Time

Prog #

Guozhi Tao¹, Renjie He¹, Sushmita Datta¹, Ponnada A. Narayana¹

¹University of Texas Medical School at Houston, Houston, Texas, USA

A diffeomorphic registration technique that explicitly includes inverse consistent constraint is developed and applied on normal human brain. The maximum inverse consistency error with our method is less than 0.07 voxels. This is ten times smaller than the recently reported values.

Neculai Archip¹, Ion-Florin Talos¹

¹Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA

This paper presents the development of a freely accessible, on-line repository of pre-and intra-operative MR-images derived from patients with hemispheric brain tumors, to serve as a validation platform for non-rigid registration algorithms.

George Iordanescu¹, ², Palamadai Venkatasubramanian¹, ², Alice M. Wyrwicz¹, ²

¹ENH Research Institute, Evanston, Illinois, USA; ²Northwestern University Feinberg School of Medicine, Evanston, Illinois, USA

We present a novel automatic method based on simulated flooding and data Laplacian to segment the AD plaques in mouse MR brain images, and show how the method can be used for a detailed analysis of plaque characteristics. The segmentation map can be used to asses individual plaques (average profile, volume, intensity variability grade), to compute the plaques load or the plaques distribution for specific brain structures. Since no assumptions are made on the plaques shape or size, the proposed method can be used to analyze data for different AD stages or mouse strains, where plaques variability should be expected.

Wilbur C. K. Wong¹, David Johnson², Chris Flask³, Paul Ernsberger², David Wilson², ³

¹The Hong Kong University of Science & Technology, Kowloon, Hong Kong; ²Case Western Reserve University, Cleveland, Ohio, USA; ³University Hospitals of Cleveland, Cleveland, Ohio, USA

We are using MR to quantify fat depots (e.g., visceral, subcutaneous, hepatic, muscular) so as to determine the role of genetic, environmental, and therapeutic factors on lipid accumulation, metabolism, and disease states. In this report, we studied lean spontaneously hypertensive rats (SHRs), a genetic variant prone to obesity (SHROBs), and animals given a high fat, high sucrose diet creating dietary obese animals (SHR-DOs). Animals were imaged with and without CHESS water-suppression. Ratio images exactly compensate for receive coil sensitivity inhomogeneity and enable the creation of gray-scale based automated analysis. A ratio image model was created for measuring lipid content in subcutaneous and visceral depots. We analyzed the statistical property of the ratio of two noisy signals and developed a maximum a posteriori (MAP) estimate of the lipid content in each voxel. We simulated the partial volume effect in a digital phantom with known fat content in each voxel. Even if relaxivities were corrected, ratio images overestimated the true volume of fat by 23%. In a cohort of rats, the MAP correction reduced visceral adipose tissue volume by 20%. We identified obesity phenotypes and characterized this model of metabolic syndrome.

Niranjan Balu¹, Vasily Yarnykh¹, William Kerwin¹, Jianming Cai², Chun Yuan¹

¹University of Washington, Seattle, USA; ²Chinese PLA General Hospital, Beijing, People's Republic of China

Contrast enhancement (CE) patterns of the carotid atherosclerotic plaque are indicative of plaque neovasculature and inflammation. Therefore quantitative CE information from black-blood CE-MRI may provide additional markers of plaque vulnerability. We demonstrate a fast, semi-automated method to obtain quantitative CE maps (QCEM) of the carotid plaque in-vivo. Luminal surface, adventitial surface, plaque shoulder and core subregions were automatically segmented for their respective QCEM. Group histograms and histogram statistics from subregion QCEM of 24 patients were compared between plaques of high and low plaque burden. Larger plaque locations showed significantly higher core and shoulder enhancement indicating possible inflammation/neovasculature in these regions.

Luca Antiga¹, David Andrew Steinman²

¹Mario Negri Institute for Pharmacological Research, Bergamo, Italy; ²University of Toronto, Toronto, Canada

MRI simulation of anatomically realistic objects can be inefficient owing to the need to discretize these objects volumetrically.  Here we present an efficient simulation technique based on integrating the MRI signal equation over a surface rather than volume discretization.

Monika Gloor¹, Klaus Scheffler¹, Oliver Bieri¹

¹University Hospital / University of Basel, Basel, Switzerland

In tissues, the signal of balanced steady-state free precession (bSSFP) is considerably reduced from magnetization transfer (MT). An extended bSSFP signal equation is derived, based on a binary spin-bath model including MT effects. Using this new bSSFP signal model, quantitative MT model parameters such as the fractional pool size, corresponding magnetization exchange rates, and relaxation times are estimated in human brain. The results show high correlation with the ones from standard methods, but benefit from bSSFP’s short acquisition times and high signal-to-noise ratios. This allows the acquisition of isotropic high resolution quantitative MT maps within clinically feasible acquisition times.

James D. Quirk¹, Alexander L. Sukstanskii¹, G Larry Bretthorst¹, Dmitriy A. Yablonskiy¹, ²

¹Washington University School of Medicine, St. Louis, Missouri, USA; ²Washington University, St. Louis, Missouri, USA

We demonstrate that joint Bayesian analysis of array coil data offers a "worry-free" method for obtaining optimal estimates of sample parameters.  A theory predicting the parameter estimate uncertainty was developed and validated on simulated data.  Comparisons between joint analysis and more traditional coil combination methods (e.g. sensitivity weighted, SOS) indicate that estimates from joint analysis have equivalent or superior precision and accuracy on phased and magnitude data, without the use of reference scans.  Combining channels without the optimal weighting factor (signal amplitude over noise power) can corrupt the parameter estimation and produce inferior results to a single channel analysis.

Christian Graff¹, Eric Clarkson¹, Eric K. Outwater¹, Maria I. Altbach¹

¹University of Arizona, Tucson, Arizona , USA

Computer-generated phantoms provide a convenient way to evaluate MR parameter estimation algorithms.  However, if the phantom data is significantly different from in vivo data, results from phantom studies may not be predictive of performance in vivo.  We have constructed a computer-generated abdomen phantom which contains realistic models for parameter variability, coil sensitivities and other practical effects.

Anita Banerji¹, Angela Caunce¹, Yvon Watson¹, Chris Rose¹, Giovanni Buonaccorsi¹, Geoff Parker¹

¹The University of Manchester, Manchester, UK

The software phantom presented in this work generates synthetic images from known ground truth for use in the validation of post-acquisition image processing algorithms, for example registration techniques for DCE-MR time series. The design of the phantom is flexible allowing images for various different imaging scenarios and modalities to be produced. A range of synthetic data sets can be generated that are relevant to the image processing algorithm in order to test robustness, accuracy and precision. Example DCE-MR images with realistic anatomy based on in-vivo data sets are shown.