<![CDATA[Ph.D. Dissertation Defense - Xinhui Li]]>28475Title: Data-Driven, Multi-View, and Multimodal Representation Learning for Neuroimaging
Committee:
Dr. Vince Calhoun, ECE, Chair, Advisor
Dr. Rogers Silva, TReNDS, Co-Advisor
Dr. Christopher Rozell, ECE
Dr. Anqi Wu, CSE
Dr. Shella Keilholz, BME
Dr. Tulay Adali, U of Maryland
]]> Daniela Staiculescu117846558892026-07-21 17:44:4917846559262026-07-21 17:45:2600eventMental disorders affect more than one billion people worldwide, and timely intervention is critical; however, current diagnostic practice remains largely subjective. Magnetic resonance imaging (MRI) provides a noninvasive way to measure brain structure and function and identify objective biomarkers of mental disorders. Despite these advances, several challenges remain in mental disorder diagnosis and neuroimaging analysis, including objective characterization of the neuropsychiatric continuum and heterogeneity, preprocessing-related variability, and heterogeneous information integration from high-dimensional, multimodal data. This dissertation develops data-driven, multi-view, and multimodal representation learning approaches, which learn low-dimensional representations from high-dimensional MRI data, to characterize the neuropsychiatric continuum and heterogeneity, mitigate preprocessing-related variability, and identify phenotypic and psychiatric biomarkers from structural and functional MRI. For unsupervised representation learning, our data-driven interpolation framework effectively characterizes individual differences within a group and continuous patterns between groups. For multi-view representation learning, our methods substantially improve both neural network representational similarity across preprocessing pipelines and prediction robustness for brain-phenotype relationships. For multimodal representation learning, our multimodal latent variable models, developed in the MISA PyTorch framework, successfully identify linked sources associated with phenotypic and psychiatric measures. Together, our methods and toolboxes contribute to reproducible neuroimaging analysis and reliable brain-behavior relationship discovery.]]>2026-07-28T10:00:00-04:002026-07-28T12:00:00-04:002026-07-28T12:00:00-04:002026-07-28 14:00:002026-07-28 16:00:002026-07-28 16:00:002026-07-28T10:00:00-04:002026-07-28T12:00:00-04:00America/New_YorkAmerica/New_Yorkdatetime2026-07-28 10:00:002026-07-28 12:00:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]><![CDATA[Zoom Link ]]><![CDATA[Ph.D. Dissertation Defense - Jonas Theumer]]>28475Title: Electromagnetic Sensing for Monitoring Physical and Embedded Systems
Committee:
Dr. Milos Prvulovic, CoC, Chair, Advisor
Dr. Morris Cohen, ECE
Dr. Ashutosh Dhekne, CoC
Dr. Gregory Durgin, ECE
Dr. Moinuddin Qureshi, CoC
]]> Daniela Staiculescu117846554362026-07-21 17:37:1617846554942026-07-21 17:38:1400eventElectromagnetic (EM) sensing is a powerful tool to non-invasively monitor internal states of systems that are otherwise hard to access. This thesis leverages EM sensing across two distinct domains: biological plants and embedded systems. First, we perform EM modeling and measurements to non-destructively track plant water content. We show that Ka-band EM transmission loss strongly correlates with leaf water content. To overcome the practical limitations of transmission measurements, we develop a model and measurement technique with a single antenna that tracks the loss due to leaf water content via backscattering. The second part of this thesis shifts the focus to systems that are themselves sources of EM signals that can be used to reveal internal operational states. Specifically, we demonstrate how to accurately classify serial protocols used by an embedded device for internal communications via unintentional EM emanations. Measuring the magnetic near field, we are able to distinguish different protocols, data rates, and even message contents with high accuracy. We are also able to perform similar classification using far-field data recorded at lower sampling rates in a noisy laboratory environment. Finally, we consider an integrated embedded platform periodically executing tasks. Instead of passively sensing EM leakage, we create the side channel by intentionally transmitting an EM wave toward the device. Just as the backscatter reflecting off leaves reveals the internal state of the underlying plant, the embedded device modulates the incident wave based on its physical operation. We are able to accurately profile the embedded platform in an unsupervised fashion by analyzing the backscatter. We further characterize the backscattering link under various configurations, and find that our unsupervised framework is highly robust to changes in the backscattering link. This approach can serve as a foundation for profiling inaccessible embedded systems. Ultimately, this thesis demonstrates how EM sensing can serve as a tool to non-invasively obtain hard to access information across disciplines.]]>2026-07-27T13:00:00-04:002026-07-27T15:00:00-04:002026-07-27T15:00:00-04:002026-07-27 17:00:002026-07-27 19:00:002026-07-27 19:00:002026-07-27T13:00:00-04:002026-07-27T15:00:00-04:00America/New_YorkAmerica/New_Yorkdatetime2026-07-27 01:00:002026-07-27 03:00:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]><![CDATA[Ph.D. Dissertation Defense - Zexian Zeng]]>28475Title: Modeling, Stability, and Resilience of Voltage-Source Converters in Hybrid AC/DC Power Systems
Committee:
Dr. Maryam Saeedifard, ECE, Chair, Advisor
Dr. Deepak Divan, ECE
Dr. Santiago Grijalva, ECE
Dr. Daniel Molzahn, ECE
Dr. Qichen Yang, UCF
]]> Daniela Staiculescu117846552892026-07-21 17:34:4917846553612026-07-21 17:36:0100eventVoltage-source converters (VSCs) are becoming the dominant interface in modern power systems as renewable generation, energy storage, high-voltage DC transmission, medium-voltage DC networks, electric-vehicle charging, and large flexible loads continue to expand. In VSC-dominated hybrid AC/DC power systems, stability and resilience are increasingly determined by converter controls rather than by the inherent electromechanical behavior of synchronous machines. This dissertation develops modeling, stability-analysis, and resilience-oriented control methods for VSC-based hybrid AC/DC power systems. The first part of the dissertation investigates small-signal AC/DC dynamic coupling across different AC- and DC-terminal control roles, including AC grid-forming/DC grid-following, AC grid-following/DC grid-forming, and dual-port grid-forming operation. The analysis shows how DC-link dynamics, modulation-reference choices, DC-network dynamics, and dual-port control parameters affect oscillatory modes and stability margins. Modular single-input single-output modeling frameworks and physical damping interpretations are developed to support scalable and design-oriented stability analysis. The second part studies large-signal resilience of P-f droop-based AC grid-forming/DC grid-following VSCs under faults, unbalanced black start, load pickup, motor startup, and protection-device interactions. Current-limiting methods, generalized three-phase droop control, distributed load restoration, and converter-aware protection behavior are examined to understand how current-limited VSCs can support practical grid functions during severe disturbances. Together, the results provide control-aware modeling tools, stability insights, and resilience-oriented design guidelines for future hybrid AC/DC grids with high penetrations of VSC-interfaced resources.]]>2026-07-23T14:30:00-04:002026-07-23T16:30:00-04:002026-07-23T16:30:00-04:002026-07-23 18:30:002026-07-23 20:30:002026-07-23 20:30:002026-07-23T14:30:00-04:002026-07-23T16:30:00-04:00America/New_YorkAmerica/New_Yorkdatetime2026-07-23 02:30:002026-07-23 04:30:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]><![CDATA[Ph.D. Dissertation Defense - Pranjal Gajare]]>28475Title: Universal Interoperable Control Framework for Inverter-Based-Resources
Committee:
Dr. Deepak Divan, ECE, Chair, Advisor
Dr. Maryam Saeedifard, ECE, Co-Advisor
Dr. Santiago Grijalva, ECE
Dr. Daniel Molzahn, ECE
Dr. Bayoun Ge, ECE
Dr. Dominic Gross, UW-Madison
]]> Daniela Staiculescu117846533102026-07-21 17:01:5017846533972026-07-21 17:03:1700eventPower systems are transitioning from synchronous generator (SG) dominated operation to fleets of inverter-based resources (IBRs), with grid-forming (GFM) control as the leading candidate for preserving voltage-source behavior. Yet GFM inverters deviate from the ideal voltage source in ways that undermine stability, spanning non-passive high-frequency behavior, poorly damped line-frequency oscillations, loss of synchronism under current limiting, and switching-induced distortion. Moreover, inverters from different vendors lack a behavioral contract, hindering stable interoperation at scale. This thesis addresses both problems. At the device level, it develops and experimentally validates controllers that resolve each instability mechanism, including delay compensation, angle restoration control, and power-angle-aware current limiting. At the system level, it synthesizes these results into a rules-and-signaling framework: a minimal, explicit behavioral contract under which GFM IBRs from different vendors interoperate stably without centralized coordination, much as SGs converged on droop and inertia over decades. The result is a path toward certifiable, vendor-heterogeneous GFM fleets.]]>2026-07-23T12:00:00-04:002026-07-23T14:00:00-04:002026-07-23T14:00:00-04:002026-07-23 16:00:002026-07-23 18:00:002026-07-23 18:00:002026-07-23T12:00:00-04:002026-07-23T14:00:00-04:00America/New_YorkAmerica/New_Yorkdatetime2026-07-23 12:00:002026-07-23 02:00:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]><![CDATA[Microsoft Teams Link ]]><![CDATA[Ph.D. Dissertation Defense - Samuel Medeiros Araujo Morais]]>28475Title: Ultrasound Shear Wave Elastography Methods for Tissue Therapy Guidance and Monitoring
Committee:
Dr. Stanislav Emelianov, ECE, Chair, Advisor
Dr. Brooks Lindsey, BME
Dr. Scott Hollister, BME
Dr. Levent Degertekin, ECE
Dr. Francisco Robles, BME
]]> Daniela Staiculescu117839298812026-07-13 08:04:4117839299742026-07-13 08:06:1400eventThe biomechanical properties of biological tissues are closely related to their physiological state, including inflammation and the development of disease. Since ancient times, clinicians have assessed tissue elasticity through palpation, making it one of the oldest diagnostic techniques in medicine. Although palpation remains a valuable clinical screening tool for detecting superficial abnormalities, the advent of medical imaging has transformed tissue elasticity assessment into a noninvasive modality capable of visualizing and quantifying tissue mechanical properties. Ultrasound elastography, in particular, combines the advantages of conventional ultrasound imaging, such as portability, real-time imaging, and cost-effectiveness, with the ability to quantitatively assess tissue mechanical properties by measuring shear wave propagation. These measurements enable not only the detection and monitoring of disease progression but also the guidance of therapeutic interventions and the evaluation of tissue response to treatment. This research expands the application of ultrasound shear wave elastography techniques to several clinically relevant challenges. In aim 1, a miniaturized source of longitudinal shear wave is investigated to enable intracardiac measurement of myocardial stiffness for heart failure therapy guidance. In aim 2, shear wave elasticity imaging is used to detect early biomechanical changes associated with skin breakdown and implant exposure following subcutaneous implantation. Finally, in aim 3, shear wave elasticity imaging is applied to assess the risk of reherniation after patch repair of congenital diaphragmatic hernia.]]>2026-07-28T14:00:00-04:002026-07-28T16:00:00-04:002026-07-28T16:00:00-04:002026-07-28 18:00:002026-07-28 20:00:002026-07-28 20:00:002026-07-28T14:00:00-04:002026-07-28T16:00:00-04:00America/New_YorkAmerica/New_Yorkdatetime2026-07-28 02:00:002026-07-28 04:00:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]><![CDATA[Zoom link]]><![CDATA[Ph.D. Dissertation Defense - Afra Nawar]]>28475Title: Enabling Robust Ambulatory Cardiovascular Monitoring inthe Context of Stress for Those with a Prior Myocardial Infarction
Committee:
Dr. Omer Inan, ECE, Chair, Advisor
Dr. Amit Shah, Emory
Dr. Mark Davenport, ECE
Dr. Matthieu Bloch, ECE
Dr. Douglas Bremmer, Emory
]]> Daniela Staiculescu117839282472026-07-13 07:37:2717839283152026-07-13 07:38:3500eventOne in four people who have experienced a myocardial infarction (MI) will go on to experience a second event in their lifetime, often resulting in early mortality. Despite this known high incidence of adverse outcomes, preventative monitoring for this population remains largely absent. In particular, there is a lack of continuous and objective assessment of the capability of the patient's cardiovascular system to withstand and adapt to daily stressors following an MI. This dissertation describes biosignal processing and algorithm development efforts towards enabling reliable ambulatory cardiovascular monitoring in the context of stress for this population. Such a system would be able to automatically identify abnormal stress-induced cardiovascular changes remotely, enabling preventative care. Towards this goal, we make the following key contributions through the research described. We first demonstrate the feasibility of quantifying complex stress-induced changes in cardiomechanical function, vascular dynamics, and eletrophysiology using a compact multimodal wearable patch, previously only possible through a disjoint collection of obtrusive wired devices. We then make several advances towards improving the usability of these signals. We validate that ECG features can be derived from the patch with equivalent fidelity to gold-standard wired devices. Utilizing the unique morphology of the SCG signal, we develop a novel TDA-based algorithm with potential to enable robust SCG signal quality assessment without reliance on an reference signal. We additionally take the first step towards the understanding the physiological origins and signal quality of core-body PPG signals as compared to peripheral measurement sites. Importantly, we find that the core-body PPG amplitude reacts in an opposite fashion to that of the peripheral site during respiratory challenges. Finally, using these insights, we create the first automated algorithm with capability to identify abnormal hemodynamic responses to stress in the post-MI population, uncovering novel potential SCG-based biomarkers of cardiac dysfunction in the process. Thus, this work presents a step towards continuous cardiovascular monitoring in at-home settings, which can provide clinicians with real-time physiological information to enable early, personalized clinical intervention for the growing population of post-MI survivors.]]>2026-07-24T14:00:00-04:002026-07-24T16:00:00-04:002026-07-24T16:00:00-04:002026-07-24 18:00:002026-07-24 20:00:002026-07-24 20:00:002026-07-24T14:00:00-04:002026-07-24T16:00:00-04:00America/New_YorkAmerica/New_Yorkdatetime2026-07-24 02:00:002026-07-24 04:00:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]><![CDATA[Zoom link]]>