Publications

2015

Ning L, Georgiou TT, Tannenbaum A, Boyd SP. Linear Models Based on Noisy Data and the Frisch Scheme. SIAM Rev Soc Ind Appl Math. 2015;57(2):167–97.
We address the problem of identifying linear relations among variables based on noisy measurements. This is a central question in the search for structure in large data sets. Often a key assumption is that measurement errors in each variable are independent. This basic formulation has its roots in the work of Charles Spearman in 1904 and of Ragnar Frisch in the 1930s. Various topics such as errors-in-variables, factor analysis, and instrumental variables all refer to alternative viewpoints on this problem and on ways to account for the anticipated way that noise enters the data. In the present paper we begin by describing certain fundamental contributions by the founders of the field and provide alternative modern proofs to certain key results. We then go on to consider a modern viewpoint and novel numerical techniques to the problem. The central theme is expressed by the Frisch-Kalman dictum, which calls for identifying a noise contribution that allows a maximal number of simultaneous linear relations among the noise-free variables-a rank minimization problem. In the years since Frisch’s original formulation, there have been several insights, including trace minimization as a convenient heuristic to replace rank minimization. We discuss convex relaxations and theoretical bounds on the rank that, when met, provide guarantees for global optimality. A complementary point of view to this minimum-rank dictum is presented in which models are sought leading to a uniformly optimal quadratic estimation error for the error-free variables. Points of contact between these formalisms are discussed, and alternative regularization schemes are presented.
Radmanesh A, Zamani AA, Whalen S, Tie Y, Suarez RO, Golby AJ. Comparison of seeding methods for visualization of the corticospinal tracts using single tensor tractography. Clin Neurol Neurosurg. 2015;129:44–9.
OBJECTIVES: To compare five different seeding methods to delineate hand, foot, and lip components of the corticospinal tract (CST) using single tensor tractography. METHODS: We studied five healthy subjects and 10 brain tumor patients. For each subject, we used five different seeding methods, from (1) cerebral peduncle (CP), (2) posterior limb of the internal capsule (PLIC), (3) white matter subjacent to functional MRI activations (fMRI), (4) whole brain and then selecting the fibers that pass through both fMRI and CP (WBF-CP), and (5) whole brain and then selecting the fibers that pass through both fMRI and PLIC (WBF-PLIC). Two blinded neuroradiologists rated delineations as anatomically successful or unsuccessful tractography. The proportions of successful trials from different methods were compared by Fisher’s exact test. RESULTS: To delineate hand motor tract, seeding through fMRI activation areas was more effective than through CP (p0.1). WBF-CP delineated hand motor tracts in a larger proportion of trials than CP alone (p
Liu S, Liu S, Cai W, Che H, Pujol S, Kikinis R, Feng D, Fulham MJ. Multimodal neuroimaging feature learning for multiclass diagnosis of Alzheimer s disease. IEEE Trans Biomed Eng. 2015;62(4):1132–40.
The accurate diagnosis of Alzheimer’s disease (AD) is essential for patient care and will be increasingly important as disease modifying agents become available, early in the course of the disease. Although studies have applied machine learning methods for the computer-aided diagnosis of AD, a bottleneck in the diagnostic performance was shown in previous methods, due to the lacking of efficient strategies for representing neuroimaging biomarkers. In this study, we designed a novel diagnostic framework with deep learning architecture to aid the diagnosis of AD. This framework uses a zero-masking strategy for data fusion to extract complementary information from multiple data modalities. Compared to the previous state-of-the-art workflows, our method is capable of fusing multimodal neuroimaging features in one setting and has the potential to require less labeled data. A performance gain was achieved in both binary classification and multiclass classification of AD. The advantages and limitations of the proposed framework are discussed.
Mandl RCW, Pasternak O, Cahn W, Kubicki M, Kahn RS, Shenton ME, Pol HEH. Comparing Free Water Imaging and Magnetization Transfer Measurements in Schizophrenia. Schizophr Res. 2015;161(1):126–32.
Diffusion weighted imaging (DWI) has been extensively used to study the microarchitecture of white matter in schizophrenia. However, popular DWI-derived measures such as fractional anisotropy (FA) may be sensitive to many types of pathologies, and thus the interpretation of reported differences in these measures remains difficult. Combining DWI with magnetization transfer ratio (MTR) - a putative measure of white matter myelination - can help us reveal the underlying mechanisms. Previous findings hypothesized that MTR differences in schizophrenia are associated with free water concentrations, which also affect the DWIs. In this study we use a recently proposed DWI-derived method called free-water imaging to assess this hypothesis. We have reanalyzed data from a previous study by using a fiber-based analysis of free-water imaging, providing a free-water fraction, as well as mean diffusivity and FA corrected for free-water, in addition to MTR along twelve major white matter fiber bundles in 40 schizophrenia patients and 40 healthy controls. We tested for group differences in each fiber bundle and for each measure separately and computed correlations between the MTR and the DWI-derived measures separately for both groups. Significant higher average MTR values in patients were found for the right uncinate fasciculus, the right arcuate fasciculus and the right inferior-frontal occipital fasciculus. No significant results were found for the other measures. No significant differences in correlations were found between MTR and the DWI-derived measures. The results suggest that MTR and free-water imaging measures can be considered complementary, promoting the acquisition of MTR in addition to DWI to identify group differences, as well as to better understand the underlying mechanisms in schizophrenia.
Jakab M, Kikinis R. CT-based Atlas of the Head and Neck. 2015.
This Head and Neck Atlas has been made available by the Surgical Planning Laboratory at Brigham and Women s Hospital. The data set consists of: 1. Reduced resolution (256x256) of the MANIX data set from the OSIRIX data sets. 2. A set of detailed label maps. 3. A set of three-dimensional models of the labeled anatomical structures. 4. Several pre-defined Scene Views (“anatomy teaching files”). 5. Annotation as supplementary information associated with a scene. 6. Anatomical model hierarchy. All in a mrb (Medical Reality Bundle) archive file that contains the mrml scene file and all data for loading into Slicer 4 for displaying the volumes in 3D Slicer version 4.0 or greater, available for download. The atlas data is made available under terms of the 3D Slicer License section B.This work is funded as part of the Neuroimaging Analysis Center, grant number P41 EB015902, by the NIH’s National Institute of Biomedical Imaging and Bioengineering (NIBIB) and the Google Faculty Research Award.Contributors: Neha Agrawal, Matthew D Artista, Susan Kikinis, Dashawn Richardson, Daniel Sachs.This atlas maybe viewed with our Open Anatomy Browser.
Donnell LJO, Pasternak O. Does diffusion MRI tell us anything about the white matter? An overview of methods and pitfalls. Schizophr Res. 2015;161(1):133–41.
One key pitfall in diffusion magnetic resonance imaging (dMRI) clinical neuroimaging research is the challenge of understanding and interpreting the results of a complex analysis pipeline. The sophisticated algorithms employed by the analysis software, combined with the relatively non-specific nature of many diffusion measurements, lead to challenges in interpretation of the results. This paper is aimed at an intended audience of clinical researchers who are learning about dMRI or trying to interpret dMRI results, and who may be wondering "Does dMRI tell us anything about the white matter?" We present a critical review of dMRI methods and measures used in clinical neuroimaging research, focusing on the most commonly used analysis methods and the most commonly reported measures. We describe important pitfalls in every section, and provide extensive references for the reader interested in more detail.
Balasubramanian M, Mulkern RV, Wells WM III, Sundaram P, Orbach DB. Magnetic Resonance Imaging of Ionic Currents in Solution: The Effect of Magnetohydrodynamic Flow. Magn Reson Med. 2015;74(4):1145–55.
PURPOSE: Reliably detecting MRI signals in the brain that are more tightly coupled to neural activity than blood-oxygen-level-dependent fMRI signals could not only prove valuable for basic scientific research but could also enhance clinical applications such as epilepsy presurgical mapping. This endeavor will likely benefit from an improved understanding of the behavior of ionic currents, the mediators of neural activity, in the presence of the strong magnetic fields that are typical of modern-day MRI scanners. THEORY: Of the various mechanisms that have been proposed to explain the behavior of ionic volume currents in a magnetic field, only one-magnetohydrodynamic flow-predicts a slow evolution of signals, on the order of a minute for normal saline in a typical MRI scanner. METHODS: This prediction was tested by scanning a volume-current phantom containing normal saline with gradient-echo-planar imaging at 3 T. RESULTS: Greater signal changes were observed in the phase of the images than in the magnitude, with the changes evolving on the order of a minute. CONCLUSION: These results provide experimental support for the MHD flow hypothesis. Furthermore, MHD-driven cerebrospinal fluid flow could provide a novel fMRI contrast mechanism.
Richolt J, Jakab M, Kikinis R. MRI-based Atlas of the Knee. 2015.
The Surgical Planning Laboratory at Brigham and Women’s Hospital, Harvard Medical School, developed the SPL Knee Atlas. The atlas was derived from a MRI scan, using semi-automated image segmentation and three-dimensional reconstruction techniques. The current version consists of: 1. the original MRI scan; 2. a set of detailed label maps; 3. a set of three-dimensional models of the labeled anatomical structures; 4. a mrml-file that allows loading all of the data into the 3D Slicer for visualization. 5. several pre-defined 3D views (“anatomy teaching files”). The SPL Knee Atlas provides important reference information for anatomy teaching, and template driven segmentation. Visualization of the data requires Slicer 3. This software package can be downloaded from here. We are pleased to make this atlas available to our colleagues for free download. Please note that the data is being distributed under the Slicer license. By downloading these data, you agree to acknowledge our contribution in any of your publications that result form the use of this atlas. The Slicer4 version archived in a mrb (Medical Reality Bundle) file that contains the mrml scene file and all data for loading into Slicer 4 for displaying the volumes in 3D Slicer version 4.0 or greater, available for download.This work is funded as part of the Neuroimaging Analysis Center, grant number P41 RR013218, by the NIH’s National Center for Research Resources (NCRR) and grant number P41 EB015902, by the NIH’s National Institute of Biomedical Imaging and Bioengineering (NIBIB) and the Google Faculty Research Award.Contributors: Matthew D’Artista, Alex Kikinis.This atlas maybe viewed with our Open Anatomy Browser.
Talos IF, Jakab M, Kikinis R. CT-based Atlas of the Abdomen. 2015.
The Surgical Planning Laboratory at Brigham and Women’s Hospital, Harvard Medical School, developed the SPL Abdominal Atlas. The atlas was derived from a computed tomography (CT) scan, using semi-automated image segmentation and three-dimensional reconstruction techniques. The current version consists of: 1. the original CT scan; 2. a set of detailed label maps; 3. a set of three-dimensional models of the labeled anatomical structures; 4. a mrml-file that allows loading all of the data into the 3D Slicer for visualization (see the tutorial associated with the atlas); 5. several pre-defined 3D-views ( anatomy teaching files ). The SPL Abdominal Atlas provides important reference information for surgical planning, anatomy teaching, and template driven segmentation. Visualization of the data requires Slicer 3. This software package can be downloaded from here. We are pleased to make this atlas available to our colleagues for free download. Please note that the data is being distributed under the Slicer license. By downloading these data, you agree to acknowledge our contribution in any of your publications that result form the use of this atlas. The Slicer4 version archived in a mrb (Medical Reality Bundle) file that contains the mrml scene file and all data for loading into Slicer 4 for displaying the volumes in 3D Slicer version 4.0 or greater, available for download.This work is funded as part of the Neuroimaging Analysis Center, grant number P41 RR013218, by the NIH’s National Center for Research Resources (NCRR) and grant number P41 EB015902, by the NIH’s National Institute of Biomedical Imaging and Bioengineering (NIBIB) and the Google Faculty Research Award.Contributors: Matthew D’Artista, Alex Kikinis, Tobias Schmidt, Svenja van der Gaag.This atlas maybe viewed with our Open Anatomy Browser.
Liu S, Liu S, Cai W, Pujol S, Kikinis R, Feng DD. Multi-Phase Feature Representation Learning for Neurodegenerative Disease Diagnosis. Artificial Life and Computational Intelligence. 2015;LNAI 8955:350–9.
Feature learning with high dimensional neuroimaging features has been explored for the applications on neurodegenerative diseases. Low-dimensional biomarkers, such as mental status test scores and cerebrospinal fluid level, are essential in clinical diagnosis of neurological disorders, because they could be simple and effective for the clinicians to assess the disorder’s progression and severity. Rather than only using the low-dimensional biomarkers as inputs for decision making systems, we believe that such low-dimensional biomarkers can be used for enhancing the feature learning pipeline. In this study, we proposed a novel feature representation learning framework, Multi-Phase Feature Representation (MPFR), with low-dimensional biomarkers embedded. MPFR learns high-level neuroimaging features by extracting the associations between the low-dimensional biomarkers and the highdimensional neuroimaging features with a deep neural network. We validated the proposed framework using the Mini-Mental-State-Examination (MMSE) scores as a low-dimensional biomarker and multi-modal neuroimaging data as the high-dimensional neuroimaging features from the ADNI baseline cohort. The proposed approach outperformed the original neural network in both binary and ternary Alzheimer’s disease classification tasks.