12023 research outputs found
Sort by
Black Hole Perturbation Theory Beyond General Relativity and Holographic Gravity in Flat Spacetime
Get PDFIn this thesis, we study two topics in using gravitational waves (GWs) to probe fundamental physics. The first topic is using black hole (BH) perturbation theory to model GW emissions by binary BH mergers in gravity theories beyond Einstein's general relativity (GR). The second topic is studying holographic quantum gravity signatures around interferometers in flat spacetime.
For BH perturbation theory beyond GR, we first construct a novel formalism based on Teukolsky's seminal work in the 1970s. Our modified Teukolsky formalism works for BHs with arbitrary spin in a broad class of beyond-GR theories that admit an effective field theory description. We derive this formalism by following Chandrasekhar's prescription to make some convenient gauge choices, under which different degrees of curvature perturbations naturally decouple. In the end, we get two decoupled and potentially separable second-order partial differential equations for the Weyl scalars Psi0 and Psi4, representing the ingoing and outgoing gravitational radiations of a perturbed BH, respectively. Our formalism works for both linear and nonlinear orders in the beyond-GR couplings. We then apply our formalism to specific examples.
In the first example, we study the isospectrality breaking of quasinormal modes (QNMs) in beyond-GR theories, where the even- and odd-parity QNMs have different frequencies. We apply the modified Teukolsky formalism and the eigenvalue perturbation method to construct a direct connection between the parity features of a theory and its structure of isospectrality breaking. In the second example, we focus on the QNMs of dynamical Chern-Simons gravity up to the first order in the slow-rotation expansion. We first directly compute the scalar field equation and the modified Teukolsky equations for Psi0 and Psi4 in the ingoing and outgoing radiation gauges, respectively. We then reduce all the equations to radial ordinary differential equations by projection to the spin-weighted spheroidal harmonics. We find that the scalar field is only coupled to the odd-parity perturbations, which is consistent with the previous studies. We then compute the QNM frequencies for the non-rotating case via the eigenvalue perturbation method. The results from the two gauges are self-consistent and agree well with previous results using metric perturbations. Since this is ongoing work, we briefly discuss the strategy for the rotating case at the end. In the third example, we apply a similar analysis to certain parametrized axisymmetric deviations of non-rotating BHs using a Weyl multipole expansion. We compute the QNM frequencies directly and analyze their connections to the multipole structure of a BH spacetime.
For holographic gravity in flat spacetime, we build an effective model for geometrical spacetime fluctuations driven by entropic fluctuations, or "geontropic fluctuations" for short, in the casual diamond defined by an interferometer. Our model involves a bosonic scalar field with some nontrivial occupation number, called "pixellon." The pixellon field characterizes all the nonlinear holographic quantum gravity fluctuations within a causal diamond in flat spacetime. We then build up a framework for computing the gauge-invariant observables of geontropic fluctuations for an interferometer with equal arms separated by arbitrary angles. We compute both the power spectral density and angular correlation of length fluctuations in such an interferometer for the pixellon model. We then use the existing or predicted noise spectra of LVK, LISA, GEO-600, and Holometer to constrain the pixellon model. In our follow-up study, we further extend the pixellon model to incorporate configurations of multiple interferometers. We then apply this extended pixellon model to calculate the power spectral density of geontropic fluctuations in Cosmic Explorer, Einstein Telescope, NEMO, and optically-levitated sensors. For Cosmic Explorer, Einstein Telescope, and NEMO, we find that the signal of the pixellon model could exceed the detector's predicted sensitivity by one or two orders of magnitudes.</p
Asymmetric Total Synthesis of Havellockate and Investigation into Chiral Palladium Enolate: Synthesis, Reactivity, and Applications
No full textResearch in the Stoltz group focuses on the synergy of complex natural product synthesis and method development in that we strive to invent new and efficient methodologies that have great synthetic potential in the use for pharmaceuticals and natural products. Herein we describe an asymmetric total synthesis of Havellockate, a polycyclic furanobutenolide-derived cembranoid diterpenoid that exhibits biological activities such as anti-inflammatory, anti-microbial, and cytotoxic. The strategy for this synthesis is highlighted by a convergent Julia–Kocienski olefination, followed by an acylation/intramolecular [4+2] cycloaddition cascade, which furnishes the main core of the natural product in high efficiency.
Another synthetic route toward the total synthesis of Havellocate is presented, using a propargyl ether as a key intermediate. Though the route was unfruitful in compleing the synthesis, the Diels–Alder cascade had significant improvement of yield and stability with this route, and it constitutes tremendous value for the synthesis of other targets within the furanobutenolide-derived natural product family.
Next, the synthesis, isolation, and reactivity of a chiral Pd enolate is described. The Pd enolate is arose by an alpha bromo acetophenone oxidative addition complex with Pd2(dba)3 and PHOX ligand. A crystal structure is obtained to show that the enolate is C- bound and highly regioselective. The novelty of this isolation can shine light on the applications of such Pd enolate for future developments.
Then, we outlined a Pd-catalyzed asymmetric vinylation of γ-lactams to construct all-carbon quaternary stereocenters. The use of canonically inactive vinyl chloride electrophiles afforded the highest yields and levels of stereoselectivity, and a range of tri-substituted vinyl chlorides were found to be proficient in promoting this transformation. These stereogenically congested products could be further elaborated to functionally rich scaffolds, proving the synthetic utility of this transformation.
Lastly, we describe the work of organizing the inaugural Day of Inclusion event of CCE. The event was orchestrated by the Diversity in Chemistry Initiative (DICI), aimed to foster cohesion and to inspire concerted efforts towards Diversity, Equity, and Inclusivity (DEI) within the Chemistry and Chemical Engineering (CCE) division at Caltech.</p
Revealing Regulatory Network Organization Through Single-Cell Perturbation Profiling and Maximum Entropy Models
Get PDFGene regulatory networks within cells modulate the expression of the genome in response to signals and changing environmental conditions. Reconstructions of gene regulatory networks can reveal the information processing and control principles used by cells to maintain homeostasis and execute cell-state transitions. In this thesis, we introduce a computational framework, D-SPIN, that generates quantitative models of gene regulatory networks from single-cell mRNA-seq datasets collected across thousands of distinct perturbation conditions. D-SPIN models the cell as a collection of interacting gene-expression programs, and constructs a probabilistic model to infer regulatory interactions between gene-expression programs and external perturbations. Using large Perturb-seq and drug-response datasets, we demonstrate that D-SPIN models reveal the organization of cellular pathways, sub-functions of macromolecular complexes, and the logic of cellular regulation of transcription, translation, metabolism, and protein degradation in response to gene knockdown perturbations. D-SPIN can also be applied to dissect drug response mechanisms in heterogeneous cell populations, elucidating how combinations of immunomodulatory drugs can induce novel cell states through additive recruitment of gene expression programs. D-SPIN provides a computational framework for constructing interpretable models of gene-regulatory networks to reveal principles of cellular information processing and physiological control
Asymmetric Pericyclic Transformations from Reactive Palladium Intermediates
Get PDFThe Pd-catalyzed decarboxylative asymmetric allylic alkylation of enolate nucleophiles is a cornerstone of our groups’ efforts to develop methodologies that directly facilitate the synthesis of stereochemically complex molecular building blocks. This thesis first focuses on our efforts to deepen our mechanistic understanding of these transformations. We then employ our insights as a base from which we expand the scope of the decarboxylative asymmetric allylic alkylation reaction, as well as develop entirely novel reaction paradigms
From Tectonic Evolution to Intraplate Stress: The Role of Structural Inheritance and Long-Wavelength Loading
Get PDFIn this thesis, I present a multifaceted exploration of various aspects of deformation and stress in the Earth's lithosphere using a variety of methods in a range of tectonic environments. I begin by examining the evolution of a young subduction zone through a combination of gravity modeling and seismological observations. Chapter 2 details the development a linear 3-D gravity inversion method capable of modelling complex geological regions such as subduction margins. Our procedure inverts satellite gravity to determine the best-fitting differential densities of spatially discretized subsurface prisms in a least-squares sense. We use a Bayesian approach to incorporate both data error and prior constraints based on seismic reflection and refraction data. Based on these data, Gaussian priors are applied to the appropriate model parameters as absolute equality constraints. To stabilize the inversion and provide relative equality constraints on the parameters, we utilize a combination of first and second order Tikhonov regularization, which enforces smoothness in the horizontal direction between seismically constrained regions, while allowing for sharper contacts in the vertical. We apply this method to the nascent Puysegur Trench, south of New Zealand, where oceanic lithosphere of the Australian Plate has under-thrust Puysegur Ridge and Solander Basin on the Pacific Plate since the Miocene. These models provide insight into the density contrasts, Moho depth, and crustal thickness in the region. The final model has a mean standard deviation on the model parameters of about 17 kg/m-3, and a mean absolute error on the predicted gravity of about 3.9 mGal, demonstrating the success of this method for even complex density distributions like those present at subduction zones. The posterior density distribution versus seismic velocity is diagnostic of compositional and structural changes and shows a thin sliver of oceanic crust emplaced between the nascent thrust and the strike slip Puysegur Fault. However, the northern end of the Puysegur Ridge, at the Snares Zone, is predominantly buoyant continental crust, despite its subsidence with respect to the rest of the ridge. These features highlight the mechanical changes unfolding during subduction initiation. Chapter 3 explores the earthquake interevent time distribution. Earthquakes are commonly assumed to result from a stationary Poisson (SIP) process. We reassess the validity of this assumption using the Quake Template Matching (QTM) catalog and the relocated SCSN catalog (HYS) for Southern California. We analyze the interevent time (IET) distribution and the Schuster spectra after declustering with the Zaliapin and Ben Zion (2013) method. Both catalogs exhibit fat-tails on the IET distribution, deviating from the expected exponential distribution. The Schuster spectra of the catalogs are also inconsistent with an SIP process. The QTM catalog shows a statistically significant seasonal signal and a drift in the Schuster probability at long periods, likely due to increased seismicity following the 2010 El Mayor-Cucapah earthquake. This increase is also evident in the yearly IET distributions of the catalog. In contrast, the HYS Schuster spectrum does not show seasonality, but the yearly IET distributions exhibit a decrease in seismicity rate over the duration of the catalog, likely due to seismic network upgrades around 1990. We use synthetic catalogs to test the origin and significance of the observed deviations from the Poisson model. Variations in the QTM annual seismicity rate, around 5.6%, are too small to generate a noticeable departure from an exponential distribution, and the SIP model can not be rejected at the 5% significance level. The synthetic catalogs also suggest the fat-tail is an artefact of incomplete declustering. Overall, variations in the IET distribution for southern California are probably the result of both 1) incomplete declustering and location uncertainty, and 2) transient non-stationarity of the background rate from viscoelastic effects of large earthquakes. However, the stationary Poisson model appears adequate for describing background seismicity at the scale of Southern California and the decadal time scale of the QTM catalog. Chapters 4 and 5 cover the primary focus of this thesis, exploring the influence of long-wavelength loading on the stress field of continental interiors and intraplate seismicity. The continental interior of eastern North America in particular has hosted many significant historical earthquakes and is undergoing both glacial isostatic adjustment (GIA) and long-wavelength subsidence due to the sinking of the Farallon slab. The regional seismicity concentrates within ancient failed rift arms and other paleo-tectonic structures, which can act as weak zones in the crust where stress accumulates. Within some of these zones, focal mechanism stress inversion shows significant rotational deviation of the maximum horizontal stress (SHmax) direction from the regional NE-SW trend, which may be explained by long-wavelength stress perturbations in the presence of lithospheric weakness. We focus on two sources of intraplate stress perturbation and seismicity and test the hypotheses that 1) mantle-flow induced epeirogenic subsidence and 2) GIA contribute to intraplate seismicity in eastern North America via reactivation of pre-existing faults. For the slab loading component of this work, we use high-resolution global, spherical finite-element flow models with CitcomS. To capture realistic temperature fields and the Farallon slab, we convert seismic tomography models to temperature using a mineralogically constrained depth-dependent scaling factor. We utilize laterally variable temperature-dependent viscosities, upon which we superimpose low-viscosity plate boundary weak zones, as well as lithospheric intraplate weak zones at the locations of failed rifts and other inherited structures in eastern North America. We parameterize the Farallon slab in terms of its buoyancy to determine the degree to which the flow induced by the sinking slab contributes to intraplate stress. Using the modeled stress tensors from instantaneous flow calculations, we compute SHmax, the stress magnitudes, and the Coulomb failure stress on mapped faults in several major seismic zones. Slab sinking drives localized mantle flow beneath the central-eastern U.S., leading to a stress amplification of 100-150 MPa across the region that peaks over the New Madrid Seismic Zone. This stress amplification introduces a pronounced continent-wide clockwise rotation of the predicted SHmax direction, reaching as much as 20° in some seismic zones, particularly when lithospheric weak zones are included. In the New Madrid, Central Virginia, Charlevoix, and Lower Saint Lawrence Seismic Zones, the presence of weak zones loaded by the Farallon slab at depth can explain the pattern of clockwise rotation of the observed focal mechanism derived SHmax relative to the regional borehole derived SHmax as reported in previous studies. However, misfits on SHmax within many of the major seismic zones suggest other sources of stress are needed to properly reproduce the observed stress trends in some areas. We also find that in order for pre-existing lithospheric weak zones to exert appreciable control on intraplate stress under the influence of mantle flow, they must be shallow/sub-crustal and in contact with the crust. These stress perturbations and rotations ultimately bring faults in the NMSZ, the Western Quebec Seismic Zone (WQSZ), and the Lower Saint Lawrence and Charlevoix Seismic Zones closer to failure. In particular, inclusion of the Farallon slab and weak zones produces positive Coulomb failure stresses on some key faults associated with major historical earthquakes, including the Reelfoot Fault in the NMSZ and the Timiskaming fault in the WQSZ. Fault instability is even more likely when assuming weaker faults with lower coefficients of friction. For the glacial unloading component of this work, we use the global, spherical finite element code CitcomSVE, which models dynamic deformation of a viscoelastic and incompressible planetary mantle in response to surface loading. We supply CitcomSVE with the same seismically constrained viscosity structures computed in the CitcomS models, including those with weak zones, and load the Earth model with the ICE-6G ice history. We perform the same suite of simulations and stress analyses as in the mantle loading problem, using the stress tensor output of the corresponding CitcomS model as the tectonic background stress. We compare the mantle flow and GIA induced stresses, with focus on the present day extant glacially derived stress field. GIA induced stress perturbations are small (~10 MPa), even in the presence of lithospheric weak zones. GIA induced SHmax alone exhibits a transition from clockwise to counterclockwise rotation moving northeast across the continent. We find that only by inclusion of the mantle flow derived background stress can we reproduce the continental scale clockwise stress rotation observed in stress data, suggesting the effect of mantle loading is more important for explaining these observations than is GIA. In the NMSZ, GIA helps promote stability on the Reelfoot Fault, in opposition to mantle flow, while promoting instability on more non-optimally oriented faults. GIA also helps localize higher Coulomb failure stress within the Charlevoix Seismic Zone and the western half of the WQSZ. In the WQSZ and LSLRS, GIA stress perturbations are large enough that even with only a small reduction in the coefficient of friction, faults that are not likely to fail under the background tectonic and geodynamic stresses alone could slip. Further investigation of the sensitivity of GIA stress to different 3D and 1D viscosity structures and the change in GIA stress with time since deglaciation is warranted to better understand how GIA affects intraplate seismicity. Ultimately, constraining how mantle flow and GIA affect stress and deformation in the presence of laterally variable viscosity is integral to quantifying how long-wavelength loading may alter the spatial distribution of seismic hazard.</p
Essays in Behavioral Economics and Game Theory
Get PDFThis thesis consists of three papers. Chapter 1 conducts experimental research on individual bounded rationality in games, Chapter 2 introduces a novel equilibrium solution concept in behavioral game theory, and Chapter 3 investigates confirmation bias within the framework of game theory.
In Chapter 1 (joint with Wei James Chen and Po-Hsuan Lin), we investigate individual strategic reasoning depths by matching human subjects with fully rational computer players in a lab, allowing for the isolation of limited reasoning ability from beliefs about opponent players and social preferences. Our findings reveal that when matched with robots, subjects demonstrate higher stability in their strategic thinking depths across games, in contrast to when matched with humans.
In Chapter 2 (joint with Po-Hsuan Lin and Thomas R. Palfrey), we investigate how players’ misunderstanding about the relationship between opponents’ private information and strategies influence their equilibrium behavior in dynamic environments. This theoretical study introduces a framework that extends the analysis of cursed equilibrium from the strategic form to multi-stage games and applies it to various applications in economics and political science.
In Chapter 3, I employ a game-theoretic framework to model how decision makers strategically interpret signals, particularly when they face a utility loss from holding beliefs that differ from their partners. The study reveals that the emergence of confirmation bias is positively associated with the strength of prior beliefs about a state, while the impact of signal accuracy remains ambiguous.</p
Wearable Sweat Sensors for Disease Monitoring and Management
Get PDFWith the emphasis of healthcare shifting towards prevention and early detection of diseases and monitoring of chronic conditions, there is a growing need for hassle‐free telemedicine sensor technologies that can be seamlessly integrated into daily life. While significant progress has been made in the development of wearable sweat and salivary biosensors to meet this need for rapid, real-time collection of physiological information, the majority of current epidermal sensing systems are unable to detect trace-level disease-relevant biomarkers accurately in biofluids and cannot be mass produced. To meet this demand for low-cost, mass-producible mHealth devices for at-home settings, we developed several fully integrated laser-engraved graphene-based biosensors for the detection of low-concentration sweat and saliva analytes including hormones (cortisol) and proteins (C-reactive protein). Several graphene surface engineering strategies are investigated for the sensitive and selective detection of targets. System-level engineering and microfluidic designs are explored to achieve on-demand sweat induction and harvesting under sedentary settings and automated sweat and reagent routing and in situ signal correction and analysis for facile operation on the skin. The utility of these fully integrated flexible mHealth systems is evaluated through multiple human studies involving healthy and various patient subgroups towards stress assessment, as well as the monitoring and management of various chronic conditions including chronic obstructive pulmonary disease, heart failure, and inflammatory bowel diseases. These fully integrated mHealth devices demonstrate a technology that can be easily adapted to monitor a broad spectrum of disease-specific proteins, cytokines, and hormones, thus advancing future applications in personalized disease diagnosis, management, and prevention
Thermomechanical Properties of Nematic Liquid Crystal Elastomers
Get PDFLiquid crystal elastomers (LCEs) are materials formed by cross-linking crystal mesogens into a flexible polymer network, and they display soft behavior and undergo large, reversible strains. The mesogenic order determines material properties, causing coupling between temperature, liquid crystalline order, and deformation, which leads to temperature-based actuation. LCEs have important applications in soft robotics and medical devices, so attempts have been made to theoretically model their behavior in order to develop new use cases. One such model, developed by Lee (2021), identifies regions of liquid crystal orientation and has agreed with initial experimental data (Lee et al., 2023). This thesis aims to characterize the behavior of isotropic-genesis polydomain LCEs across various temperatures, strain rates, and crosslinking densities and further test the model by comparing the experimental data against it.
Tensile tests were run across five strain rates (10⁻¹/s, 5×10⁻²/s, 10⁻²/s, 5×10⁻³/s, 10⁻³/s), three temperatures (26◦C, 55◦C, 90◦C), and two crosslinking densities (50 mol%, 25 mol%). A custom tensile rig with a heated chamber made by Lee (2021) was modified for the purpose of this thesis to allow for digital image correlation and trials across temperatures.
These tensile tests revealed that stiffness increased with faster strain rates, and, as temperature increased, soft behavior was reduced at 55◦C and vanished at above the nematic transition temperature. Additionally, residual strain decreased with increasing temperature, at ∼1.5 at 26◦C, ∼0.75 at 55◦C, and ∼0.1 at 90◦C. Reducing the crosslinking density more than doubled the strain at failure and drastically
increased the region of soft behavior.
Experimental data across three strain rates (10⁻²/s, 5×10⁻³/s, 10⁻³/s), three temperatures, and at 50 mol% crosslinking density were compared against the model developed by Lee (2021). The soft behavior of the LCE was generally well characterized by the model, however, the model deviated from experimental data above two strain, as the Neo Hookean-based model was unable to capture strain hardening. Since higher temperature trials were run to lower strains, the model was able to better capture the full behavior of the LCE at higher temperatures, even with the loss of soft behavior at 90◦C. This model is therefore a useful tool for modeling LCE soft behavior across various temperatures.</p
Design and Implementation of a Microparticle Delivery Device for the Cornea
Get PDFBiolistic drug delivery offers an alternative path for delivering therapeutics into the cornea. Until now, none of the commercially available gene guns are suitable for clinical delivery of therapeutics due to tissue damage caused by high speed gas used to accelerate microparticles. Here, we demonstrated the use of a device that both eliminates the exit gas, only allowing high speed particles through, and one that works in a clinical setting.
Microparticles ranging from 5 to 22 μm were accelerated and delivered into both the agarose gels and ex vivo corneas. In gels, we found that normalized penetration depth was proportional to particle diameter and density. As the standoff distance between the device and the target increased, more particles were left stranded at the surface, as their penetrating power decreased, and their dispersion from the center of mass on the target increased. The orifice size served to control both the number of particles and the amount of exit gas. Increasing the inlet pressure did not a show significant increase in the penetration depth of microparticles.
In the cornea, we found that we were able to use our device to deliver particles into both the epithelium and the stroma, although only higher density particles were able to enter the stroma. There was little to no damage to the cornea due to particle delivery. If epithelial defects were detected in the cornea due to particle penetration, they were quickly resolved within 30 minutes. Our device demonstrated performance (penetration depth) comparable to previous biolistic delivery methods in the cornea, while also maintaining clinical relevance by eliminating exit gas flow. </p
Energy-Efficient and Robust Algorithms for Biomedical Applications
Get PDFMedical devices play a critical role in improving the quality of life for patients and assisting physicians by monitoring, detecting, and helping manage chronic conditions such as epilepsy and spinal cord injuries. To perform these functions effectively, these devices must extract the most relevant information from complex medical data. However, the functionality of these medical devices has been limited by the existing challenges in medical applications. Some of these challenges include the complexity in the analysis of raw medical data, adaptability, non-stationarity, noise, large data volumes, real-time processing, limited resources, and high accuracy demands. Moreover, considering factors such as individual differences, environmental influences, and genetic variations, medical data will cause numerous variations and uncertainties in analyzing and interpreting the medical conditions in different biomedical applications. Medical data analysis is already complex and is further complicated by issues like non-stationarity and noise, especially when using traditional and manual methods. When it comes to the designing, implementation, and utilization of wearable and implantable medical devices, efficiency, accuracy, and adaptability become crucial. Particularly, applications that require fast control of equipment, such as brain-machine interfaces (BMIs), make the need for fast decision-making evident. Medical data have been conventionally managed by reliance on extensive manual labor. However, such manual data management techniques are not scalable, have inefficient procedures, and are more likely to produce errors. Therefore, more advanced, automated methods are required immediately considering the existing challenges of the current medical data analysis techniques.
Such a shift in data processing and management will lead to more trustable procedures that can significantly improve the accuracy and efficiency of medical data analysis. Other than being just an improvement, such transformation signifies a noteworthy point in the development of medical devices. In this view, it is essential to introduce advanced technology and novel methods for medical data processing as well as automation. Therefore, it becomes critical that these high-performance and advanced techniques can efficiently be implemented with minimum effects on hardware for clinical applications. Currently, artificial intelligence (AI) and its subfield machine learning (ML) has led to major transformations in designing and utilization of various medical devices. Among all these biomedical applications, three major area are addressed in this thesis: Brain Machine Interfaces (BMIs), seizure detection, and classification of arrhythmias in cardiac rhythms. We selected these three applications due to their significance and ability to improve patient treatment further. Additionally, we showed how we used machine learning algorithms for each of these applications to address their current challenges.
In our work related to Brain-Machine Interfaces (BMIs), we have been focused on improving the quality of life for individuals with spinal cord injury (SCI) through two studies. In our initial study, we have designed and implemented a deep multi-state Dynamic Recurrent Neural Network (DRNN) decoder for BMI applications. This algorithm decodes neural data recorded from the posterior parietal cortex (PPC) and the motor cortex (M1) of human participants to appropriate control signals to predict computer cursor kinematics on the computer screen. By reducing the amount of history used in predicting the movement kinematics from the recorded neural data, we have demonstrated that improved performance and robustness are preserved while memory and power consumption are reduced. We then compared the performance of DRNN with other decoding techniques to demonstrate that when operating on wavelet-based neural features, our proposed DRNN-based decoder outperforms other decoding techniques. Therefore, DRNN have the potential to be used for more efficient and effective BMIs. After developing DRNN as a decoding technique for BMI applications, we have implemented an efficient feature extraction technique, referred to as Feature Extraction Network (FENet), which has been designed by using convolutional neural networks for optimizing feature extraction and decoding to ensure consistency across electrodes when decoding the recorded neural data to the movement kinematics in BMI systems. After being tested with data recorded from the posterior parietal and motor cortices of three human participants, FENet outperformed existing feature extraction techniques such as threshold crossings and wavelet transforms, and it significantly enhanced both closed- and open-loop cursor controls. We have also evaluated the generalizability of FENet when applied to different datasets, brain regions, and participants. Therefore, the results of our research in BMI technology have the potential to promise the improvement of the quality of life for spinal cord injury (SCI) patients.
Second, we co-designed EKGNet, a convolutional network that combines analog computing and deep learning for detecting heartbeat arrhythmia. EKGNet demonstrated high accuracy while minimizing power consumption, effectively overcoming challenges related to analog circuitry and real-time processing. The experimental findings, using PhysionNet’s MIT-BIH and PTB Diagnostics datasets, showed an average balanced accuracy of 95% for intra-patient arrhythmia classification and 94.25% for myocardial infarction (MI) classification.
Finally, we designed a real-time seizure detector by using XGboost as a technique relies on gradient boosted trees, which can help with the fast and accurate diagnosis of seizure for epileptic patients. With an averaged detection latency of 1.1 seconds, this design attained average F1 scores of 99.23% and 87.86% under various data splitting methods. The energy-area-latency product was 27× lower than the current state-of-the-art solutions, which allowed for adjustments that were specific to each patient and significantly reduced energy consumption.
The results presented in this dissertation demonstrate the potential of AI in addressing the existing challenges in three biomedical applications: brain-machine interfaces (BMI), seizure detection, and heartbeat arrhythmia detection. By addressing these existing challenges including complex biological data management, real-time processing constraints, and limited resources in biomedical applications, AI has the potential to improve the quality of life for patients suffering from neurological disorders and medical conditions. Moreover, the improved precision, operational efficiency, and flexibility caused by the integration of AI into the design of the future biomedical systems will potentially assist healthcare providers to offer enhanced support and treatment to patients. While we have focused on the three above-mentioned biomedical applications, the principles learned from our analysis may be relevant and can be extended to other biomedical applications.</p