IntellMR 2014
Home Page Image
** Extended deadline **
June 29, 2014 (midnight PST):
Paper Submission

July 9, 2014:
Notification of Acceptance

July 18, 2014:
Camera-Ready Papers
September 18, 2014:



Turning Nuclear Magnetic Resonance (NMR) into Magnetic Resonance Imaging (MRI):
What a Long Strange Trip It’s Been
Robert V. Mulkern
Department of Radiology
Children's Hospital
Harvard Medical School
Boston , MA

The dawn of NMR occurred circa the 1940’s as scientists learned that, by placing a sample of nuclear spins in a static magnetic field and then manipulating the orientation of the nuclear spins with an appropriate radio-frequency (RF) field, the spins would then send out a signal which could be received and studied. Specific properties of the signal revealed much about the microenvironment within which the nuclear spins lived and for forty years NMR was used by physicists and chemists to study the chemical composition and various microstructural and dynamical aspects of materials in a manner hitherto unheard of, advancing humankind’s knowledge of said materials dramatically. Biological materials were also studied with the technique and, circa the late 70’s, several investigators began to realize that with the addition of magnetic field gradients to the basic NMR equipment (magnet, RF transmitters and receivers), images of the object under scrutiny could be made, particularly using the proton signals from the abundant water and lipid molecules in biological materials, like us. Thus MRI was born and has grown enormously since then, becoming a multi-billion dollar industry indispensable to the medical profession for providing unparalleled details of our inner workings – safely and without ionizing radiation. Here we review the basic physics of the NMR experiment from Field to Sample to Signal and then capture the essence of imaging, MRI, as discovered in the late 70’s by considering how field gradients applied across the sample can endow the signal with the spatial information needed to generate magnetic resonance images, a medical miracle indeed.


MRI reconstruction: Parallel imaging and Simultaneous MultiSlice acquisition
and their application to Connectomics and beyond.
Kawin Setsompop
Department of Radiology
Harvard Medical School

This educational talk will provide a basic explanation to how images are created from MR signals through the use of Fourier encoding. A portion of this talk will focus on parallel imaging, a popular and widely used technique to accelerate MR acquisitions. Such technique achieved acquisition time saving by acquiring undersampled Fourier data and make use of the spatial encoding information in the multiple receiver coils to aid the image reconstruction. In addition to traditional parallel imaging, this talk will also cover Simultaneous Multislice (SMS aka Multiband) acquisition, a parallel imaging technique that has gained much interest in recent years due to its ability to help increase the temporal sampling of fMRI by an order of magnitude and speed up diffusion imaging by a factor of 3-4x. Controlled Aliasing techniques that have been keyed to this technology will be covered and application of SMS to data acquisition in Connectomic (fMRI and diffusion), perfusion, anatomical imaging, phase contrast, and other area of interest will be discussed.



Reduction of motion-related artifacts in MRI data using motion-insensitive acquisition
and advanced post- processing algorithms.
Nan-kuei Chen
Duke University

In the past two decades, numerous efforts have been invested to improve motion tolerance of MRI. However, even for cooperative patients, unavoidable motion-related artifacts often limit the MRI image quality and diagnostic value. Therefore, there is still a strong need for further developing techniques to effectively eliminate motion-related artifact, which is the major bottleneck in achieving high-quality clinical MRI for challenging patient populations such as children, tremor-dominant Parkinson’s patients and seriously ill patients among others.

First, I will discuss conventional motion artifact correction methods and their limitations: 1) The use of embedded low-resolution navigator echoes may reduce rigid body motion induced aliasing artifact in structural MRI. The local and nonlinear motion, however, may not be accurately measured and corrected by the navigator echoes of low spatial resolution; 2) The respiratory-gating and breath-holding may be used to reduce motion artifacts in abdominal MRI. However, the image quality may still be degraded in gated acquisition when the respiratory frequency changes significantly over time (e.g., in older adults; seriously ill patients), and the breath-holding is impractical for many patients; 3) The motion- related artifacts can generally be reduced with alternative imaging strategies such as PROPELLER or radially-sampled imaging. However, the residual artifacts resulting from motion of challenging patterns (e.g., head tremor at 3 Hz in Parkinson’s patients) may still be pronounced; 4) It has been shown that a real-time slice prescription may be implemented to reduce motion-related artifacts. However, this approach (with delay time of at least a TR) can only correct for infrequent subject motion, and may not be effective in addressing continual motion (e.g., free-breathing abdominal imaging; neuro imaging in the presence of head tremor); 5) It is possible to significantly reduce motion artifacts, even for challenging subject populations, by using a single-shot EPI pulse sequence. However, the image quality and anatomic resolvability are limited in single- shot EPI, which is not well suited for clinical structural imaging that requires high spatial-resolution and fidelity.

Second, I will discuss new and promising approaches that may be further integrated to suppress MRI motion artifacts: 1) Using the RF coil sensitivity profiles as the constraint in image reconstruction, motion related artifacts may be effectively reduced even for challenging applications (e.g., interleaved diffusion-weighted MRI in the presence of subject motion); 2) Using the repeated k-t-domain acquisition and bootstrapping reconstruction, motion related artifacts can be suppressed for various types of subject motion (e.g., free-breathing abdominal imaging; head tremor at 3 Hz in Parkinson’s patients).

Third, I will discuss how motion-insensitive MRI pulse sequences and advanced post-processing algorithms may be integrated to further improve the motion tolerance of MRI.


Susceptibility Weighted Data Acquisition and Processing
Berkin Bilgic
Harvard Medical School
Massachusetts General Hospital
Martinos Center for Biomedical Imaging

Gradient echo (GRE) phase imaging has demonstrated the ability to provide exquisite anatomical contrast complementary to the magnitude signal, and reflects the local changes in the main magnetic field. Susceptibility Weighted Imaging (SWI) employs this phase information to further enhance the T2* weighted image contrast, and offers complementary information valuable in the diagnosis and treatment of neurovascular and neurodegenerative disorders (1). The local phase variation utilized in SWI is incurred by the tissue magnetic susceptibility distribution, which is related to the observed phase via an ill-posed convolution operation. Quantitative Susceptibility Mapping (QSM) estimates this underlying susceptibility map from the tissue phase through deconvolution, and yields critical information about the tissue iron concentration and blood oxygenation fraction in vessels. This allowed QSM to find recent application in probing and monitoring of neurodegenerative diseases such as multiple sclerosis (2) and Alzheimer’s (3), as well as mapping of oxygen saturation along cerebral venous vasculature (4). This tutorial will introduce data acquisition and processing methods that span a pipeline starting from raw k-space data and ending with susceptibility maps. Data acquisition will focus on acquisition parameter setting for optimal phase contrast and rapid susceptibility imaging using the efficient spiral (5), echo-planar (6) and Wave-CAIPI (7) trajectories with parallel imaging. The processing side will begin with phase-sensitive coil combination strategies and phase unwrapping, and will include background phase removal using harmonic filtering and dipole deconvolution methods for susceptibility mapping. The tutorial will conclude with and introduction to the novel modalities of multi-orientation QSM (8) and Susceptibility Tensor Imaging (STI) (9).

1. Mittal S, Wu Z, Neelavalli J, Haacke EM. Susceptibility-weighted imaging: technical aspects and clinical applications, part 2. Am. J. Neuroradiol. 2009;30:232–252.
2. Langkammer C, Liu T, Khalil M, Enzinger C, Jehna M, Fuchs S, Fazekas F, Wang Y, Ropele S. Quantitative Susceptibility Mapping in Multiple Sclerosis. Radiology 2013;267:551–9. doi: 10.1148/radiol.12120707.
3. Acosta-Cabronero J, Williams G. In Vivo Quantitative Susceptibility Mapping (QSM) in Alzheimer’s Disease. PLoS One 2013.
4. Fan AP, Bilgic B, Gagnon L, Witzel T, Bhat H, Rosen BR, Adalsteinsson E. Quantitative oxygenation venography from MRI phase. Magn. Reson. Med. 2014;72:149–59. doi: 10.1002/mrm.24918.
5. Wu B, Li W, Avram AV, Gho S-M, Liu C. Fast and tissue-optimized mapping of magnetic susceptibility and T2* with multi-echo and multi-shot spirals. Neuroimage 2012;59:297–305.
6. Poser B, Koopmans P, Witzel T, Wald L, Barth M. Three dimensional echo-planar imaging at 7 Tesla. Neuroimage 2010;51:261–266.
7. Bilgic B, Gagoski B, Cauley S, Fan AP, Polimeni J, Grant P, Wald L, Setsompop K. Wave-CAIPI for highly accelerated 3D imaging. Magn. Reson. Med. 2014. doi: 10.1002/mrm.25347.
8. Liu T, Spincemaille P, de Rochefort L, Kressler B, Wang Y. Calculation of susceptibility through multiple orientation sampling (COSMOS): a method for conditioning the inverse problem from measured magnetic field map to susceptibility source image in MRI. Magn. Reson. Med. 2009;61:196–204.
9. Liu C. Susceptibility tensor imaging. Magn. Reson. Med. 2010;63:1471–1477.


Accelerated Imaging and Compressed Sensing in structural and functional imaging
Ashish Raj
Image Data Evaluation and Analytics Lab (IDEAL)
Weill Cornell Medical College

Due to slow acquisition speed, MR imaging can limit clinical applications. One way to speed up acquisition is to deliberately under-sample k-space. However, naïve under sampling leads to severe aliasing artifacts, whose removal is essential. The inferring of missing data from under sampled k-space acquisitions is one of the main goals of advanced accelerated imaging reconstruction algorithms [1]. This difficult problem has attracted a lot of algorithmic firepower in the last decade, and now several algorithms are available, all of which assume some sort of prior knowledge about the image. Each prior assumption leads to a constrained minimization problem, which is usually computationally challenging due to large problem size and high ill-posedness. We will describe the MR imaging problem and present the classical reconstruction using the SENSE method [1,2]. Next, we will discuss recently popular Compressive Sensing (CS) based methods, which employ sparsity constraints and convex optimization routines [3]. We will then discuss a new approach called graphcuts, which can solve non-convex minimization problems resulting from desirable but challenging piecewise smooth prior constraints [4]. We will conclude with some examples of compressed sensing in structural and functional MRI.

[1] K. P. Pruessmann, M. Weiger, M. B. Scheidegger, and P. Boesiger, “SENSE: sensitivity encoding for fast MRI.,” Magnetic resonance in medicine, vol. 42, no. 5, pp. 952-62, Nov. 1999
[2] K. P. Pruessmann, M. Weiger, P. Börnert, and P. Boesiger, “Advances in sensitivity encoding with arbitrary k-space trajectories.,” Magnetic resonance in medicine, vol. 46, no. 4, pp. 638-51, Oct. 2001
[3] M. Lustig, D. Donoho, and J. M. Pauly, “Sparse MRI: The application of compressed sensing for rapid MR imaging.,” Magnetic resonance in medicine?: official journal of the Society of Magnetic Resonance in Medicine / Society of Magnetic Resonance in Medicine, vol. 58, no. 6, pp. 1182-95, Dec. 2007
[4] A. Raj et al., “Bayesian parallel imaging with edge-preserving priors.,” Magnetic resonance in medicine, vol. 57, no. 1, pp. 8-21, Jan. 2007


Simultaneous Estimation of T1, T2 and B1 Maps From Relaxometry MR Sequences
Fang Cao, Olivier Commowick, Elise Bannier, Christian Barillot

Interest in quantitative MRI and relaxometry imaging is rapidly increasing to enable the discovery of new MRI disease imaging biomarkers. While DESPOT1 is a robust method for rapid whole-brain voxel-wise mapping of the longitudinal relaxation time (T1), the approach is inherently sensitive to inaccuracies in the transmitted flip angles, de- fined by the B1 inhomogeneity field, which become more severe at high field strengths (e.g., 3T). We propose a new approach for simultaneously mapping the B1 field, M0 (proton density), T1 and T2 relaxation times based on regular fast T1 and T2 relaxometry sequences. The new method is based on the intrinsic correlation between the T1 and T2 relaxometry sequences to jointly estimate all maps. It requires no additional sequence for the B1 correction. We evaluated our proposed algorithm on simulated and in-vivo data at 3T, demonstrating its improved accuracy with re- spect to regular separate estimation methods.


Simultaneous Registration and Bilateral Differential Bias Correction in Brain MRI
Bin Zou, Akshay Pai, Lauge Sørensen, and Mads Nielsen
University of Copenhagen and Biomediq A/S, Copenhagen, Denmark

Differential bias correction is an important tool when simultaneous assessments of longitudinal scans are made. Among the existing methods, the bias model is applied to only one of the images. This may lead to inconsistent atrophy estimation depending on which image it is applied to. In this paper, we propose a B-spline free-form deformation based two-image differential bias correction method where both images in the registration process are corrected for bias simultaneously. Further, symmetry in bias correction is achieved via a new regularization term. In a simulated experiment, reproducibility of atrophy measurements in a single-image bias correction method largely depended on the choice of the image that was corrected while this choice did not matter with our proposed two-image bias correction method. On Alzheimer’s disease neuroimaging initiative data, the two-image bias correction method performed superior when compared to registration of separately bias cor- rected images.


A System Identification Approach to Estimating a Dynamic Model of Head Motion for MRI Motion Correction
Burak Erem, Onur Afacan1, Ali Gholipour, Sanjay P. Prabhu, and Simon K. Warfield
Computational Radiology Laboratory and Advanced Image Analysis Lab, Radiology Department, Boston Children’s Hospital

Motion-compensated MRI is a promising technique to mitigate the effects of motion on MRI. This works best when accurate real- time motion measurements are available. Many measurement techniques track motion with a delay and produce noisy measurements. We propose an estimator that uses a dynamic system identification approach to estimate rigid body head motion from concurrent measurements of position and orientation, which can be used to predict motion shortly into the future. We compare our method to static estimates and a Kalman filter-based method in our experiments, in which we evaluate the effects of motion using real and simulated tracking data.


Parallel imaging for aligned multishot MR reconstruction
Lucilio Cordero-Grande, Emer Hughes, Rui Pedro A. G. Teixeira, Anthony Price, and Joseph V. Hajnal?
King’s College London, London, UK

We propose a method to retrospectively correct for rigid within plane motion in multishot MR images. The rationale behind our proposal is to make use of the information provided by multiple receiver coils in order to estimate the position of the imaged object throughout the acquisition. The estimated motion is incorporated into the reconstruction model in an iterative manner to obtain a motion-free image. The method is parameter-free, uses no prior model for the image to be reconstructed, avoids blurred images due to resampling, does not make use of external sensors, and does not introduce modifications in the acquisition sequence. Results on synthetic data have shown that we are able to fully recover highly motion corrupted brain images on customary acquisition protocol settings. Additionally, the method is robust to noise, the use of parallel imaging acceleration, and the presence of some other sources of motion. Results on its application to neonatal MR brain imaging have shown a general improvement of the quality of the reconstructed images without any observed adverse side effect.