccomini², Isabelle Raynal¹, Philippe Robert¹, Franck Lethimonier², Marc Port¹, and Claire Corot^1
1Guerbet Research, Paris, France, ²Neurospin, CEA, Saclay, France

During the last decade, different approaches were designed to detect USPIO. “Positive contrast” imaging based on a single off resonance RF excitation suffers of low sensitivity and miss registrated spatial localization. A way to avoid these drawbacks is the ORS sequence, based on a long RF saturation to affect off resonance water diffusible protons. The purpose of this study was to investigate a new sequence initially performed for CEST imaging, WASSR, to combined strong contrast, high sensitivity and quantification. This sequence was evaluated on Iron and Gadolinium based contrast agents in in vitro conditions at clinical and research fields.

Jean-Sebastien Raynaud¹, Antony Lee¹, Xavier Violas¹, Robin Santus¹, Gaelle Louin¹, Isabelle Raynal¹, Philippe Robert¹, Marc Port¹, and Claire Corot^1
1Guerbet Research, Paris, France

In this study, we purpose to adapt a modified ORS sequence, WASSR (WAter Saturation Shift Referencing), initially applied on CEST contrast, to investigate in vitro and in vivo detection of a targeted USPIO in a clinical field. Preliminary result on U87 mice tumors, shows the ability of the WASSR sequence to provide an original approach for in vivo USPIO imaging combining contrast, sensitivity, quantification and appropriate spatial resolution, in clinical considerations (clinical field and low SAR).

Sean Smart¹, Danny Allen¹, John Beech¹, Sally Hill¹, Veerle Kersemans¹, Lei Zhao¹, and Ruth Muschel^1
1Gray Institute, Oxford University, Oxford, Oxfordshire, United Kingdom

We describe a slice ordering scheme used in conjuntion with cardio-respiratory triggering that is used to generate T2 weighted images of liver that are free of motion artefact. The scheme allows high resolution imaging of the whole liver in the mouse within an acceptable scan time.

Mohamed Tachrount¹, Virginie Callot¹, Patrick J. Cozzone¹, and Guillaume Duhamel^1
1CRMBM / CNRS 6612, Faculté de Médecine, Université de la Méditerranée, Marseille, France

Fast imaging sequences (e.g. EPI) are required for small animal studies which include multimodal approaches. With the increasing fields, EPI may not be applicable due to the increase of image distortions or strong susceptibility artifacts. Spiral imaging could be a good alternative as it offers similar acquisition speed performance with less sensitivity to susceptibility artifacts and motion. However, its successful application can be challenging because of B0 inhomogeneities and gradients imperfections. This works presents the investigation of spiral imaging feasibility at very high field (11.75T) for mouse brain imaging and, its application for brain perfusion study perfomed with ASL.