Room 255 EF

13:30 - 15:30

Moderators:

James C. Gee, Joseph V. Hajnal

Weiwei Chen¹, Susan A. Gauthier², Ajay Gupta², Joseph Comunale², Tian Liu², Shuai Wang³, Mengchao Pei², David Pitt⁴, and Yi Wang^2
1Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China, ²Weill Cornell Medical College, New York, NY, United States, ³School of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, Sichuan, China, ⁴The Ohio State University, Columbus, Ohio, United States

A total of 293 MS lesions were detected in 12 clinically definite MS patients who underwent twice quantitative susceptibility mapping (QSM) from 8/2011 to 5/2012. This longitudinal study of susceptibilities of MS lesions of different ages using QSM suggests the following findings. 1) There are 6 patterns of lesions manifested in MRI, with various patterns in individual patients. 2) QSM can detect lesions that are not detectable on conventional MRI (pattern Q). 3) Susceptibilities of MS lesions increase at early stage, peak in 1-3 yrs, subsequently decrease in 3-7 yrs, and return to normal after 7 yrs.

Chunlei Liu¹ and Wei Li^1
1Brain Imaging and Analysis Center, Duke University, Durham, NC, United States

A method was demonstrated to extract sub-voxel tissue magnetic response in vivo. A set of magnetic multipoles were obtained by analyzing a single volume of gradient-echo images in the spectral space (p-space). An algorithm was developed for multipole analysis of images acquired with multiple coils. A number of unique findings were described: 1) multipole response of white matter exhibits strong dependence on p-value; 2) multipole response of white matter is anisotropic; 3) multipole tensors provide distinctive contrasts and 4) are indicative of tissue microstructure. Most importantly, this unique information can be obtained from a single volume of GRE images.

Serena Y. Yeung¹, Kie Tae Kwon¹, Bob S. Hu^1,2, and Dwight G. Nishimura^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Palo Alto Medical Foundation, Palo Alto, CA, United States

Magnetization-prepared 3D SSFP sequences have shown promise for non-contrast-enhanced flow-independent angiography (FIA), where intrinsic tissue parameters such as T1, T2, and chemical shifts are exploited to generate stable vessel contrast even under slow flow conditions. However, an important challenge with this approach is sufficient artery-vein contrast, which is crucial for artery visualization and the diagnosis of arterial disease. In this work, we apply the Maximally Stable Extremal Regions (MSER) detector and k-means clustering to perform unsupervised segmentation and removal of the femoral veins in 3D FIA datasets of the lower extremities.

Andreas Deistung¹, Ferdinand Schweser², and Jürgen R. Reichenbach^2
1Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany, ²Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University, Jena, Germany

We present a supervised approach for creating image contrast that enhances specific structures of interest. To this end, we apply linear discriminant analysis (LDA) to optimally combine multiple image contrasts generated from complex-valued multi-echo GRE information (magnetic susceptibility, R2*, and S0) to create a novel contrast with increased cerebral venous vessel contrast. This approach can easily be adjusted to emphasize other tissue properties by changing the definition of classes in the training step.

Robert Grimm¹, Li Feng², Christoph Forman³, Jana Hutter¹, Berthold Kiefer⁴, Joachim Hornegger¹, and Kai Tobias Block^5
1Pattern Recognition Lab, FAU Erlangen-Nuremberg, Erlangen, Germany, ²Department of Radiology, New York University Langone Medical Center, New York City, NY, United States, ³Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nürnberg, Erlangen, Germany, ⁴Siemens Healthcare, Erlangen, Germany, ⁵Department of Radiology, NYU Langone Medical Center, New York City, NY, United States

Compressed Sensing reconstruction of DCE-MRI with radial stack-of-stars GRE acquisition suffers from two shortcomings: First, the acquisition cannot be combined with conventional bolus-detection techniques to confirm successful bolus administration. Second, in abdominal examinations only few of the up to 100 reconstructed volumes are relevant for clinical diagnosis. An automatic, k-space based method is presented that addresses both problems. A bolus signal reflecting the course of global contrast enhancement is extracted and used to accurately determine the bolus arrival time. Using population-based estimates for the delay of arterial and venous enhancement in the liver, the corresponding images can be identified.

Bo Xu¹, Sam Tilsen², Pascal Spincemaille³, Madhur Srivastava², Peter Doerschuk², and Yi Wang^1
1Biomedical Engineering, Cornell University, Ithaca, New York, United States, ²Cornell University, Ithaca, New York, United States, ³Weill Cornell Medical College, New York, New York, United States

In this study, the real-time spiral MRI combined with the TRACER reconstruction is used for articulation analysis. This method maintains coverage and spatial resolution but achieves a high temporal frame rate that is shown to be advantageous for the investigation of speech motor control.

Alexandra Tobisch¹ and Hui Zhang^2
1Department of Medical Physics and Bioengineering, University College London, London, United Kingdom, ²Department of Computer Science and Centre for Medical Image Computing, University College London, London, United Kingdom

This work develops a super-resolution reconstruction (SRR) technique that constructs isotropic high-resolution diffusion-weighted images (DWI) from multiple anisotropic low-resolution acquisitions. The technique will enable the mapping of tissue microstructure for fine brain structures without the need for prohibitively long imaging time. It advances the state-of-the-art by adopting a model-based approach to directly super-resolve the parameter maps of the underlying tissue microstructure.

Brian Hansen¹, Torben E. Lund¹, Ryan Sangill¹, and Sune N. Jespersen^1,2
1CFIN, Aarhus University, Aarhus C, Denmark, ²Department of Physics, Aarhus University, Aarhus C, Denmark

Diffusion kurtosis imaging is a popular extension of diffusion tensor imaging accounting for nongaussian aspects of diffusion in biological tissue. Recently, several studies have indicated enhanced sensitivity of mean kurtosis (MK) to pathology, including stroke. However, lengthy acquisition time and postprocessing remains an obstacle for the exploration of further clinical applications. Here we propose a very fast acquisition and postprocessing scheme based on a linear combination of 13 diffusion weighted images for estimation of a new mean kurtosis metric, the trace of the kurtosis tensor, which is then shown to have very similar contrast to MK in the human brain.

Chantal M.W. Tax¹, Willem M. Otte¹, Max A. Viergever¹, Rick M. Dijkhuizen¹, and Alexander Leemans^1
1Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands

Diffusion kurtosis imaging (DKI) provides new avenues for an accurate and complete tissue characterization within clinically feasible scanning times. In a clinical setting, however, such benefits are often nullified by numerous acquisition artifacts. In this work, we propose to extend the popular Robust Estimation of Tensors by Outlier Rejection (RESTORE) approach, which is widely used in diffusion tensor imaging (DTI), to DKI. In addition, a linearized framework, coined REKINDLE (Robust Extraction of Kurtosis INDices with Linear Estimation), has been developed that drastically reduces the computational cost without compromising the estimation reliability.

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

A computational modeling framework has been developed which is able to analyze and compare image quality across different sequences, trajectories and reconstruction techniques. The image quality metrics are based on practical analysis tasks which emulate the complex uses of clinical MR. Using these metrics we show how even complex reconstruction methods such as compressed sensing can be analyzed in a rigorous manner, which is not possible with traditional image quality metrics.