yigang pei¹, and daoyu hu^2
1Department of Radiology, Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology, wuhan, Hubei, China, People's Republic of, ²1Department of Radiology, Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology

The recipient vessels after orthotopic liver transplantation (OLT) are often evaluated by CT and MRI, but the recipient vessels either may be contaminated by other vessel or may not be presented distinctly due to the individual hemodynamics changes. In addition, Gd or iodine contrast media are applied on CT and MRI for displaying the recipient vessel which may cause nephrogenic systemic fibrosis (NSF) or contrast-Induced Nephropathy. Therefore, a non-contrast magnetic resonance angiography (NC-MRA) is desirable to develop for presenting the recipient vessels. In this study, a new NC-MRA (Spatial LabEling with multiple invErsion pulses, SLEEK) has been applied to delineate the recipient vessel after OLT.

Pauline Wong Worters¹, Manojkumar Saranathan¹, Alan Xu¹, and Shreyas Vasanawala^1
1Stanford University, Stanford, CA, United States

A sequence is developed to provide high spatial resolution renal angiograms within breath-holding times. Balanced SSFP has good flow properties, and high T2/T1 contrast and SNR efficiency. IR preparation provides inflow sensitivity, while Dixon technique allows for excellent fat suppression. An efficient k-space trajectory enabled acquisition in 20-30 seconds. We evaluated IR-prep, Dixon bSSFP by comparing it to a commercially available, respiratory-gated technique Inhance IFIR (2-5 min) in 9 patients at 3T. No significant difference in artery visualization or image quality was found between Inhance IFIR and ECG-gated, breath-hold IR-prep, Dixon bSSFP. We have shown a robust sequence that provides diagnostic-quality renal angiograms comparable to a commercially available sequence, with significantly shorter scan times and without compromising spatial coverage and resolution.

Mehdi Moradi¹, Septimiu E Salcudean¹, Silvia D Chang², Edward C Jones³, S Larry Goldenberg^4,5, and Piotr Kozlowski^2,6
1Electrical and Computer Engineering, University of British Columbia, Vancouver, BC, Canada, ²Radiology, University of British Columbia, ³Pathology and Laboratory Medicine, University of British Columbia, ⁴Urologic Sciences, University of British Columbia, ⁵Vancouver Prostate Centre, University of British Columbia, ⁶MRI Research Centre, University of British Columbia, Vancouver, Canada

We use five parameters extracted from Diffusion Tensor Imaging (DTI) and Dynamic Contrast Enhanced (DCE) MRI for prostate cancer detection. 29 patients were involved. The method is based on support vector machine classification and calculation of posterior class probabilities. These cancer probability estimates are used for creating cancer maps validated based on histopathologic analysis of biopsy samples. We also found a correlation between the proposed measure of cancer probability and the Gleason grade of the tumors. The average probability value was 0.555 for tumors of grade 3+3, 0.778 for tumors of grades 3+4 and 4+3, and 0.963 for grade 4+5.

Sang Ho Lee¹, Jong Hyo Kim^1,2, Jeong Yeon Cho^2,3, Sang Youn Kim^2,3, In Chan Song³, Hyeon Jin Kim³, and Seung Hyup Kim^2,3
1Interdisciplinary Program in Radiation Applied Life Science, Seoul National University College of Medicine, Seoul, Korea, Republic of, ²Department of Radiology, Seoul National University College of Medicine, Seoul, Korea, Republic of, ³Department of Radiology, Seoul National University Hospital, Seoul, Korea, Republic of

DCE-MRI plays an essential role for cancer detection and characterization. The significant tissue parameters for lesion classification may take the form of synergistic subsets in the combinatorial space of multifarious kinetic features. In this study, we postulate that extending the individual analysis schemes of contrast enhancement kinetics to a hybrid analysis scheme and that selecting a meaningful feature subset from a combined feature pool may allow an improved performance for lesion classification. Based on this postulation, we presented a novel approach to prostate MRI computer-aided diagnosis (CAD) using multifarious kinetic parameters in DCE-MR images.

Pallavi Tiwari¹, John Kurhanewicz², and Anant Madabhushi^1
1Biomedical Engineering, Rutgers University, Piscataway, NJ, United States, ²Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, United States

In this work we present a novel multi-protocol MRI classifier, semi-supervised multi-kernel (SeSMiK), for quantitatively combining features from T2-w magnetic resonance (MR) imaging (T2 -w MRI) and MR spectroscopy (MRS) data to identify high grade prostate cancer (CaP) in vivo. High grade CaP is known to be correlated with more biologically aggressive prostate cancer; low grade CaP is typically associated with indolent disease. A computerized decision support (CDS) tool that can distinguish high grade from low grade prostate cancer in vivo could help identify those patients who might benefit from a “wait and watch policy” as opposed to those who might be better suited to application of more aggressive treatment strategies. The objective of this work is to build a CDS system that can distinguish high from low grade CaP in vivo, on a per voxel basis, using quantitative integration of T2 -w MRI and MRS. The SeSMiK strategy leverages multi-kernel learning (MKL) and dimensionality reduction (DR) to provide a unified framework for quantitative integration of T2-w MRI and MRS. Texture and metabolic features are extracted from T2-w MRI and MRS respectively. Extracted features are transformed to a similarity kernel space and then combined using MKL. The combined T2-w MRI, MRS data is then reduced to a low dimensional space using a DR scheme. A probabilistic boosting tree classifier which individually evaluated (1) T2-w MRI texture features, (2) MRS metabolic features, and (3) combined low dimensional features obtained via SeSMiK, suggest a higher accuracy and area under the receiver operating curve from SeSMiK compared to the individual T2-w MRI,MRS modalities. Our results suggest that the SeSMiK classifier might be able to ultimately identify biologically aggressive CaP in vivo.

Namkug Kim¹, and JeongKon Kim^1
1Radiology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Seoul, Korea, Republic of

The prostate is anatomically composed of central, peripheral, and transitional zones. In the peripheral zone, 70% of prostate cancers arise. In addition, 20% of prostate cancers arise in the transitional zone. To exploit this tumor occurrence information, we evaluated accuracy enhancement for prostate tumor detection of automated classifier using deformable registration based atlas information as well as permeability parameters. In thirty seven patients with radical prostatectomy, MR images were obtained, including T2WI and dynamic contrast enhanced MR imaging for Brix permeability analysis. Each prostate was manually segmented into transitional zone and other zone by an expert radiologist and registered by FSL FNIRT deformable registration method. Sensitivity, specificity, accuracy, and AUC of ROC were significantly greater in automated classifier with atlas information (86.5¡¾0.02%, 95.5¡¾0.01%, 92.2¡¾1.08%, 92.9¡¾0.00 respectively) than in that without atlas information.

Satish Viswanath¹, B. Nicolas Bloch², Jonathan Chappelow¹, Pratik Patel¹, Neil Rofsky³, Robert Lenkinski⁴, Elisabeth Genega⁴, and Anant Madabhushi^1
1Biomedical Engineering, Rutgers University, Piscataway, NJ, United States, ²Boston Medical Center, ³UT Southwestern Medical Center, ⁴Beth Israel Deaconess Medical Center

We present a novel technique, Enhanced Multi-Protocol Analysis via Intelligent Supervised Embedding (EMPrAvISE), for building a computerized meta-classifier to predict the spatial extent of prostate cancer (CaP) in vivo via multi-protocol (T2-weighted, Dynamic Contrast Enhanced, Diffusion-weighted) MRI data. We employ automated registration, quantitative image descriptors, and a novel ensemble representation technique in our methodology. Evaluation of our automated predictions for spatial extent of CaP at a pixel level (against registered extents of CaP on MRI) results in EMPrAvISE showing a statistically significant improvement (AUC=0.73) over individual protocols (T2w, DCE,DWI), as well as simple multi-protocol feature concatenation.

Robert James Toth¹, B Nicholas Bloch², Elizabeth M Genega³, Neil M Rofsky³, Robert E Lenkinski³, Mark A Rosen⁴, and Anant Madabhushi^1
1Biomedical Engineering, Rutgers University, New Brunswick, NJ, United States, ²Boston Medical Center, Bostom, MA, United States, ³Beth Israel Deaconess Medical Center, Boston, MA, United States, ⁴Hospital at the University of Pennsylvania, Philadelphia, PA, United States

We present a method for using an advanced Active Shape Model based segmentation scheme to determine prostate volume on T2-weighted MR imagery. Our method was compared to the clinically standard ellipsoid technique and showed volume estimates more consistent with the ground truth volumes over 34 studies.

Aravinthan Jegatheesan¹, Michael D Noseworthy^1,2, Colm Boylan³, Robert Shayegan⁴, and John F MacGregor^5
1School of Biomedical Engineering, McMaster University, Hamilton, Ontario, Canada, ²Electrical and Computer Engineering, McMaster University, Hamilton, Ontario, Canada, ³St. Joseph's Healthcare, Hamilton, Ontario, Canada, ⁴Dept. of Urology, St. Joseph's Healthcare, Hamilton, Ontario, Canada, ⁵Chemical Engineering, McMaster University, Hamilton, Ontario, Canada

Principal component analysis was used to analyze BOLD-MRI of 5 patients with prostate tumors. The analysis attempts to identify the most significant temporal signatures to attempt to differentiate normal and tumor tissue. The results indicate that while the method maybe sensitive to a subset of tumors, it is not robust for detecting all tumors.

Gaoyu Xiao¹, B. Nicolas Bloch², Jonathan Chappelow¹, Elizabeth Genega³, Neil Rofsky³, Robert Lenkinski³, John Tomaszewski⁴, Michael Feldman⁴, Mark Rosen⁴, Arjun Kalyanpur⁵, and Anant Madabhushi^1
1Rutgers University, Piscataway, NJ, United States, ²Boston Medical Center, MA, USA, ³Beth Israel Deaconess Medical Center, MA, USA, ⁴University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA, USA., ⁵Teleradiology Solutions Pvt. Ltd. Whitefield, Bangalore, 560048, India

We present an automated computerized system to automatically determine slice correspondence between images from histology and MRI for the purpose of mapping spatial prostate cancer extent from wholemount histology slices to their corresponding in vivo T2 MRI slices. The mapping of prostate cancer extent is important in constructing computer-aided-diagnosis system and training radiology residents. The explicit determination of the histology-MRI slice correspondences is indispensable when an accurate 3D reconstruction of the histological volume cannot be achieved because of limited tissue slices with unknown inter-slice spacing, and histological image. The image slice correspondences obtained using our method were very close to the ground truth by medical experts.

Yingxuan Zhu¹, Ming-Ching Chang², Fiona M Fennessy³, and Sandeep Narendra Gupta^4
1Dept. of EECS, Syracuse University, Syracuse, NY, United States, ²Vis. and Comp. Vision Lab, GE Global Research Center, Niskayuna, NY, United States, ³Dept. of Radiology, Brigham and Women's Hospital, Boston, MA, United States, ⁴Functional Imaging Lab, GE Global Research Center, Niskayuna, NY, United States

Dynamic Contrast Enhanced MRI has shown promise in non-invasive assessment of tumor vascular properties for prostate cancer staging and treatment monitoring. Accurate quantification from DCEMRI requires the determination of arterial input function. AIF is commonly measured manually by selecting an ROI, which is time consuming and subjective. Automated ways of measuring AIF would be useful. This is challenging because of extreme intensity non-uniformity. Therefore existing methods do not perform well. Here, we use both temporal and spatial information to determine the AIF. We validate our method on clinical data and compare our approach with expert user defined manual AIFs.

Satish Viswanath¹, Jonathan Chappelow¹, Pallavi Tiwari¹, John Kurhanewicz², and Anant Madabhushi^1
1Biomedical Engineering, Rutgers University, Piscataway, NJ, United States, ²University of California, San Francisco

We present CADOnc©, a novel computerized segmentation, registration, and classification framework for quantifying changes in the prostate due to radiation therapy (RT) via multi-parametric MRI (Magnetic Resonance Spectroscopy, T2-weighted, Diffusion-weighted). On a small preliminary cohort of patient data, CADOnc© accurately quantified changes in specific multi-parametric MRI biomarkers, for both pre- and post-RT data. Further, quantitatively integrating these markers shows excellent utility in (1) predicting disease extent on pre-RT data, and (2) quantifying both residual and possible new foci of disease, post-RT.

Lawrence Patrick Panych¹, Renxin Chu¹, Yi Tang¹, Stephan E Maier¹, Clare M Tempany¹, and Robert V Mulkern^2
1Radiology, Brigham and Womens Hospital, Boston, MA, United States, ²Radiology, Children's Hospital, Boston, MA, United States

A reduced field-of-view single-shot fast-spin-echo sequence was implemented and applied for T2 quantification in the prostate. Six patients with biopsy-proven prostate cancer were imaged, acquiring whole-gland multiple-TE images in around one minute. Apparent T2 values were estimated from the multi-TE data for regions selected as being suspected healthy (SH) and suspected cancerous (SC). The apparent T2 in the SH regions was estimated at 224 +/- 67 msec and 129 +/- 25 msec in the SC regions. Sensitivity and specificity for detection of cancerous tissue is estimated to be 82% and 80% with this sequence.

Tokunori Kimura¹, and Mitsukazu Kamata^1
1MRI development department, Toshiba Medical Systems corp., Otawara, Tochigi, Japan

A real-part imaging for inversion recovery (real-IR) is known that clinically useful to provide higher T1W contrasts than the magnitude-based IR especially for brain. Fluid attenuated inversion recovery (FLAIR) is one of inversion recovery (IR) sequences. In this study, we applied real-IR technique to Fast-Spin echo (FSE) based interleaved FLAIR sequence and assessed for tissue contrasts by volunteer brain study on 1.5T and 3T. We concluded that real-FLAIR imaging has more advantageous than the magnitude-based FLAIR imaging from the points of tissue contrast enhancement, robust against TI setting, and in addition, reducing CSF inflow artifacts.

Zhikui Xiao¹, Lou Xin², Shen Hao¹, and Cao Guang^1
1Global Applied Science Laboratory, GE Healthcare, Beijing, Beijing, China, People's Republic of, ²Department of Radiology, PLA General Hospital, Beijing, Beijing, China, People's Republic of

In this work, we implemented a 3D flow-dephasing prepared FSE sequence by adding a flow dephasing preparation pulse prior to 3D-FSE acquisition. This enables better blood signal suppression for isotropic high spatial resolution nerve imaging at 3.0T. In-vivo lumbosacral plexus results are shown to evaluate the proposed scheme. We show that flow-dephasing prepared image present significantly better blood vessel signal suppression with high spatial resolution.

Amy Kathleen McGuinness¹, Christina Malamateniou¹, Joanna M Allsop¹, Serena J Counsell¹, Rita G Nunes¹, Zhi Q Wu¹, Nora Tusor², Ash Ederies², Jo V Hajnal¹, and Mary A Rutherford^1
1Imaging Sciences Department, MRC Clinical Sciences Centre, Hammersmith Hospital, Imperial College London, London, United Kingdom, ²Neonatal Imaging Group, Hammersmith Hospital, Imperial College London, London, United Kingdom

SNAPIR, an optimized single-shot T1-weighted acquisition, improves normal brain anatomy delineation in cases of fetal motion compared to a standard T1-weighted gradient echo (T1W-GE) protocol. However, the role of SNAPIR in neonatal imaging where motion may also be an issue has not yet been established. The aim of this study was to evaluate and compare neonatal brain pathology delineation using SNAPIR and a standard T1W-GE protocol, the MP-RAGE. In the absence of motion MP-RAGE is the preferred sequence. However, SNAPIR is almost as sensitive in many pathologies and is a robust technique when motion degrades T1W GE protocols.

Paul Polak¹, Robert Zivadinov^1,2, and Guy Poloni^1
1Buffalo Neuroimaging Analysis Center, Department of Neurology, University at Buffalo, State University of New York, Buffalo, NY, United States, ²The Jacobs Neurological Institute, Department of Neurology, University at Buffalo, State University of New York, Buffalo, NY, United States

Multiple Sclerosis is believed to be a chronic, autoimmune, neurodegenerative disease of the CNS. Approaches that monitor disease progression in the white matter tissue via T2-weighed imaging suffer from the clinical/radiological paradox, in that poor or no correlation is found between clinical outcomes and MR metrics. We propose using an experimental method to optimize lesion contrast in white and grey matter tissues, using regions of interest analysis. Using the resulting data we propose a set of 3D DIR sequence parameters designed to improve the detection of clinically significant lesions in both white and grey matter.

Paul Polak¹, Robert Zivadinov^1,2, and Guy Poloni^1
1Buffalo Neuroimaging Analysis Center, Department of Neurology, University at Buffalo, State University of New York, Buffalo, NY, United States, ²The Jacobs Neurological Institute, Department of Neurology, University at Buffalo, State University of New York, Buffalo, NY, United States

Detection of brain tissue lesions via T2-weighted imaging is standard clinical practice in diagnosing and prognosis of multiple sclerosis. However, these methods suffer from the clinical/radiological paradox, in that poor or no correlation is found between clinical outcomes and MR metrics. Improving lesion sensitivity in MR techniques may help resolve this conflict. We propose an experimental method to optimize lesion contrast in a 3D FLAIR sequence using a region of interest analysis to optimize sequence parameters. The derived 3D FLAIR imaging sequence is superior to current clinical 2D and 3D FLAIR sequences, in both detection power and resolution.

Albert Kir^1,2, and Alan Blair McMillan^1
1Magnetic Resonance Research Center, Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD, United States, ²Electrical Engineering and Computer Science, University of Maryland Baltimore County, Baltimore, MD, United States

Previous studies investigating optimized rapid 3D inversion-prepared gradient echo imaging sequences have relied upon the assumption of ideal radio-frequency (RF) spoiling. In this work, a new simulation approach based on the extended phase graph (EPG) formalism, which considers the effect of residual transverse magnetization, is used to predict both signal levels and contrast. We conducted phantom and in vivo brain studies to validate and verify the improvement of the proposed approach over approaches based on ideal RF spoiling, to generate images with optimal SNR, tissue contrast, and acquisition time.

Bhashwati Roy¹, Sanjay Verma², Rishi Awasthi¹, Ram KS Rathore², Ramesh Venkatesan³, SA Yoganathan⁴, KJ Maria Das⁴, Kashi Nath Prasad⁵, and Rakesh Kumar Gupta^1
1Radiodiagnosis, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India, Lucknow, Uttar Pradesh, India, ²Mathematics & Statistics, Indian Institute of Technology, Kanpur, Kanpur, Uttar Pradesh, India, ³Wipro-GE Healthcare, Bangalore, Karnataka, India, ⁴Radiotherapy, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India, Lucknow, Uttar Pradesh, India, ⁵Microbiology, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India, Lucknow, Uttar Pradesh, India

A total of 35 patients having fifty two calcified cysts were imaged using conventional and T2 star weighted angiography, used for extracting phase images. A significant correlation between CT-Hounsfield values and phase values proves the sensitivity of phase imaging in detecting presence calcification in a lesion as that of CT. In addition its ability to differentiate between diamagnetic and paramagnetic minerals gives phase imaging an edge over CT in demonstrating these lesions.

Yunhong Shu¹, Joshua D Trzasko¹, John III Huston¹, Armando Manduca¹, and Matt A Bernstein^1
1Radiology, Mayo Clinic, Rochester, MN, United States

The purpose of this work is to explore the feasibility of applying an undersampled spherical SWIRLS trajectory to CE-MRA applications at 3T. NUFFT reconstruction and off-resonance correction was used to further improve the image quality.