28068 research outputs found
Sort by
Retrospective Quantitative T1 Imaging to Examine Characteristics of Multiple Sclerosis Lesions
No full textQuantitative MRI plays an essential role in assessing tissue abnormality and diseaseprogression in multiple sclerosis (MS). Specifically, T1 relaxometry is gaining popularity
as elevated T1 values have been shown to correlate with increased inflammation,
demyelination, and gliosis. The predominant issue is that relaxometry requires parametric
mapping through advanced imaging techniques not commonly included in standard clinical
protocols. This leaves an information gap in large clinical datasets from which quantitative
mapping could have been performed.
We introduce T1-REQUIRE, a retrospective T1 mapping method that approximates
T1 values from a single T1-weighted MR image. This method has already been shown to
be accurate within 10% of a clinically available reference standard in healthy controls but
will be further validated in MS cohorts. We also further aim to determine T1-REQUIRE’s
statistical significance as a unique biomarker for the assessment of MS lesions as they
relate to clinical disability and disease burden.
A 14-subject comparison between T1-REQUIRE maps derived from 3D T1
weighted turbo field echoes (3D T1w TFE) and an inversion-recovery fast field echo (IRFFE) revealed a whole-brain voxel-wise Pearson’s correlation of r = 0.89 (p < 0.001) and
mean bias of 3.99%. In MS white matter lesions, r = 0.81, R2 = 0.65 (p < 0.001, N = 159),
bias = 10.07%, and in normal appearing white matter (NAWM), r = 0.82, R
2 = 0.67 (p <
0.001), bias = 9.48%.
Mean lesional T1-REQUIRE and MTR correlated significantly (r = -0.68, p <
0.001, N = 587) similar to previously published literature. Median lesional MTR correlated
significantly with EDSS (rho = -0.34, p = 0.037), and lesional T1-REQUIRE exhibited
xiii
significant correlations with global brain tissue atrophy as measured by brain parenchymal
fraction (BPF) (r = -0.41, p = 0.010, N = 38). Multivariate linear regressions between T1-
REQUIRE NAWM provided meaningful statistical relationships with EDSS (β = 0.03, p
= 0.027, N = 38), as well as did mean MTR values in the Thalamus (β = -0.27, p = 0.037,
N = 38).
A new spoiled gradient echo variation of T1-REQUIRE was assessed as a proof of concept
in a small 5-subject MS cohort compared with IR-FFE T1 maps, with a whole brain voxel-wise
correlation of r = 0.88, R2 = 0.77 (p < 0.001), and Bias = 0.19%. Lesional T1 comparisons
reached a correlation of r = 0.75, R2 = 0.56 (p < 0.001, N = 42), and Bias = 10.81%.
The significance of these findings means that there is the potential to provide
supplementary quantitative information in clinical datasets where quantitative protocols
were not implemented. Large MS data repositories previously only containing structural
T1 weighted images now may be used in big data relaxometric studies with the potential
to lead to new findings in newly uncovered datasets. Furthermore, T1-REQUIRE has the
potential for immediate use in clinics where standard T1 mapping sequences aren’t able to
be readily implemented
Co-design Framework for Design and Optimization of Switched Reluctance Motors
No full textThis thesis developed a co-design optimization framework for Switched Reluctance Motors (SRM) where geometric and control parameters are jointly optimized to improve the overall performance of the motor. This algorithm aims to increase the average torque, decrease the ripple of the torque profile, and enhance operating efficiency as well as acoustic noise. This co-design methodology allows the motor electromagnetic design to be combined with the control strategy and presents a solution for multi-physics optimization. A magnetic equivalent circuit (MEC) based approach is proposed to analyze the electromagnetic characteristics which greatly improves the accuracy over conventional MEC for arbitrary geometries. The natural frequencies of the stator system are identified using a system approach, which is suitable for motors with complex frames. The vibration response and acoustic noise prediction are analyzed and evaluated by the analytical approach.
To demonstrate the effectiveness of this approach, case studies were performed that compare optimized SRM designs obtained from their framework to those designed traditionally. Experimental validation is performed to evaluate the accuracy of the proposed approach
Defense-in-Depth for Cyber-Secure Network Architectures of Industrial Control Systems
No full textDigitization and modernization efforts have yielded greater efficiency, safety, and cost-savings for Industrial Control Systems (ICS). To achieve these gains, the Internet of Things (IoT) has become an integral component of network infrastructures. However, integrating embedded devices expands the network footprint and softens cyberattack resilience. Additionally, legacy devices and improper security configurations are weak points for ICS networks. As a result, ICSs are a valuable target for hackers searching for monetary gains or planning to cause destruction and chaos. Furthermore, recent attacks demonstrate a heightened understanding of ICS network configurations within hacking communities. A Defense-in-Depth strategy is the solution to these threats, applying multiple security layers to detect, interrupt, and prevent cyber threats before they cause damage. Our solution detects threats by deploying an Enhanced Data Historian for Detecting Cyberattacks. By introducing Machine Learning (ML), we enhance cyberattack detection by fusing network traffic and sensor data. Two computing models are examined: 1) a distributed computing model and 2) a localized computing model. The distributed computing model is powered by Apache Spark, introducing redundancy for detecting cyberattacks. In contrast, the localized computing model relies on a network traffic visualization methodology for efficiently detecting cyberattacks with a Convolutional Neural Network. These applications are effective in detecting cyberattacks with nearly 100% accuracy.
Next, we prevent eavesdropping by applying Homomorphic Encryption for Secure Computing. HE cryptosystems are a unique family of public key algorithms that permit operations on encrypted data without revealing the underlying information. Through the Microsoft SEAL implementation of the CKKS algorithm, we explored the challenges of introducing Homomorphic Encryption to real-world applications. Despite these challenges, we implemented two ML models: 1) a Neural Network and 2) Principal Component Analysis.
Finally, we hinder attackers by integrating a Cyberattack Lockdown Network with Secure Ultrasonic Communication. When a cyberattack is detected, communication for safety-critical elements is redirected through an ultrasonic communication channel, establishing physical network segmentation with compromised devices. We present proof-of-concept work in transmitting video via ultrasonic communication over an Aluminum Rectangular Bar. Within industrial environments, existing piping infrastructure presents an optimal solution for cost-effectively preventing eavesdropping. The effectiveness of these solutions is discussed within the scope of the nuclear industry
Nanopore sensing for environmental and biomarker analysis
No full textNanopore stochastic sensing is a powerful analytical tool for detecting target molecules through a nanoscale pore. The analyte and electrolyte ions are subjected to a voltage bias which drives them to translocate through the nanopore, resulting in disruptions in the ionic current. These disruptions are translated to blockage events which can serve as a signature of the analyte. Owing to its unique features of single-molecule and label-free sensing, nanopore technique has been exploited in a wide array of applications such as detection of metal ions, proteins, DNA, microRNA, toxic agents etc. In this dissertation, projects showcasing nanopore’s sensing capability of different biomarkers and in the detection of a wide range of target molecules based on non-covalent interactions are presented. Particularly in the first two projects, nanopore detection of ferric ions relevant to environmental regulation as well as a biomarker for human health and a miRNA-based biomarker for oral cancer and oral related diseases are summarized. Ferric ions, which are benign if present in balanced quantities but can be toxic otherwise, are detected by using an engineered multifunctional nanopore and a chelating organophosphonic acid ligand. The chelate complex formed after ferric ions bind to ligand gives significantly different event signatures than the free ligand in the solution enabling ferric ion detection. Even in the presence of interfering ions, the ferric ions could be recognized easily because of the conformational changes brought in the nanopore lumen by the interaction of the interfering metal ions with the His-tags of the nanopore which in turn resulted in variations in the characteristics of blocking events. In the second project, miR31, an oral cancer biomarker, is selectively detected with the help of an engineered nanopore, and a DNA based probe. Several probes with variations in length, composition and position of the overhangs or probes with no overhangs were compared and studied as the probes play a crucial role in capturing the target of interest with high specificity. Our strategically designed probe emerged as the most effective in capturing the target even in presence of large background from human saliva samples and enhanced the sensitivity of the system. In the first two projects, nanopores are utilized for selective and specific detection of certain target molecules. However, in order to analyze diverse range of analytes, numerous sensing systems have to be constructed which can be a time-consuming and challenging task. To circumvent this limitation, in the third project, diverse recognition sites based on various non-covalent interactions are incorporated into the α-hemolysin protein pore to achieve detection of not just a single analyte but broad category of molecules such as cations, anions, aromatic and hydrophobic compounds
Resolvent analysis of turbulent flows: Extensions, improvements and applications
No full textThis thesis presents several advances in both physics-based and data-driven modeling of turbulent fluid flows. In particular, the present thesis focuses on resolvent analysis, a physics-based framework that identifies the coherent structures that are most amplified by the Navier-Stokes equations when they are linearized about a known turbulent mean flow via a singular value decomposition (SVD) of a discretized operator. This method has proven to effectively capture energetically-relevant features observed in various flows. However, it has some shortcomings that the present work intends to alleviate. First, the original formulation of resolvent analysis is restricted to statistically-stationary or time-periodic mean flows. To expand the applicability of this framework, this thesis presents a spatiotemporal variant of resolvent analysis that is able to account for time-varying systems. Moreover, sparsity (which manifests in localization) is also incorporated to the analysis through the addition of an l1-norm penalization term to the optimization associated with the SVD. This allows for the identification of energetically-relevant coherent structures that correspond to spatio-temporally localized amplification mechanisms, for flows with either a time-varying or stationary mean. The high computational cost associated with the discretization and analysis of a large discretized of the mean-linearized Navier-Stokes operator represents the second drawback of resolvent analysis. As a second contribution, this thesis provides an analytic form of resolvent analysis for planar flows based on wavepacket pseudomode theory, avoiding the numerical computations required in the original framework. The third contribution focuses on the characterization of the energetically-dominant coherent structures that arise in turbulent flow traveling through straight ducts with square and rectangular cross-sections. First, resolvent analysis is applied to predict the coherent structures that arise in this flow, and to study the sensitivity of this methodology to the secondary mean flow components that display a distinct pattern near the duct corners. Next, a data-driven causality analysis is performed to understand the physical mechanisms involved in the evolution of coherent structures near the duct corners. To do this, a nonlinear Granger causality analysis method is developed and applied to proper orthogonal decomposition coefficients of direct numerical simulation data, revealing that the structures associated with the secondary velocity components are behind the formation and translation of the near-wall and near-corner streamwise structures. A general discussion and future prospects are discussed at the end of this thesis
Defense-in-Depth for Cyber-Secure Network Architectures of Industrial Control Systems
No full textDigitization and modernization efforts have yielded greater efficiency, safety, and cost-savings for Industrial Control Systems (ICS). To achieve these gains, the Internet of Things (IoT) has become an integral component of network infrastructures. However, integrating embedded devices expands the network footprint and softens cyberattack resilience. Additionally, legacy devices and improper security configurations are weak points for ICS networks. As a result, ICSs are a valuable target for hackers searching for monetary gains or planning to cause destruction and chaos. Furthermore, recent attacks demonstrate a heightened understanding of ICS network configurations within hacking communities. A Defense-in-Depth strategy is the solution to these threats, applying multiple security layers to detect, interrupt, and prevent cyber threats before they cause damage. Our solution detects threats by deploying an Enhanced Data Historian for Detecting Cyberattacks. By introducing Machine Learning (ML), we enhance cyberattack detection by fusing network traffic and sensor data. Two computing models are examined: 1) a distributed computing model and 2) a localized computing model. The distributed computing model is powered by Apache Spark, introducing redundancy for detecting cyberattacks. In contrast, the localized computing model relies on a network traffic visualization methodology for efficiently detecting cyberattacks with a Convolutional Neural Network. These applications are effective in detecting cyberattacks with nearly 100% accuracy.
Next, we prevent eavesdropping by applying Homomorphic Encryption for Secure Computing. HE cryptosystems are a unique family of public key algorithms that permit operations on encrypted data without revealing the underlying information. Through the Microsoft SEAL implementation of the CKKS algorithm, we explored the challenges of introducing Homomorphic Encryption to real-world applications. Despite these challenges, we implemented two ML models: 1) a Neural Network and 2) Principal Component Analysis.
Finally, we hinder attackers by integrating a Cyberattack Lockdown Network with Secure Ultrasonic Communication. When a cyberattack is detected, communication for safety-critical elements is redirected through an ultrasonic communication channel, establishing physical network segmentation with compromised devices. We present proof-of-concept work in transmitting video via ultrasonic communication over an Aluminum Rectangular Bar. Within industrial environments, existing piping infrastructure presents an optimal solution for cost-effectively preventing eavesdropping. The effectiveness of these solutions is discussed within the scope of the nuclear industry
Large Language Model Based Machine Learning Techniques for Fake News Detection
No full textWith advanced technology, it’s widely recognized that everyone owns one or more personal devices. Consequently, people are evolving into content creators on social media or the streaming platforms sharing their personal ideas regardless of their education or expertise level. Distinguishing fake news is becoming increasingly crucial. However, the recent research only presents comparisons of detecting fake news between one or more models across different datasets. In this work, we applied Natural Language Processing (NLP) techniques with Naïve Bayes and DistilBERT machine learning method combing and augmenting four datasets. The results show that the balanced accuracy is higher than the average in the recent studies. This suggests that our approach holds for improving fake news detection in the era of widespread content creation
The Double-edged Sword of Executive Pay: How the CEO-TMT Pay Gap Influences Firm Performance
No full textThis study examines the relationship between the chief executive officer (CEO) and top management team (TMT) pay gap and consequent firm performance. Drawing on tournament theory and equity theory, I argue that the effect of the CEO-TMT pay gap on consequent firm performance is non-monotonic. Using data from 1995 to 2022 from S&P 1500 US firms, I explicate an inverted U-shaped relationship, such that an increase in the pay gap leads to an increase in firm performance up to a certain point, after which it declines. Additionally, multilevel analyses reveal that this curvilinear relationship is moderated by attributes of the TMT, and the industry in which the firm competes. My findings show that firms with higher TMT gender diversity suffer lower performance loss due to wider pay gaps. Furthermore, when firm executives are paid more compared to the industry norms, or when the firm has a long-tenured CEO, firm performance becomes less sensitive to larger CEO-TMT pay gaps. Lastly, when the firm competes in a masculine industry, firm performance is more negatively affected by larger CEO-TMT pay gaps. Contrary to my expectations, firm gender-diversity friendly policies failed to influence the CEO-TMT pay gap-firm performance relationship
The Double-edged Sword of Executive Pay: How the CEO-TMT Pay Gap Influences Firm Performance
No full textThis study examines the relationship between the chief executive officer (CEO) and top management team (TMT) pay gap and consequent firm performance. Drawing on tournament theory and equity theory, I argue that the effect of the CEO-TMT pay gap on consequent firm performance is non-monotonic. Using data from 1995 to 2022 from S&P 1500 US firms, I explicate an inverted U-shaped relationship, such that an increase in the pay gap leads to an increase in firm performance up to a certain point, after which it declines. Additionally, multilevel analyses reveal that this curvilinear relationship is moderated by attributes of the TMT, and the industry in which the firm competes. My findings show that firms with higher TMT gender diversity suffer lower performance loss due to wider pay gaps. Furthermore, when firm executives are paid more compared to the industry norms, or when the firm has a long-tenured CEO, firm performance becomes less sensitive to larger CEO-TMT pay gaps. Lastly, when the firm competes in a masculine industry, firm performance is more negatively affected by larger CEO-TMT pay gaps. Contrary to my expectations, firm gender-diversity friendly policies failed to influence the CEO-TMT pay gap-firm performance relationship
Resolvent analysis of turbulent flows: Extensions, improvements and applications
No full textThis thesis presents several advances in both physics-based and data-driven modeling of turbulent fluid flows. In particular, the present thesis focuses on resolvent analysis, a physics-based framework that identifies the coherent structures that are most amplified by the Navier-Stokes equations when they are linearized about a known turbulent mean flow via a singular value decomposition (SVD) of a discretized operator. This method has proven to effectively capture energetically-relevant features observed in various flows. However, it has some shortcomings that the present work intends to alleviate. First, the original formulation of resolvent analysis is restricted to statistically-stationary or time-periodic mean flows. To expand the applicability of this framework, this thesis presents a spatiotemporal variant of resolvent analysis that is able to account for time-varying systems. Moreover, sparsity (which manifests in localization) is also incorporated to the analysis through the addition of an l1-norm penalization term to the optimization associated with the SVD. This allows for the identification of energetically-relevant coherent structures that correspond to spatio-temporally localized amplification mechanisms, for flows with either a time-varying or stationary mean. The high computational cost associated with the discretization and analysis of a large discretized of the mean-linearized Navier-Stokes operator represents the second drawback of resolvent analysis. As a second contribution, this thesis provides an analytic form of resolvent analysis for planar flows based on wavepacket pseudomode theory, avoiding the numerical computations required in the original framework. The third contribution focuses on the characterization of the energetically-dominant coherent structures that arise in turbulent flow traveling through straight ducts with square and rectangular cross-sections. First, resolvent analysis is applied to predict the coherent structures that arise in this flow, and to study the sensitivity of this methodology to the secondary mean flow components that display a distinct pattern near the duct corners. Next, a data-driven causality analysis is performed to understand the physical mechanisms involved in the evolution of coherent structures near the duct corners. To do this, a nonlinear Granger causality analysis method is developed and applied to proper orthogonal decomposition coefficients of direct numerical simulation data, revealing that the structures associated with the secondary velocity components are behind the formation and translation of the near-wall and near-corner streamwise structures. A general discussion and future prospects are discussed at the end of this thesis