Washington University Medical Center
Washington University St. Louis: Open ScholarshipNot a member yet
26344 research outputs found
Sort by
A Mechanistic Understanding of Transcutaneous Auricular Vagus Nerve Stimulation
No full textTranscutaneous auricular vagus nerve stimulation (taVNS) offers great potential as a noninvasive neuromodulation technique with expanding therapeutic potential across a range of neurological and systemic indications. Yet despite strong clinical interest, the mechanistic foundations of taVNS remain incompletely understood. In contrast to the invasive form of vagus nerve stimulation (VNS), taVNS has seen more limited parametric characterization, with many studies relying on parameters adopted from invasive protocols without systematic validation at the noninvasive level. This dissertation focuses on developing a more comprehensive understanding of the underlying mechanisms of taVNS. It extends existing literature by leveraging intracranial electrophysiology to directly quantify neural responses to both traditional electrical taVNS and a novel vibrotactile modality. Furthermore, it broadens the scope of mechanistic research to consider the central nervous system in its entirety by probing responses in spinal motoneuron excitability. Using stereotactic electroencephalography (sEEG), we first investigated local activity in deep brain regions across a range of stimulation intensities. We observed increased gamma power during 1.0 and 1.5 mA active taVNS in several limbic regions, predominantly concentrated in the left hemisphere. Additionally, we observed that this increased gamma power was sustained throughout the duration of stimulation in both the insula and orbitofrontal cortex (OFC). Notably, lower intensity stimulation (0.5 mA) resulted in a subtle decrease in gamma power in some regions. These contrasting, amplitude-specific responses provide rationale for the critical role parameter selection plays in the efficacy of neuromodulation and may partially explain the variability of results present in VNS and taVNS literature. To establish a more generalizable method of noninvasive vagal activation, we next designed and built a custom earpiece intended to stimulate the cymba concha mechanically via vibration. We again used sEEG recordings to investigate brain responses across five vibration frequencies, observing significant increases in global low-frequency coherence during 6, 20, and 40 Hz stimulation. These findings support the use of vibrotactile stimulation as a viable alternative modality for noninvasive VNS, motivating further investigation into behavioral, clinical, and sham-controlled mechanistic studies. Finally, to complement the two studies focusing on brain responses to taVNS, we also investigated the role of spinal circuits, and more specifically spinal motoneuron excitability, in mediating the effects of taVNS. Addressing a notable gap in the literature, we designed a protocol using evoked electromyography (EMG) to measure changes in motoneuron excitability in healthy individuals during concurrent taVNS. In this pilot study, we did not observe significant changes in motoneuron excitability, quantified via F-wave persistence and amplitude, across either active taVNS or sham stimulation. However, notable individual variability suggests that a larger sample size may reveal more subtle or context-dependent spinal effects. Collectively, this dissertation reveals the nuanced interactions between taVNS and the central nervous system, underscoring the need for continued mechanistic research. Careful optimization of stimulation parameters – including intensity and modality – will be essential for realizing the full therapeutic potential of taVNS
Learning From Conditional Data Distributions
No full textTraditional machine learning paradigms often rely on a single global model trained on an entire dataset, aiming for broad generalization across all instances. However, in many real-world applications, the underlying data distribution is heterogeneous, and meaningful predictions often require models that focus on specific subpopulations rather than treating the data as a whole. This motivates the study of learning from conditional distributions, a framework where predictive models are designed to capture the structure and properties of restricted subsets of the data, leading to improved accuracy, fairness, and interpretability. This dissertation explores three key subproblems that exemplify different aspects of learning from conditional distributions: fairness auditing, conditional linear classification, and personalized prediction. These problems share a common goal of modeling and verifying properties of subpopulations within a broader distribution but present distinct computational challenges. First, we examine fairness auditing, which assesses whether a classifier exhibits statistical parity across certain subgroups. This can be seen as verifying whether the performance of a model remains stable when conditioned on specific subsets of the data. We develop an efficient auditing framework using a reduction to agnostic learning under distributional assumptions. More importantly, we establish distribution-specific lower bound, for auditing halfspace subgroups by reducing from the Learning With Errors problem, demonstrating that verifying fairness over the family of general halfspace is computationally intractable even under nice distributions, such as standard normal ones. Next, we investigate conditional linear classification, where the objective is to learn a classifier that minimizes classification error within a well-defined subgroup. Unlike fairness auditing, which focuses on post-hoc verification of properties, conditional classification actively seeks an optimal classifier-subgroup pair. We present efficient approximation algorithms for this problem under well-behaved and standard normal distributional settings. However, we also discovered nontrivial distribution-specific lower bounds for conditional linear classification by reduction from agnostic linear classification, which turns out to be computationally intractable even under Gaussian distributions. Building on conditional classification, we extend our study to personalized prediction, which seeks to optimize predictions for an individual rather than a predefined subgroup. The key distinction is that in personalized prediction, the subgroup or reference class must always contain the reference point we aim to classify. This introduces an additional constraint but also highlights a natural extension of conditional learning. We show that with modifications to our learning algorithms, we can efficiently construct reference classes that satisfy this requirement, bridging the gap between conditional classification and truly individualized predictions. Together, our analysis of the three problems establish a rich theoretical foundation for learning from conditional distributions, which should give the audience a good overview of this domain. In particular, our findings provide both algorithmic insights and computational lower bounds, illustrating the trade-offs between computational hardness and the expressive power of the subgroup classes. By leveraging techniques from gradient descent and cryptographic hardness assumptions, this dissertation deepens our understanding of conditional learning and its potential for practical applications in fairness auditing, interpretable classification, and personalized decision-making
Optimizing Noninvasive Uterine Electrophysiological Imaging: Advancements in Electrode Reduction, Signal Enhancement, and Robust Data Processing
No full textIn recent years, noninvasive electrophysiological imaging systems have significantly advanced the visualization and analysis of electrical activity in human organs without the need for invasive procedures. Among these innovative systems, Electromyometrial Imaging (EMMI) and Uterine Peristalsis Imaging (UPI) have emerged as powerful tools for studying uterine electrical behavior — EMMI during pregnancy and UPI throughout the menstrual cycle. By enabling the reconstruction of detailed four-dimensional electrical activation patterns, these technologies offer unprecedented insights into uterine physiology and function. However, current procedures for noninvasive electrophysiological imaging systems face several challenges, including the need for a large number of electrodes, low-quality electrical measurements due to noise and motion artifacts, and a lack of robust algorithms for signal processing. This dissertation aims to address these issues by optimizing the critical components of the workflow: electrode reduction, signal enhancement, and data processing. First, we propose a method to reduce the number of electrodes while maintaining imaging quality, thereby improving clinical feasibility. Second, we enhance signal quality by leveraging the relationship between slow and fast wave components in the uterine electrical signals, improving the signal-to-noise ratio to enable better detection of uterine contractions. Finally, we introduce improved post-processing techniques tailored to the complex, multi-channel, and geometry-dependent nature of EMMI data, allowing for more accurate and robust identification of uterine contractions
An Introduction to Geospatial Thinking and Open Source GIS
No full textAn Introduction to Geospatial Thinking and Open Source GIS is a beginners guide to understanding geospatial concepts and tools. As we move into an age of AI generated research, this foundational learning is essential for understanding the nuts and bolts of conducting analyses as well as evaluating the analyses of others. Part one is focused largely on concepts and part two is focused on practice using open source tools. This is an Open Educational Resource (OER). Part one was adapted mainly from Essentials of Geographic Information Systems, Campbell, 2011, and Spatial Thinking in Planning Practice: An Introduction to GIS, Fang, 2014. Part two is newly developed content focused on providing readers, scaffolded introductory exercises.
Available online through Pressbooks.https://openscholarship.wustl.edu/books/1070/thumbnail.jp
Housing Options as We Age
No full textAlmost 75% of adults ages 50 and older want to stay in their homes and communities as they age. This is referred to as “aging in place.” Our homes and communities can help us feel safe and give us a sense of belonging and connection. Furthermore, our living situations effect our health and quality of life as well as our ability to live independently. Changes in our physical, psychological, social, and financial well-being later in life can affect our abilities to age in place. Finding the best housing based on our resources, interests, and needs is important for living well as we get older
Viability as Abortion-Rights Orthodoxy
No full textBorrowing a quotation from Justice Jackson’s influential opinion in West Virginia Board of Education v. Barnette, Professors Linda McClain and James Fleming have named their ambitious and illuminating new book “What Shall Be Orthodox” in Polarized Times. As Justice Jackson wrote in applying the First Amendment to protect school children with familial objections to a required flag salute: “If there is any fixed star in our constitutional constellation, it is that no official, high or petty, can prescribe what shall be orthodox in politics, nationalism, religion, or other matters of opinion or force citizens to confess by word or act their faith therein.” Although Justice Jackson takes aim at government prescribed orthodoxy and Professors McClain and Fleming understandably center such concerns, this essay for the symposium on their book takes a broader look at orthodoxy. As I show, orthodoxies can form in subtle ways without government prescription, and official speech—although not itself subject to the First Amendment—can sometimes shape them.
This essay explores, as an illustration, the inclusion of a viability limit in the successful November 2024 ballot initiative that placed a right to reproductive freedom in the Missouri Constitution. The orthodoxy lens prompted by McClain and Fleming offers new insights on how viability, a judicial compromise from the now overruled 1973 decision in Roe v. Wade, came to constrain efforts to reimagine reproductive justice after Dobbs v. Jackson Women’s Health Organization. The campaign to restore abortion in Missouri makes plain the tenacity of the viability limit, notwithstanding both strong arguments by those who opposed any gestational limit and at best tepid support in favor of one based on fetal viability. This “case study” illustrates how orthodoxies can arise from state actors even in the absence of any government-imposed restriction or compulsion
Expansion Limits of Meshed Split-Thickness Skin Grafts
No full textSplit-thickness skin grafts are widely used to treat chronic wounds. Procedure design requires surgeons to predict how much a patch of the patient\u27s own skin expands when it is meshed with rows of slits and stretched over a larger wound area. Accurate prediction of graft expansion remains a challenge, with current models overestimating the actual expansion, leading to suboptimal outcomes. Inspired by the principles of mechanical metamaterials, we developed a model that distinguishes between the kinematic rearrangement of structural elements and their stretching, providing a more accurate prediction of skin graft expansion. Our model was validated against extensive data from skin graft surgeries, demonstrating vastly superior predictive capability compared to existing methods. This metamaterial-inspired approach enables informed decision-making for potentially improving healing outcomes
Judicial Reform from the Inside Out
No full textPrepared for the Notre Dame Law Review’s Spring 2025 Federal Courts Symposium on the 100th Anniversary of the Judiciary Act of 1925.The Judiciary Act of 1925, the subject of this Symposium, is known as “The Judges Bill” for a reason. The Justices of the Supreme Court, and Chief Justice Taft in particular, produced the Act and persuaded Congress to enact it. To modern eyes, such efforts seem indecorous; perhaps even scandalous. But in fact, Supreme Court Justices and other federal judges have been extensively involved in judicial reform throughout American history. This Essay examines participation by federal judges in judicial reform efforts—what we call judicial reform from the inside out.We survey examples of judges participating in reform debates from across different historical eras and different levels of the federal judiciary. We then use our descriptive account as a platform for theoretical and normative analysis. We begin by drawing some general lessons from the historical narrative. We then identify the overarching costs and benefits of judicial participation in reform, as well as the many factors for which one must account in normatively assessing any one instance of inside-out judicial reform. Relying on that framework, we offer some tentative recommendations for how inside-out judicial reform can be appropriately channeled. Studying judicial reform from the inside out can help us understand court administration, the judicial role, and the relationship between the judiciary and the political branches—as well as shedding light on the current debate over reform of the Supreme Court
Sex Differences in the Risk of Developing Alzheimer Disease, Cognitive Trajectories, and Relationships between Cognition and Pathological Burden
No full textPrior work suggests females may develop Alzheimer disease (AD) dementia at a higher rate than males, and that this effect may interact with apolipoprotein (APOE) ε4 genotype. However, this work has often been limited by imprecise clinical diagnoses and a lack of AD biomarkers. In addition to the frequency of dementia, there are mixed reports on whether sex impacts rates of cognitive decline. Further, females consistently exhibit greater accumulation of AD pathological biomarkers, particularly tau, than males. It has been theorized that females initially demonstrate resilience to AD pathology, maintaining cognitive functioning longer than males, but experiencing a steeper cognitive decline as the disease progresses. To understand the role sex plays in AD we conducted two studies. The first Study (Study 1) investigated the influence of sex and APOE ε4 status on risk of AD, as well as the effects of sex on cross-sectional and longitudinal cognition. The second Study (Study 2) explored sex differences in the relationships between AD pathology levels and cognition cross-sectionally and longitudinally. Participants were selected from the Charles F. and Joanne Knight Alzheimer Disease Research Center. In Study 1, we characterized cognition in two ways. Dementia severity was assessed using Clinical Dementia Rating (CDR) scores and cognitive performance was assessed using neuropsychological tests combined into a Preclinical Alzheimer Cognitive Composite (PACC). Based upon the epidemiological literature we were interested in whether sex impacted the incidence rate of developing dementia. Dementia was defined using three approaches to mirror different standards in the field. First, all-cause dementia, or any sign of decline, provided insight into sex differences in dementia risk within general populations. Second, restricting analyses to clinically diagnosed AD accounted for clinician-based identification of the disease. Finally, defining AD using biomarker-confirmed amyloid-beta (Aβ) status, alongside cognitive impairment, aligned with the biological framework of AD. For risk analyses, only participants who were cognitively normal at baseline (CDR = 0) were included. Cox proportional Hazards were used to estimate risk of developing dementia. As expected, overtime, APOE ε4 carriers had a significantly higher risk of all cause dementia compared to non-carriers with an 84.5% increased risk (HR = 1.85 [1.50–2.27], p \u3c 0.0001), a 97% increased risk of clinical AD dementia (HR = 1.97 [1.55–2.51], p \u3c 0.0001), and a 276.3% increased risk of dementia in individuals with biomarker-confirmed Aβ, (HR = 3.76 [2.38–5.94], p \u3c 0.0001). There were no significant differences between males and females in the likelihood of developing all-cause dementia (HR = 0.93 [0.76-1.14], p = 0.5), clinical AD dementia (HR = 0.95 [0.75-1.2], p = 0.6), or dementia in Aβ-positive participants (HR = 1.1 [0.70-1.7], p = 0.7). Additionally, APOE ε4 status did not differ by sex in predicting dementia risk (HR = 0.96 [0.6-1.4] p = 0.8, HR = 0.99 [0.6-1.6] p = 1.0, HR = 0.85 [0.3-2.1] p = 0.7, respectively). As a compliment to analyzing overt dementia in Study 1, we analyzed continuous cognitive performance using a neuropsychological composite. To assess sex differences, we stratified the cohort into cognitively unimpaired individuals and a combined group of unimpaired and impaired individuals. We then further restricted our analysis to Aβ-positive participants. This allowed us to examine sex differences in normative cognition and determine whether cognitively impaired females experience faster decline than impaired males. By focusing on Aβ-positive individuals, we assessed sex differences and cognitive decline in individuals on an AD pathological trajectory. Cross-sectional analyses utilized linear regression models and longitudinal analyses employed linear mixed effects models. Females demonstrated superior cognitive performance at baseline in the unimpaired subgroup (β = 0.12, t = 4.7, p \u3c 0.0001), Aβ-positive unimpaired subgroup (β = 0.15, t = 4.5, p \u3c 0.0001), unimpaired and impaired subgroup (β = 0.14, t = 4.5, p \u3c 0.0001) and Aβ-positive unimpaired and impaired subgroup (β = 0.2, t = 4.7, p \u3c 0.0001). There was no evidence to suggest sex-based differences in cognitive decline over time in the unimpaired subgroup (β = -0.002, t = -0.6, p = 0.5) or the Aβ-positive unimpaired subgroup (β = -0.001, t = -0.3, p = 0.8). Additionally, there were no significant differences in PACC slopes between males and females in the unimpaired subgroup (t = -1.03, p = 0.3) or in the Aβ-positive unimpaired subgroup (t = 0.3, p = 0.7). In both the unimpaired and impaired subgroup (β = -0.003, t = -0.6, p = 0.6) and the Aβ-positive unimpaired and impaired subgroup (β = 0.003, t = 0.3, p = 0.8), impaired females did not exhibit a faster rate of cognitive decline than impaired males. Further, the effect of sex on PACC slopes did not significantly differ across CDR groups in both the unimpaired and impaired subgroup (F=1.6, p=0.2) and the Aβ-positive unimpaired and impaired subgroup (F=1.6, p=0.2). In Study 2, we examined whether there was any evidence that sex infers an inherent resistance or vulnerability to AD pathology. To test this question, we tested if sex alters the relationship between levels of baseline AD pathology and both cross-sectional and longitudinal cognition. First, we studied the relationship between tau and cognition using the PACC neuropsychological composite. We then studied the relationship between Aβ pathology and cognition. We also further restricted our analysis to Aβ-positive participants to examine effects in individuals on an AD pathological trajectory. Cross-sectional analyses utilized linear regression models and longitudinal analyses employed linear mixed effects models. Greater baseline tau was associated with worse PACC scores at baseline in the full group (β = -1.1, t = - 6.2, p \u3c 0.001) and in the Aβ-positive subgroup (β = -1.3, t = -5.3, p \u3c 0.001). Longitudinally, greater baseline tau was associated with steeper rate of decline on PACC scores in the full group (β = -0.2, t = -3.2, p = 0.001) and in the Aβ-positive subgroup (β = -0.3, t = -2.7, p = 0.007). The adverse effect of baseline tau on baseline cognition was more pronounced in females compared to males in the full group (β = -0.48, t = -2.0, p = 0.04) and Aβ-positive subgroup (β = -0.84, t = -2.5, p = 0.01). Compared to males, females demonstrated slower longitudinal slopes relative to baseline tau, indicative of greater cognitive resilience to similar levels of baseline tau pathology over time in both the full cohort (β = 0.3, t = 3.3, p = 0.0009) and the Aβ-positive subgroup (β = 0.5, t = 3.5, p = 0.0005). This pattern aligns with the hypothesis that females may maintain better cognitive functioning than males despite the presence of AD pathology. Greater baseline Aβ was also associated with worse baseline PACC scores in both the full group (β = -0.006, t = -6.9, p \u3c 0.001) and the Aβ-positive subgroup (β = -0.0081, t = -5.0, p \u3c 0.0001). Longitudinally, greater baseline Aβ was associated with a steeper decline in PACC scores, but this relationship was not statistically significant in either the full cohort (β = -0.0004, t = -1.9, p = 0.06) or the Aβ-positive subgroup (β = -0.0004, t = -0.9, p = 0.4). No sex differences were observed in baseline PACC performance in relation to baseline Aβ pathology in either the full group (β = 0.002, t = 1.5, p = 0.1) or the Aβ-positive subgroup (β = 0.004, t = 1.7, p = 0.09). Males and females showed similar longitudinal slopes relative to baseline Aβ in both the full cohort (β = 0.0002, t = 0.7, p =0.5) and the Aβ-positive subgroup (β = 0.0005, t = 0.8, p = 0.4), indicating no significant sex differences in the effect of baseline Aβ burden on PACC decline over time
Cumulative Lifespan Stress and Inflammation are Associated with Black-White Racial Disparities in Mortality Among Americans
No full textBlack Americans disproportionately experience higher rates of health challenges and mortality compared to White Americans, yet the mechanisms underlying these disparities remain inadequately understood. Prominent theoretical models highlight stress and resulting allostatic load as putative mechanisms through which these Black-White racial disparities emerge; however, empirical data supporting such models with regard to mortality remains sparse. The current study examined the potential role of cumulative stress exposure across the life span and elevated levels of C-Reactive Protein, a biomarker of inflammation, in contributing to the longstanding increased mortality risk among Black relative to White Americans. Data were drawn from the Saint Louis Personality and Aging Network (SPAN) study, a longitudinal study that has followed a cohort of older adults from the St. Louis community (n = 1,577; 32.7% Black, 67.3% White; 55.1% female; mean age at baseline = 58.08 ± 2.95years). Cox hazard proportional mediational models showed that Black participants had significantly higher mortality risk than White participants (HR = 1.77, 95% CI [1.36, 2.31], p \u3c 0.001), along with greater cumulative stress (b = 0.558, [0.487, 0.628], p \u3c 0.001) and CRP levels (b = 0.184, 95% CI [0.132, 0.236], p \u3c .001). Additionally, the serial model revealed that indirect effects through cumulative stress, CRP, and their serial pathway significantly mediated the relationship between race and mortality risk, collectively accounting for 35.85% of the disparity. These findings suggest that heightened stress exposure and inflammation are plausible mechanisms contributing to existing Black-White disparities in mortality risk, underscoring the importance of structural interventions aimed at reducing long-term stressors disproportionately experienced by Black Americans