Matching Items (103)
Description
Recent advances in medical imaging technology have greatly enhanced imaging based diagnosis which requires computational effective and accurate algorithms to process the images (e.g., measure the objects) for quantitative assessment. In this dissertation, one type of imaging objects is of interest: small blobs. Example small blob objects are cells in histopathology images, small breast lesions in ultrasound images, glomeruli in kidney MR images etc. This problem is particularly challenging because the small blobs often have inhomogeneous intensity distribution and indistinct boundary against the background.
This research develops a generalized four-phased system for small blob detections. The system includes (1) raw image transformation, (2) Hessian pre-segmentation, (3) feature extraction and (4) unsupervised clustering for post-pruning. First, detecting blobs from 2D images is studied where a Hessian-based Laplacian of Gaussian (HLoG) detector is proposed. Using the scale space theory as foundation, the image is smoothed via LoG. Hessian analysis is then launched to identify the single optimal scale based on which a pre-segmentation is conducted. Novel Regional features are extracted from pre-segmented blob candidates and fed to Variational Bayesian Gaussian Mixture Models (VBGMM) for post pruning. Sixteen cell histology images and two hundred cell fluorescent images are tested to demonstrate the performances of HLoG. Next, as an extension, Hessian-based Difference of Gaussians (HDoG) is proposed which is capable to identify the small blobs from 3D images. Specifically, kidney glomeruli segmentation from 3D MRI (6 rats, 3 humans) is investigated. The experimental results show that HDoG has the potential to automatically detect glomeruli, enabling new measurements of renal microstructures and pathology in preclinical and clinical studies. Realizing the computation time is a key factor impacting the clinical adoption, the last phase of this research is to investigate the data reduction technique for VBGMM in HDoG to handle large-scale datasets. A new coreset algorithm is developed for variational Bayesian mixture models. Using the same MRI dataset, it is observed that the four-phased system with coreset-VBGMM has similar performance as using the full dataset but about 20 times faster.
This research develops a generalized four-phased system for small blob detections. The system includes (1) raw image transformation, (2) Hessian pre-segmentation, (3) feature extraction and (4) unsupervised clustering for post-pruning. First, detecting blobs from 2D images is studied where a Hessian-based Laplacian of Gaussian (HLoG) detector is proposed. Using the scale space theory as foundation, the image is smoothed via LoG. Hessian analysis is then launched to identify the single optimal scale based on which a pre-segmentation is conducted. Novel Regional features are extracted from pre-segmented blob candidates and fed to Variational Bayesian Gaussian Mixture Models (VBGMM) for post pruning. Sixteen cell histology images and two hundred cell fluorescent images are tested to demonstrate the performances of HLoG. Next, as an extension, Hessian-based Difference of Gaussians (HDoG) is proposed which is capable to identify the small blobs from 3D images. Specifically, kidney glomeruli segmentation from 3D MRI (6 rats, 3 humans) is investigated. The experimental results show that HDoG has the potential to automatically detect glomeruli, enabling new measurements of renal microstructures and pathology in preclinical and clinical studies. Realizing the computation time is a key factor impacting the clinical adoption, the last phase of this research is to investigate the data reduction technique for VBGMM in HDoG to handle large-scale datasets. A new coreset algorithm is developed for variational Bayesian mixture models. Using the same MRI dataset, it is observed that the four-phased system with coreset-VBGMM has similar performance as using the full dataset but about 20 times faster.
ContributorsZhang, Min (Author) / Wu, Teresa (Thesis advisor) / Li, Jing (Committee member) / Pavlicek, William (Committee member) / Askin, Ronald (Committee member) / Arizona State University (Publisher)
Created2015
Description
Discriminative learning when training and test data belong to different distributions is a challenging and complex task. Often times we have very few or no labeled data from the test or target distribution, but we may have plenty of labeled data from one or multiple related sources with different distributions. Due to its capability of migrating knowledge from related domains, transfer learning has shown to be effective for cross-domain learning problems. In this dissertation, I carry out research along this direction with a particular focus on designing efficient and effective algorithms for BioImaging and Bilingual applications. Specifically, I propose deep transfer learning algorithms which combine transfer learning and deep learning to improve image annotation performance. Firstly, I propose to generate the deep features for the Drosophila embryo images via pretrained deep models and build linear classifiers on top of the deep features. Secondly, I propose to fine-tune the pretrained model with a small amount of labeled images. The time complexity and performance of deep transfer learning methodologies are investigated. Promising results have demonstrated the knowledge transfer ability of proposed deep transfer algorithms. Moreover, I propose a novel Robust Principal Component Analysis (RPCA) approach to process the noisy images in advance. In addition, I also present a two-stage re-weighting framework for general domain adaptation problems. The distribution of source domain is mapped towards the target domain in the first stage, and an adaptive learning model is proposed in the second stage to incorporate label information from the target domain if it is available. Then the proposed model is applied to tackle cross lingual spam detection problem at LinkedIn’s website. Our experimental results on real data demonstrate the efficiency and effectiveness of the proposed algorithms.
ContributorsSun, Qian (Author) / Ye, Jieping (Committee member) / Xue, Guoliang (Committee member) / Liu, Huan (Committee member) / Li, Jing (Committee member) / Arizona State University (Publisher)
Created2015
Description
The presence of a rich set of embedded sensors on mobile devices has been fuelling various sensing applications regarding the activities of individuals and their surrounding environment, and these ubiquitous sensing-capable mobile devices are pushing the new paradigm of Mobile Crowd Sensing (MCS) from concept to reality. MCS aims to outsource sensing data collection to mobile users and it could revolutionize the traditional ways of sensing data collection and processing. In the meantime, cloud computing provides cloud-backed infrastructures for mobile devices to provision their capabilities with network access. With enormous computational and storage resources along with sufficient bandwidth, it functions as the hub to handle the sensing service requests from sensing service consumers and coordinate sensing task assignment among eligible mobile users to reach a desired quality of sensing service. This paper studies the problem of sensing task assignment to mobile device owners with specific spatio-temporal traits to minimize the cost and maximize the utility in MCS while adhering to QoS constraints. Greedy approaches and hybrid solutions combined with bee algorithms are explored to address the problem.
Moreover, the privacy concerns arise with the widespread deployment of MCS from both the data contributors and the sensing service consumers. The uploaded sensing data, especially those tagged with spatio-temporal information, will disclose the personal information of the data contributors. In addition, the sensing service requests can reveal the personal interests of service consumers. To address the privacy issues, this paper constructs a new framework named Privacy-Preserving Mobile Crowd Sensing (PP-MCS) to leverage the sensing capabilities of ubiquitous mobile devices and cloud infrastructures. PP-MCS has a distributed architecture without relying on trusted third parties for privacy-preservation. In PP-MCS, the sensing service consumers can retrieve data without revealing the real data contributors. Besides, the individual sensing records can be compared against the aggregation result while keeping the values of sensing records unknown, and the k-nearest neighbors could be approximately identified without privacy leaks. As such, the privacy of the data contributors and the sensing service consumers can be protected to the greatest extent possible.
Moreover, the privacy concerns arise with the widespread deployment of MCS from both the data contributors and the sensing service consumers. The uploaded sensing data, especially those tagged with spatio-temporal information, will disclose the personal information of the data contributors. In addition, the sensing service requests can reveal the personal interests of service consumers. To address the privacy issues, this paper constructs a new framework named Privacy-Preserving Mobile Crowd Sensing (PP-MCS) to leverage the sensing capabilities of ubiquitous mobile devices and cloud infrastructures. PP-MCS has a distributed architecture without relying on trusted third parties for privacy-preservation. In PP-MCS, the sensing service consumers can retrieve data without revealing the real data contributors. Besides, the individual sensing records can be compared against the aggregation result while keeping the values of sensing records unknown, and the k-nearest neighbors could be approximately identified without privacy leaks. As such, the privacy of the data contributors and the sensing service consumers can be protected to the greatest extent possible.
ContributorsWang, Zhijie (Thesis advisor) / Xue, Guoliang (Committee member) / Sen, Arunabha (Committee member) / Li, Jing (Committee member) / Arizona State University (Publisher)
Created2016
Description
In this thesis, I present the study of nucleon structure from distinct perspectives. I start by elaborating the motivations behind the endeavors and then introducing the key concept, namely the generalized parton distribution functions (GPDs), which serves as the frame- work describing hadronic particles in terms of their fundamental constituents. The second chapter is then devoted to a detailed phenomenological study of the Virtual Compton Scattering (VCS) process, where a more comprehensive parametrization is suggested. In the third chapter, the renormalization kernels that enters the QCD evolution equations at twist- 4 accuracy are computed in terms of Feynman diagrams in momentum space, which can be viewed as an extension of the work by Bukhvostov, Frolov, Lipatov, and Kuraev (BKLK). The results can be used for determining the QCD background interaction for future precision measurements.
ContributorsJi, Yao, Ph. D (Author) / Belitsky, Andrei (Thesis advisor) / Lebed, Richard (Committee member) / Schmidt, Kevin E (Committee member) / Vachaspati, Tanmay (Committee member) / Arizona State University (Publisher)
Created2016
Description
The work presented in this dissertation examines three different nonequilibrium particle physics processes that could play a role in answering the question “how was the particle content of today’s universe produced after the big bang?” Cosmic strings produced from spontaneous breaking of a hidden sector $U(1)_{\rm X}$ symmetry could couple to Standard Model fields through Higgs Portal or Kinetic Mixing operators and radiate particles that contribute to the diffuse gamma ray background. In this work we calculate the properties of these strings, including finding effective couplings between the strings and Standard Model fields. Explosive particle production after inflation, known as preheating, would have produced a stochastic background of gravitational waves (GW). This work shows how the presence of realistic additional fields and interactions can affect this prediction dramatically. Specifically, it considers the inflaton to be coupled to a light scalar field, and shows that even a very small quartic self-interaction term will reduce the amplitude of the gravitational wave spectrum. For self-coupling $\lambda_{\chi} \gtrsim g^2$, where $g^2$ is the inflaton-scalar coupling, the peak energy density goes as $\Omega_{\rm GW}^{(\lambda_{\chi})} / \Omega_{\rm GW}^{(\lambda_{\chi}=0)} \sim (g^2/\lambda_{\chi})^{2}$. Finally, leptonic charge-parity (CP) violation could be an important clue to understanding the origin of our universe's matter-antimatter asymmetry, and long-baseline neutrino oscillation experiments in the coming decade may uncover this. The CP violating effects of a possible fourth ``sterile" neutrino can interfere with the usual three neutrinos; this work shows how combinations of various measurements can help break those degeneracies.
ContributorsHyde, Jeffrey Morgan (Author) / Vachaspati, Tanmay (Thesis advisor) / Easson, Damien (Committee member) / Belitsky, Andrei (Committee member) / Comfort, Joseph (Committee member) / Arizona State University (Publisher)
Created2016
Description
Transfer learning refers to statistical machine learning methods that integrate the knowledge of one domain (source domain) and the data of another domain (target domain) in an appropriate way, in order to develop a model for the target domain that is better than a model using the data of the target domain alone. Transfer learning emerged because classic machine learning, when used to model different domains, has to take on one of two mechanical approaches. That is, it will either assume the data distributions of the different domains to be the same and thereby developing one model that fits all, or develop one model for each domain independently. Transfer learning, on the other hand, aims to mitigate the limitations of the two approaches by accounting for both the similarity and specificity of related domains. The objective of my dissertation research is to develop new transfer learning methods and demonstrate the utility of the methods in real-world applications. Specifically, in my methodological development, I focus on two different transfer learning scenarios: spatial transfer learning across different domains and temporal transfer learning along time in the same domain. Furthermore, I apply the proposed spatial transfer learning approach to modeling of degenerate biological systems.Degeneracy is a well-known characteristic, widely-existing in many biological systems, and contributes to the heterogeneity, complexity, and robustness of biological systems. In particular, I study the application of one degenerate biological system which is to use transcription factor (TF) binding sites to predict gene expression across multiple cell lines. Also, I apply the proposed temporal transfer learning approach to change detection of dynamic network data. Change detection is a classic research area in Statistical Process Control (SPC), but change detection in network data has been limited studied. I integrate the temporal transfer learning method called the Network State Space Model (NSSM) and SPC and formulate the problem of change detection from dynamic networks into a covariance monitoring problem. I demonstrate the performance of the NSSM in change detection of dynamic social networks.
ContributorsZou, Na (Author) / Li, Jing (Thesis advisor) / Baydogan, Mustafa (Committee member) / Borror, Connie (Committee member) / Montgomery, Douglas C. (Committee member) / Wu, Teresa (Committee member) / Arizona State University (Publisher)
Created2015
Description
Network traffic analysis by means of Quality of Service (QoS) is a popular research and development area among researchers for a long time. It is becoming even more relevant recently due to ever increasing use of the Internet and other public and private communication networks. Fast and precise QoS analysis is a vital task in mission-critical communication networks (MCCNs), where providing a certain level of QoS is essential for national security, safety or economic vitality. In this thesis, the details of all aspects of a comprehensive computational framework for QoS analysis in MCCNs are provided. There are three main QoS analysis tasks in MCCNs; QoS measurement, QoS visualization and QoS prediction. Definitions of these tasks are provided and for each of those, complete solutions are suggested either by referring to an existing work or providing novel methods.
A scalable and accurate passive one-way QoS measurement algorithm is proposed. It is shown that accurate QoS measurements are possible using network flow data.
Requirements of a good QoS visualization platform are listed. Implementations of the capabilities of a complete visualization platform are presented.
Steps of QoS prediction task in MCCNs are defined. The details of feature selection, class balancing through sampling and assessing classification algorithms for this task are outlined. Moreover, a novel tree based logistic regression method for knowledge discovery is introduced. Developed prediction framework is capable of making very accurate packet level QoS predictions and giving valuable insights to network administrators.
A scalable and accurate passive one-way QoS measurement algorithm is proposed. It is shown that accurate QoS measurements are possible using network flow data.
Requirements of a good QoS visualization platform are listed. Implementations of the capabilities of a complete visualization platform are presented.
Steps of QoS prediction task in MCCNs are defined. The details of feature selection, class balancing through sampling and assessing classification algorithms for this task are outlined. Moreover, a novel tree based logistic regression method for knowledge discovery is introduced. Developed prediction framework is capable of making very accurate packet level QoS predictions and giving valuable insights to network administrators.
ContributorsSenturk, Muhammet Burhan (Author) / Li, Jing (Thesis advisor) / Baydogan, Mustafa G (Committee member) / Wu, Teresa (Committee member) / Arizona State University (Publisher)
Created2014
Description
In a healthcare setting, the Sterile Processing Department (SPD) provides ancillary services to the Operating Room (OR), Emergency Room, Labor & Delivery, and off-site clinics. SPD's function is to reprocess reusable surgical instruments and return them to their home departments. The management of surgical instruments and medical devices can impact patient safety and hospital revenue. Any time instrumentation or devices are not available or are not fit for use, patient safety and revenue can be negatively impacted. One step of the instrument reprocessing cycle is sterilization. Steam sterilization is the sterilization method used for the majority of surgical instruments and is preferred to immediate use steam sterilization (IUSS) because terminally sterilized items can be stored until needed. IUSS Items must be used promptly and cannot be stored for later use. IUSS is intended for emergency situations and not as regular course of action. Unfortunately, IUSS is used to compensate for inadequate inventory levels, scheduling conflicts, and miscommunications. If IUSS is viewed as an adverse event, then monitoring IUSS incidences can help healthcare organizations meet patient safety goals and financial goals along with aiding in process improvement efforts. This work recommends statistical process control methods to IUSS incidents and illustrates the use of control charts for IUSS occurrences through a case study and analysis of the control charts for data from a health care provider. Furthermore, this work considers the application of data mining methods to IUSS occurrences and presents a representative example of data mining to the IUSS occurrences. This extends the application of statistical process control and data mining in healthcare applications.
ContributorsWeart, Gail (Author) / Runger, George C. (Thesis advisor) / Li, Jing (Committee member) / Shunk, Dan (Committee member) / Arizona State University (Publisher)
Created2014
Description
With the discovery of the Higgs Boson in 2012, particle physics has decidedly moved beyond the Standard Model into a new epoch. Though the Standard Model particle content is now completely accounted for, there remain many theoretical issues about the structure of the theory in need of resolution. Among these is the hierarchy problem: since the renormalized Higgs mass receives quadratic corrections from a higher cutoff scale, what keeps the Higgs boson light? Many possible solutions to this problem have been advanced, such as supersymmetry, Randall-Sundrum models, or sub-millimeter corrections to gravity. One such solution has been advanced by the Lee-Wick Standard Model. In this theory, higher-derivative operators are added to the Lagrangian for each Standard Model field, which result in propagators that possess two physical poles and fall off more rapidly in the ultraviolet regime. It can be shown by an auxiliary field transformation that the higher-derivative theory is identical to positing a second, manifestly renormalizable theory in which new fields with opposite-sign kinetic and mass terms are found. These so-called Lee-Wick fields have opposite-sign propagators, and famously cancel off the quadratic divergences that plague the renormalized Higgs mass. The states in the Hilbert space corresponding to Lee-Wick particles have negative norm, and implications for causality and unitarity are examined.
This dissertation explores a variant of the theory called the N = 3 Lee-Wick
Standard Model. The Lagrangian of this theory features a yet-higher derivative operator, which produces a propagator with three physical poles and possesses even better high-energy behavior than the minimal Lee-Wick theory. An analogous auxiliary field transformation takes this higher-derivative theory into a renormalizable theory with states of alternating positive, negative, and positive norm. The phenomenology of this theory is examined in detail, with particular emphasis on the collider signatures of Lee-Wick particles, electroweak precision constraints on the masses that the new particles can take on, and scenarios in early-universe cosmology in which Lee-Wick particles can play a significant role.
This dissertation explores a variant of the theory called the N = 3 Lee-Wick
Standard Model. The Lagrangian of this theory features a yet-higher derivative operator, which produces a propagator with three physical poles and possesses even better high-energy behavior than the minimal Lee-Wick theory. An analogous auxiliary field transformation takes this higher-derivative theory into a renormalizable theory with states of alternating positive, negative, and positive norm. The phenomenology of this theory is examined in detail, with particular emphasis on the collider signatures of Lee-Wick particles, electroweak precision constraints on the masses that the new particles can take on, and scenarios in early-universe cosmology in which Lee-Wick particles can play a significant role.
ContributorsTerBeek, Russell Henry (Author) / Lebed, Richard F (Thesis advisor) / Alarcon, Ricardo (Committee member) / Belitsky, Andrei (Committee member) / Chamberlin, Ralph (Committee member) / Parikh, Maulik (Committee member) / Arizona State University (Publisher)
Created2015
Description
Data imbalance and data noise often coexist in real world datasets. Data imbalance affects the learning classifier by degrading the recognition power of the classifier on the minority class, while data noise affects the learning classifier by providing inaccurate information and thus misleads the classifier. Because of these differences, data imbalance and data noise have been treated separately in the data mining field. Yet, such approach ignores the mutual effects and as a result may lead to new problems. A desirable solution is to tackle these two issues jointly. Noting the complementary nature of generative and discriminative models, this research proposes a unified model fusion based framework to handle the imbalanced classification with noisy dataset.
The phase I study focuses on the imbalanced classification problem. A generative classifier, Gaussian Mixture Model (GMM) is studied which can learn the distribution of the imbalance data to improve the discrimination power on imbalanced classes. By fusing this knowledge into cost SVM (cSVM), a CSG method is proposed. Experimental results show the effectiveness of CSG in dealing with imbalanced classification problems.
The phase II study expands the research scope to include the noisy dataset into the imbalanced classification problem. A model fusion based framework, K Nearest Gaussian (KNG) is proposed. KNG employs a generative modeling method, GMM, to model the training data as Gaussian mixtures and form adjustable confidence regions which are less sensitive to data imbalance and noise. Motivated by the K-nearest neighbor algorithm, the neighboring Gaussians are used to classify the testing instances. Experimental results show KNG method greatly outperforms traditional classification methods in dealing with imbalanced classification problems with noisy dataset.
The phase III study addresses the issues of feature selection and parameter tuning of KNG algorithm. To further improve the performance of KNG algorithm, a Particle Swarm Optimization based method (PSO-KNG) is proposed. PSO-KNG formulates model parameters and data features into the same particle vector and thus can search the best feature and parameter combination jointly. The experimental results show that PSO can greatly improve the performance of KNG with better accuracy and much lower computational cost.
The phase I study focuses on the imbalanced classification problem. A generative classifier, Gaussian Mixture Model (GMM) is studied which can learn the distribution of the imbalance data to improve the discrimination power on imbalanced classes. By fusing this knowledge into cost SVM (cSVM), a CSG method is proposed. Experimental results show the effectiveness of CSG in dealing with imbalanced classification problems.
The phase II study expands the research scope to include the noisy dataset into the imbalanced classification problem. A model fusion based framework, K Nearest Gaussian (KNG) is proposed. KNG employs a generative modeling method, GMM, to model the training data as Gaussian mixtures and form adjustable confidence regions which are less sensitive to data imbalance and noise. Motivated by the K-nearest neighbor algorithm, the neighboring Gaussians are used to classify the testing instances. Experimental results show KNG method greatly outperforms traditional classification methods in dealing with imbalanced classification problems with noisy dataset.
The phase III study addresses the issues of feature selection and parameter tuning of KNG algorithm. To further improve the performance of KNG algorithm, a Particle Swarm Optimization based method (PSO-KNG) is proposed. PSO-KNG formulates model parameters and data features into the same particle vector and thus can search the best feature and parameter combination jointly. The experimental results show that PSO can greatly improve the performance of KNG with better accuracy and much lower computational cost.
ContributorsHe, Miao (Author) / Wu, Teresa (Thesis advisor) / Li, Jing (Committee member) / Silva, Alvin (Committee member) / Borror, Connie (Committee member) / Arizona State University (Publisher)
Created2014