22689 research outputs found
Sort by
BRIDGING THE GAP IN ROBOT-ASSISTED SURGICAL TRAINING: A SYSTEMS APPROACH
Get PDFThe central goal of any surgical residency training program is to prepare physicians and surgeons for independent practice within an outcome-based, peer-reviewed system. Despite significant reforms by the American Council of Graduate Medical Education (ACGME) to establish competency-based education (CBE) models, there is severe variability in the quality of robotic-assisted surgery (RAS) training for surgical residents, leading to disparities in their preparedness for independent practice. This dissertation examines urology residents' perceived training and assessment variability, focusing on their confidence and competence in RAS procedures.
Using Bronfenbrenner’s Ecological Systems Theory (EST) as the theoretical framework, this study analyzes the multi-layered influences on residents' training experiences within an urban academic medical center. Data were collected from 16 faculty and 16 residents through surveys, revealing three pivotal themes for refining RAS training: enhancing structured training programs, increasing simulation-based training, and integrating early exposure to RAS techniques.
Key findings indicate that residents report diverse experiences and varying levels of exposure to RAS procedures, with significant gaps in mentorship and feedback. Both residents and faculty recognize the importance of tailored training programs, but discrepancies exist in perceptions of intraoperative danger, time efficiency, and skill acquisition rates. Identified barriers include limited autonomy for residents, competing institutional priorities, and the need for extra time for effective teaching.
The study underscores the need for standardized and comprehensive RAS training programs to ensure consistent, high-quality education. Addressing identified barriers and aligning training approaches with resident and faculty expectations can enhance residents' preparedness, competence, and confidence in robotic surgery, ultimately improving patient outcomes and advancing the field of urology.
This dissertation provides actionable recommendations for developing structured training pathways, incorporating diverse educational tools, and fostering a cohesive educational environment. This research aims to equip future surgeons with the necessary skills to excel in advanced minimally invasive procedures by bridging the gap in RAS training
NEUROMORPHIC SLAM SYSTEM
Get PDFMost animals, including humans, have the capability to navigate through complex environments. While traversing through unfamiliar environments, they can keep track of their location and form a memory that correlates various environmental information including visual landmarks and scent, home, food, or predators and obstacles with their location, thus encoding a map in their brain. Furthermore, upon their next trip, they can utilize the map to navigate to a desired location efficiently with consideration of distance and probable threats. Such capability is of interest to the robotics community and is known as Simultaneous Localization and Mapping, or SLAM [1-8]. Yet SLAM systems typically get inputs from many sensors and often use complex mathematical algorithms, requiring high computation capability and thus high energy consumption. However, animals can achieve high-performance SLAM without, to our understanding, the ability to perform any overt mathematical computation. Neuroscience researchers are attempting to explain the mechanism of spatial encoding neurons and how they contribute to path planning and navigation.
Neuromorphic engineering is a subject that brings neuroscience research findings into the application of engineering for more efficient and smart solutions as well as enhancing the understanding of neural systems from the perspective of an electrical designer. This thesis attempts to construct both a theoretical model of neural circuit structure according to neuroscience discoveries, and an electrical realization of the spatial encoding and navigation neurons for navigation purposes.
The dissertation proposed a neuromorphic model for how spatial encoding cells can be formed by lower-level movement-encoding neurons with variations in their behavior, and even utilized such variations for better spatial encoding uniqueness. It provides a way for digitizing continuous neural signals for simpler implementation.
This work also provided a hardware implementation demonstration for the effectiveness of the proposed neuromorphic model. A chip design imitating the behavior of movement-encoding neurons has been proven effective in the spatial encoding process of the neuromorphic SLAM system. A neural circuit model of path planning functionality is subsequently composed for a hypothesis on how the spatial encoding neurons can participate in navigation for animals
IN SILICO MODELING OF FLOW DIVERTER STENT HEMODYNAMICS FOR ENDOVASCULAR TREATMENT OF CEREBRAL ANEURYSMS
No full textIt is estimated that 1 to 5% of the adult population in the world suffers from cerebral aneurysms. These aneurysms, when left untreated, may burst resulting in subarachnoid hemorrhage which has a 50% mortality rate. The classical treatment for cerebral aneurysms is by a surgical procedure known as microvascular clipping. In this open brain surgical procedure, a clip is placed at the aneurysms neck to cut off the supply of blood, promoting thrombogenesis. The treatment of cerebral aneurysm has, however, improved significantly over the years. Cerebral aneurysms are now often treated using endovascular treatment through the insertion of a flow diverting stent, which reduces the blood flow into the aneurysms. This treatment induces stasis and intra-saccular thrombosis, diminishing the chances of aneurysm rupture.
Currently, device selection for endovascular treatment heavily relies on the expertise of clinicians. In this study, we perform an in-silico modeling to predict the effect of flow diversion device on the blood hemodynamics and occlusion of the cerebral aneurysm. The stent is modeled as an additional momentum source in incompressible Navier-Stokes equation. Two stent models are presented, porous-medium stent and screen force stent. The cerebral aneurysm model is derived from CT scan data provided by the Department of Neurosurgery, the Johns Hopkins Hospital. Additional scalars are added for thrombogenesis model, with the chemical transport modeled using advection-diffusion-reaction equation. The immersed boundary solver, ViCar3D, is utilized to solve the computational model which yields the flow pattern and occlusion inside the cerebral aneurysm.
The result shows a successful flow suppression for both porous-medium stent and screen force stent, with the latter being more practical. The velocity profile at the aneurysm neck shows the blood velocity is reduced up to 82.90% for porous medium stent and 20.97% for screen force stent. However, the thrombogenesis model does not show any growth of blood clot in the aneurysms. This study provides insights on the thrombotic occlusion of flow-diverter treatments of cerebral aneurysm and represents a capability that could be used by clinicians to evaluate endovascular treatments
THE IMPACT OF THE TUMOR MICROENVIRONMENT ON BREAST TUMORIGENESIS
Get PDFAlthough therapies for breast cancer have developed rapidly in the last decade, breast
cancer is still the main global cause of women’s death due to cancer. As with all the other cancer
cells, breast cancer is caused by the accumulation of mutations which causes unregulated cell
proliferation. The tumor microenvironment has long been recognized to support cancer cell
survival, growth, and metastasis. Accumulating evidence demonstrates that extracellular matrix
(ECM), a major structural component of tumor microenvironment, modulates oncogenic
signaling cascades through the interaction between key proteins and cell-surface receptors. We
utilized fibroblast-derived ECM scaffolds to investigate the impact of the ECM on breast
epithelial cells. The mRNA expression of CD24 and CRIP2 was largely decreased in MCF-10A cells
cultured on ECM scaffolds while the protein expression of CRIP2 was upregulated. In the
meantime, most downstream effectors of NFκB were inhibited when cells were cultured in an
ECM-rich environment. The O2 concentration is another important feature of the tumor
microenvironment. We discovered that the mRNA expression of FUT11, DNAH11, and TACF2
was increased in MDA-MB-231, MCF-7, and ZR-75-1 cell lines, whereas CASP14 expression was
only upregulated under hypoxic conditions in MCF-7 and ZR-75-1 cell lines. In addition, we
found that the upregulation of FUT11, DNAH11, and TCAF2 mRNA expression under hypoxic
conditions required HIF-1α; whereas in MCF-7 cell lines, only CASP14 and FUT11 required HIF-1
α expression
ENHANCING CLINICAL DECISION-MAKING IN LOW-RESOURCE SETTINGS: COMPARING MORTALITY RISK SCORES FOR ADULT CRITICAL CARE PATIENTS IN LESOTHO DURING THE COVID-19 PANDEMIC
Get PDFAbstract
Introduction:
Lesotho faced challenges amid the COVID-19 pandemic. Mortality or severe illness risk scores offer potential in aiding patient triage and resource allocation. Our study aims to evaluate the performance of these scores in Lesotho's COVID-19 context. By investigating factors distinguishing mortality from survival and assessing score effectiveness, we seek to address the gap in understanding their applicability in low-resource settings during the pandemic.
Methods:
Berea and Mafeteng hospitals were the main COVID-19 treatment centers during the pandemic in Lesotho. This retrospective cohort study focused on adult critical care admissions and predicting their survival outcomes using mortality risk scores. Logistic regression analyzed odds of death by clinical features and three mortality risk scores: Universal Vital Assessment (UVA)1, Modified Early Warning Score (MEWS)2, and Mortality Probability Admission Model (MPM)3. Predicted probabilities and ROC curves evaluated model performance, with optimal thresholds determined for classification.
Results:
From March 2020 to May 2022, 1,426 patients were received, with 449 deaths recorded. About 59% of all COVID tests were positive (n=844), 23% were suspected (n=330) and the remaining were negative (18%, n=252). UVA’s high-risk category had greater odds of death than the medium risk category (aOR: 2.81, 95% CI: 6.01, 13.87, aOR: 1.82 95% CI: 2.14, 3.97,
respectively). MEWS and mMPM24 scores showed increased mortality odds in high versus low categories (aOR: 1.78; 95% CI: 1.26, 2.57 and aOR: 2.31 95% CI: 1.3, 4.0, respectively). All three mortality risk scores had poor discrimination (mMPM24, C-statistic: 0.55; UVA, C-statistic: 0.65, MEWS: C-statistic: 0.53). UVA exhibited consistency in mortality prediction across all COVID-19 statuses, with most patients falling under medium risk (n= 860), unlike MEWS and mMPM24, where majority were at high risk (n= 1228 and n= 1224, respectively).
Conclusion:
Amid COVID-19 waves, these scores can help guide interventions for those most in need. However, their utility depends on continuous validation efforts. By recognizing their role in stratifying risk and addressing inherent limitations, due to data availability or score accuracy, these tools can be used to efficiently distribute constrained resources
THE ASSOCIATION OF SOCIAL DETERMINANTS OF HEALTH AND RISK OF INFECTION AMONG ADULTS WITH CHRONIC KIDNEY DISEASE: THE CHRONIC RENAL INSUFFICIENCY (CRIC) COHORT
Get PDFBackground: Individuals with chronic kidney disease (CKD) are at increased risk of infection. Social determinants of health (SDoH) may affect both kidney health and risk of infection, but the role of SDoH on the risk of infection has not been well-characterized in a CKD population.
Methods: Among 5,499 adults (mean age 59.5 (SD 10.7) years, 43.4% female, 44.0% non-Hispanic Black, and mean eGFR 48.4 ml/min/1.73m2) from the Chronic Renal Insufficiency Cohort (CRIC) Study, we evaluated the association of SDoH, as assessed by income, educational attainment, marital status, and health insurance statuses, with the risk of hospitalization with infection using Cox proportional hazard models in the overall population and stratified by race and ethnicity.
Results: During follow-up (median 5.7 years), 2913 participants were hospitalized with infection (incidence rate 95.76, 95% CI [92.34 – 89.05]). Black participants had higher risk of infection than White participants (IRs 107.8, 95% CI [102.2 – 113.6] and 78.0, 95% CI [73.6 – 82.7] per 1,000 person-years), but the association was attenuated in multivariable Cox model (HR 1.06, 95% CI [0.97-1.16]). By SDoH, the risk of infection was higher among those with adverse SDoH, such as lower income, less educational attainment, and never or formally married status, with the strongest association observed for income. In multivariable Cox models, the HRs were 1.77 (1.49-2.09) for income <100,000; 1.32 (1.16-1.50) for education less than high school vs. college degree, and 1.20 (1.09-1.31) for formally vs current married status. The association was not significant for insurance status (HR 0.97, 95% CI (0.81 – 1.15)), although 94.3% had insurance. These associations were consistent across race and ethnicity without significant interactions, and across infection subtypes.
Conclusions: Adverse SDoH, particularly low income, were significantly associated with an increased risk of infection across race and ethnicity groups. Addressing disparities in economic stability may be key to mitigating infection risks among adults living with CKD in the US
EFFICIENT LEARNING ALGORITHMS FOR STOCHASTIC PROCESSES, REINFORCEMENT LEARNING, AND APPLICATIONS
No full textThis thesis develops efficient learning algorithms for stochastic dynamical systems, reinforcement learning, and medical applications. These complex domains exhibit multiscale structures and high dimensionality that pose challenges like the curse of dimensionality for traditional methods. By exploiting various problem structures through dimensionality reduction, randomization, and multiscale methods, we derive scalable, sample-efficient techniques with broad applicability.
Part I of this thesis introduces a nonlinear stochastic model reduction technique for high- dimensional stochastic dynamical systems having a low-dimensional invariant effective manifold with slow dynamics, and high-dimensional, large fast modes. Given only access to short bursts of simulation, we design an algorithm that outputs an estimate of the invariant manifold, a process of the effective stochastic dynamics on it, which has averaged out the fast modes, and a simulator thereof. The algorithm and the estimation can be performed on-the-fly, leading to efficient exploration of the effective state space, without losing consistency with the underlying dynamics. This construction enables fast, efficient simulation of the effective dynamics, plus accurate estimation of crucial features and observables of such dynamics. Theoretical analysis demonstrates favorable sampling complexity scaling linearly in the high dimension of the state space, indicating that our approach overcomes the curse of dimensionality. Extensions of the technique motivated by some averaging and linear approximation theory address the nonlinearity of fast modes.
Part II of this thesis introduces a fast multiscale procedure for repeatedly compressing Markov decision processes (MDPs), wherein a hierarchy of sub-problems at different scales is automatically determined by using parametric families of policies to abstract sub-problems at finer scales and a special outer product operation to bridge those policies. Coarsened MDPs are themselves independent MDPs and may be solved using existing algorithms. The multiscale representation can lead to substantial improvements in convergence rates both locally within sub-problems and globally across sub-problems. These multiscale decompositions also yield new transfer opportunities across different levels and different problems, by summarizing useful skills and higher-order functions from previously learned policies, which further enable systematic curriculum learning. In addition, we provide additional features like virtual policies and recursion. Finally, we demonstrate all the features above in a collection of illustrative domains.
Part III of this thesis studies dyslipidemia management, where there are currently two major classifications, Fredrickson-Levy-Lee (FL) and Sniderman (S). While based on lipid profiles, both were designed using clinical information. With the availability of massive lipid profile databases, new classifications may be designed using unsupervised machine learning techniques, which extract patterns from raw unlabeled data. We use the very large database of lipids (VLDL), to (i) quantify the agreements and differences between FL and S classifications, (ii) optimize such agreement by tuning the thresholds involved, and (iii) use k-means clustering to generate, in a completely unsupervised fashion, a novel classification, which we compare to FL and S classifications
After Racialization: Neoliberalism and the Limits of Racial Justice
Get PDFThis dissertation interrogates the practices by which race becomes apparent as a seeming reflection of a reality it in fact constructs. Beginning from Frantz Fanon’s insistence that race does not ontologically exist but that racial terms constitute a discursive situation in which action unfolds, I demonstrate how absorbing that claim orients us within social space by rendering us newly aware, and thus responsible, for the ways we participate, individually and collectively, in racializing practices. If racial categories do not automatically reflect external reality, we must examine the social processes that produce reality as raced. Many theorists respond to this need by emphasizing the historical link between racialization and racist hierarchy as the very meaning of race. Turning first to the philosopher Charles Mills, then to other theorists in the social sciences and humanities, I argue that while race concepts regularly abet the construction of such hierarchy, practices of racialization and practices of racist hierarchization are not identical, even when hierarchies reflect ostensibly racial categories. As an alternative, I present a theory of racialization as racial ascription: the practice that produces race by citing seemingly inalterable characteristics (how people look; specific aspects of their ancestry) as evidence for their assignment to a racial group. Importantly, the evidence and the group are not identical. Additionally, while racial ascription often functions as an aspect of oppression, justifying the racially unequal distribution of benefits and burdens, it need not necessarily do so. Constraining our understanding of racemaking processes to those that construct race as a seemingly natural fact in turn allows us better to understand the function of race within contemporary liberalism. Antiracist liberalism holds that there are races and they must be equal. In so doing, it promotes a particular form of equality—equality across races and other identity groups—only by presupposing, and thus reinstalling, the legitimacy of nonracial forms of inequality. This possible use of race demonstrates that not all hierarchy produced by race is itself racial. Consequently, the most plausible strategy for eradicating racialization’s negative effects is to pursue antiracist egalitarianism, which seeks political, social, and economic equality for all
Development and validation of non-contrast MRI techniques for the measurement of cerebral blood volume
Get PDFThe inflow-based vascular space occupancy (iVASO) MRI technique was developed to measure arterial cerebral blood volume (CBVa). The integration of vascular crushers within iVASO amplifies signal sensitivity, particularly targeting small pial arteries and arterioles. In this dissertation, we aim to achieve several objectives: firstly, to systematically optimize and evaluate 3D iVASO sequences on both 3T and 7T MRI scanners for quantifying CBVa values in the brain; secondly, to compare and evaluate different vascular crushers and their impact on iVASO images; additionally, scans were conducted on both a 7T human scanner and an 11.7T animal scanner to compare iVASO with a contrast-based method, validating the arterial origin of iVASO signals; furthermore, in this study, iVASO was utilized to evaluate the vasodilation effect of CO2 by monitoring CBVa changes; lastly, a MATLAB toolbox was developed to standardize the iVASO data processing pipeline
MOLECULAR AND CELLULAR ANALYSIS ON GENETICALLY MODIFIED MOUSE MODELS FOR PPK
No full textPalmoplantar Keratodermas (PPKs) are a collection of rare hereditary skin disorders characterized by epidermis thickening in the palms and soles. Under some conditions, this could lead to persistent pain for human patients. In this study, we used a loss-of-function mouse model of Mal de Meleda (MdM), SLURP2X-/-, to exhibit these phenotypes and study its pain molecular mechanism to identify future therapeutic targets. While the underlying molecular mechanism of persistent pain is overall not fully understood, it is well known that the immune system is essential to the emergence and maintenance of numerous chronic pain disorders. Many immune cells, such as macrophages, mast cells (MCs), and T cells are undoubtedly important participants in immune-related pain. In addition, they are also the contributors to neuropathic pain through neuroimmune crosstalk. In our study, we focused on macrophages and analyzed their composition in PPK-affected skin tissues. Based on immune cell alterations our lab found earlier in this mouse model, we ablated macrophages through multiple methods to test out if they are essential in PPK pain hypersensitivity. However, pain sensitivity was not improved after the ablation. We also explored the effects of macrophages on other immune cell like MCs and T cells. Though there was a trend showing that increased MCs and decreased T cells in paw skin tissues of SLURP2X-/- mice, not enough evidence supports a correlation between macrophage depletion and infiltration of other immune cell types. More data will be needed to arrive at definitive answers to these questions, but as we learn more about the PPK pain mechanism and identify more useful molecular targets, we may facilitate prevention and improved treatments for many other pain-related disorders