22689 research outputs found
Sort by
TWO TYPES OF LOCUS COERULEUS NOREPINEPHRINE NEURONS DRIVE REINFORCEMENT LEARNING
No full textThe cerebral cortex generates flexible behavior by learning. Reinforcement learning is though to be driven by error signals in midbrain dopamine neurons. However, they project more densely to basal ganglia than cortex, leaving open the possibility of another source of learning signals for cortex. The locus coeruleus (LC) contains most of the brain’s norepinephrine (NE) neurons and project broadly to cortex. We measured activity from identified mouse LC-NE neurons during a behavioral task requiring ongoing learning from reward prediction errors (RPEs). We found two types of LC-NE neurons: neurons with wide action potentials (type I) were excited by positive RPE and showed an increasing relationship with change of
choice likelihood. Neurons with thin action potentials (type II) were excited by lack of reward and showed a decreasing relationship with change of choice likelihood. Silencing LC-NE
neurons changed future choices, as predicted from the electrophysiological recordings and a model of how RPEs are used to guide learning. We reveal functional heterogeneity of a neuromodulatory system in the brain and show that NE inputs to cortex act as a quantitative learning signal for flexible behavior
THE USE OF NEAR REAL-TIME MICROBIOGRAM FOR TREATMENT DECISION OF PATIENTS WITH PRESUMED SEPSIS
No full textSepsis is a severe condition caused by a series of host responses toward infection and may eventually result in death. High occurrence, mortality, and expensive treatment make it one of the most severe diseases. Despite deeper understanding of sepsis, effective therapies have saved many patients’ lives from it, yet mortality and cost are still high compared to many other diseases. Antimicrobials are effective against infections and mitigate sepsis progression if applied early and accurately. However, such treatment is hard since it needs clinicians’ expertise and a systematic understanding of the patient information. Due to the advance of data science focused on sepsis prediction, we created a near real-time microbiogram dashboard to facilitate clinicians with antimicrobial prescriptions after they have diagnosed presumed sepsis and begin therapy. It is done by displaying environmental data in Maryland and Electronic Health Record Data from Johns Hopkins Medical Institutions together to show clinicians community infection levels for different viruses. A comprehensive keyword grouping with other data management techniques is applied using Structured Query Language (SQL) and Python for data cleaning, data grouping, and filter creation. Data are visualized with Microsoft PowerBI and ESRI ArcGIS, where the information dashboard, mapping dashboard, and user interface are displayed. The microbiogram shows infection rates with precision up to census block group spatially and up to weeks temporally, allowing treatment advice at both macroscopic and microscopic levels. The filter section and reference map layers offer clinicians the freedom to customize settings and only select points of interest. Despite some trivial limitations, the microbiogram is a great starting point that combines patients’ demographics and microbiology lab tests to provide clinicians with more information about patients’ living environments. This may help clinicians to give more accurate initial antimicrobials before blood cultures results are available, and thus reduce mortality rate, reduce the cost, and improve clinical outcomes for sepsis. In addition, it’s a good platform for public health research, which can eventually benefit sepsis treatment. Therefore, it is worthy to research more on expanding the data elements, finding relationships among those, and exploring the potential to add practicability to the microbiogram
ASSOCIATIONS BETWEEN AREA-LEVEL ARSENIC EXPOSURE AND ADVERSE BIRTH OUTCOMES: AN ECHO-WIDE COHORT ANALYSIS
No full textBackground: Drinking water is a common source of exposure to inorganic arsenic globally. In the US, the Safe Drinking Water Act (SDWA) was enacted to protect consumers from exposure to contaminants, including arsenic, in public water systems (PWS). The reproductive effects of preconception and prenatal arsenic exposure in regions with low to moderate arsenic concentrations are not well understood.
Objectives: This study examined associations between preconception and prenatal exposure to arsenic violations in water, measured via residence in a county with an arsenic violation in a regulated PWS during pregnancy, and five birth outcomes: birth weight, gestational age at birth, preterm birth, small for gestational age (SGA), and large for gestational age (LGA).
Methods: Data for arsenic violations in PWS, defined as concentrations exceeding 10 parts per billion, were obtained from the Safe Drinking Water Information System. Participants of the Environmental Influences on Child Health Outcomes Cohort Study were matched to arsenic violations by time and location based on residential history data. Multivariable, mixed effects regression models were used to assess the relationship between preconception and prenatal exposure to arsenic violations in drinking water and birth outcomes.
Results: Compared to unexposed infants, continuous exposure to arsenic from three months prior to conception through birth was associated with 88.8 grams higher mean birth weight (95% CI: 8.2, 169.5), after adjusting for individual-level confounders. No statistically significant associations were observed between any preconception or prenatal violations exposure and gestational age at birth, preterm birth, SGA, or LGA.
Conclusions: Our study did not identify associations between preconception and prenatal arsenic exposure, defined by drinking water exceedances, and adverse birth outcomes. Exposure to arsenic violations in drinking water was associated with higher birth weight. These results may be partially attributable to protective SDWA regulations. Future studies would benefit from more precise geodata of water system service areas, direct household drinking water measurements, and exposure biomarkers
ENGINEERING DRUG DELIVERY PLATFORM FOR OCULAR DRUG DELIVERY
No full textDrug delivery to the eye is important when treating diseases such as glaucoma. It is important to study the role of melanin in ocular drug delivery as it can be used to prolong the therapeutic levels of drugs in the eye. It is abundant in pigmented cells in the eye such as the retinal pigment epithelium, choroid, iris stroma, and ciliary stroma. The use of peptide conjugation and machine learning can be used to create drug conjugates with high melanin binding and high cell-penetrating properties. A synergistic effect was found between melanin binding and cell-penetrating properties. The quantification of such results like retinal gangling cell images is also an important step in understanding the efficacy of ocular drugs. However, it is a largely time-consuming process due to the variability of cell imagery. The use of object detection softer ware such as Faster R-CNN can be used to create a process that is able to quantify large batches of images with a high prediction rate. This software was used to study the efficacy of the drug peptide conjugate of HR-97 Sunitinib. HR-97 was shown to extend retinal ganglion cell protection for a longer period than just the drug Sunitinib itself. High melanin binding and cell-penetrating drug peptide conjugates can be used to increase drug retention in pigmented cells while allowing for therapeutic levels of the drug to reach the retinal ganglion cells in non-pigmented cells. This approach can be used to deliver drugs to treat ocular diseases more effectively and can make ocular drug delivery more efficient
Investigating compositional visual knowledge through challenging visual tasks
No full textHuman vision manifests remarkable robustness to recognize objects from the visual world filled with a chaotic, dynamic assortment of information. Computationally, our visual system is challenged by the enormous variability in two-dimensional projected images as a function of viewpoint, lighting, material, articulation as well as occlusion. Many past research investigated the underlying representations and computational principles that support human vision robustness with controlled and simplified visual stimuli. Nevertheless, the generality of these findings was unclear until tested on more challenging and more naturalistic stimuli.
In this thesis, I study human vision robustness with several challenging visual tasks and more naturalistic stimuli, including the recognition of occluded objects and the recognition of non-rigid human bodies from natural images of scenes. I use psychophysics, functional magnetic resonance imaging as well as computational modeling approaches to measure human vision robustness and examine the hierarchical, compositional framework as the underlying principle where the representation of the whole is composed of the representation of its parts through different hierarchies. I show that human vision has impressive abilities to recognize heavily occluded natural objects, and the human behavioral performance is better explained by compositional models rather than standard deep convolutional neural networks. In addition, I also show that human vision can rapidly and robustly extract information about spatial relationships between human body parts and discriminate three-dimensional non-rigid human poses even from a mere glance. Lastly, I show that there exists a distributed cortical network that encodes compositional pose representations with different view invariance and depth sensitivity, and the difference in these neural representations might be driven by the diversity of the supported behavior tasks. Taken together, this thesis demonstrates that human vision manifests great robustness even in these challenging visual tasks, and that the hierarchical, compositional framework may be one of the underlying principles supporting such robustness
Fluoroscopic Navigation for Robot-Assisted Orthopedic Surgery
No full textRobot-assisted orthopedic surgery has gained increasing attention due to its improved accuracy and stability in minimally-invasive interventions compared to a surgeon's manual operation. An effective navigation system is critical, which estimates the intra-operative tool-to-tissue pose relationship to guide the robotic surgical device. However, most existing navigation systems use fiducial markers, such as bone pin markers, to close the calibration loop, which requires a clear line of sight and is not ideal for patients. This dissertation presents fiducial-free, fluoroscopic image-based navigation pipelines for three robot-assisted orthopedic applications: femoroplasty, core decompression of the hip, and transforaminal lumbar epidural injections. We propose custom-designed image intensity-based 2D/3D registration algorithms for pose estimation of bone anatomies, including femur and spine, and pose estimation of a rigid surgical tool and a flexible continuum manipulator. We performed system calibration and integration into a surgical robotic platform. We validated the navigation system's performance in comprehensive simulation and ex vivo cadaveric experiments. Our results suggest the feasibility of applying our proposed navigation methods for robot-assisted orthopedic applications.
We also investigated machine learning approaches that can benefit the medical imaging analysis, automate the navigation component or address the registration challenges. We present a synthetic X-ray data generation pipeline called SyntheX, which enables large-scale machine learning model training. SyntheX was used to train feature detection tasks of the pelvis anatomy and the continuum manipulator, which were used to initialize the registration pipelines. Last but not least, we propose a projective spatial transformer module that learns a convex shape similarity function and extends the registration capture range.
We believe that our image-based navigation solutions can benefit and inspire related orthopedic robot-assisted system designs and eventually be used in the operating rooms to improve patient outcomes
Country selectivity in aid allocation: Evidence on need and effectiveness
No full textOver $200 billion was spent in foreign assistance in 2022, yet most donors do not have explicit criteria for allocating their resources. The misallocation of foreign aid resources can create huge inefficiencies that potentially stifle its effectiveness. This study produces evidence for the optimal allocation of foreign assistance across the dimensions of country need and potential effectiveness for donors that seek to maximize their impact on poverty reduction through economic growth.
My research uses quantitative methods to examine the two most relevant allocation factors for the goal of poverty reduction through growth: need and effectiveness. The first chapter reviews the aid allocation literature and proposes a conceptual framework to guide the rest of the analysis.
The second chapter explores a needs-based approach across the dimensions of the development challenge and the resources available. The two different components of country need both suggest a strong focus on allocating assistance towards the poorest countries.
The third chapter examines the differential effectiveness of aid at a macro level for different criteria employed by performance-based approaches – i.e., is foreign aid more effective in promoting economic growth in better-governed and more democratic countries? I find that the aid-growth relationship is much stronger for worse-governed and less democratic countries.
The fourth chapter exploits micro data to examine differences in project-level outcomes for both needs-based and performance-based aid allocation criteria. In find that good governance is the most important country-level factor, followed by higher average income, corruption is insignificant, and democracy may be a detriment to achieving project outcomes.
The fifth chapter concludes by comparing the current allocation of assistance to more evidence-based optimal allocation models, and I find that too much assistance goes to richer countries and strategic partners. I conclude by providing recommendations for improving the effectiveness of foreign assistance through revised allocation criteria based on the findings of the preceding chapters
Comparing Tumor-Specific CD8+ T Cells in the Bone Marrow with Tumor Infiltrating Lymphocytes: Implications for Adoptive Cell Therapy in Solid Tumors
No full textThe variety of diseases encompassed by the term “cancer” result in substantial global disease burden, comprising one of the leading causes of premature mortality. Immunotherapy, comprised largely of immune checkpoint blockade and T cell therapy, has progressed immensely in recent years, becoming an integral strategy in cancer care. Strategies employing adoptive cell therapy (ACT) utilize cells derived from a patient or from a donor to develop a cellular product for infusion. Current ACT approaches have seen the establishment of chimeric antigen receptor-T cells (CAR-T cells) using peripheral blood lymphocytes as well as tumor infiltrating lymphocytes (TILs) with significant clinical results.
Alongside marked progress, the limitations of current strategies are also emerging. Notably, current ACT modalities face challenges of persistence following infusion and duration of clinical responses. TIL therapy specifically requires candidate patients to be at an advanced disease stage, decreasing the likelihood of a positive outcome following therapeutic intervention. Additionally, although TILs have seen clinical success in limited settings, they are also a dysfunctional population of T cells. As the quality of an initial T cell population has been linked to improved outcomes following adoptive transfer, this implies that TILs may be a sub-optimal T cell source for broad clinical applications. Due to these limitations, novel approaches are needed.
The bone marrow (BM) is an immunological niche that houses T cells with specificity for previously encountered antigens, including tumor-associated antigens from certain solid cancers. BM T cells possess an enhanced reactivity to a variety of hematologic and solid cancers as compared to T cells from the peripheral blood. As TILs are an approved treatment modality for at least two types of solid tumors, t¬¬his project sought to improve our understanding of the biology of tumor-specific BM T cells in the context of solid cancer by comparing them with TILs. In doing so, we intended to further the rationale for using the BM as a source of T cells for ACT against solid malignancies.
Herein, we used the B16.OVA murine melanoma model to profile tumor specific CD8+ T cells from the BM and contrast them with TILs. Through immunophenotyping via polychromatic flow cytometry, bulk RNA-sequencing, and a combination of in vitro and in vivo assays, we demonstrate that T cells from the BM exhibit characteristics that could make them a superior source of cells for cellular therapy as compared to TILs; these include a stem-like memory phenotype, improved effector function, enhanced persistence within a tumor-bearing host, and increased tumor infiltration. Additionally, our studies indicate that the BM appears to maintain a stable population of tumor-specific T cells during disease progression. If this holds true in humans, this could make the BM a reservoir of tumor-specific T cells that is accessible in and harvestable from all patients, even at early, pre-metastatic stages of disease.
Collectively, the data presented here provide a foundation for further exploring the BM as a source of tumor-specific T cells for ACT in solid malignancies
Three Essays on Agent Interaction and Policy Interventions in the Healthcare Market
No full textThe focus of my dissertation is the interplay of agents, mediums, and policy interventions within the healthcare market. The first chapter examines the interaction of patient and physicians within the context of mental health treatment. Utilizing a model where physicians are depicted as imperfect agents and patients don't always adhere to their treatments, I find that patients value their mental health while responding to treatment costs, and physicians tend to prioritize their own benefits to a greater extent compared to patient well-being. Moreover, despite the growing focus on policies to improve patient compliance and physician altruism, I find that targeting either channel independently has limited impact on patient mental health outcomes. The second chapter explores the interaction between online and offline pharmacies. By exploiting an exogenous policy shock, we find that decreases in offline prices do not trigger a decline in online demand as predicted by channel substitution theory. Instead, our results reflect substantial increase in online demand, indicating channel complementarity. Additionally, even though the offline price reductions were localized, they prompt a nationwide decrease in prices due to the online pharmacy's policy of geographical price uniformity. The third chapter examines the repercussions of improved drug availability on the consistency with which consumers follow their prescribed treatments. We leverage a policy that significantly boosted the availability of certain drugs in 11 cities to reveal that this policy led to a 43% higher increase in non-compliance in these cities compared to other areas
AN EXAMINATION OF THE ENGLISH LANGUAGE ARTS CLASSROOM AND MODERN EDUCATIONAL POLICY TO DEVELOP A WRITING WORKSHOP
No full textThe modern English Language Arts (ELA) classroom has changed substantially due to state and federal demands such as the Common Core State Standards and high-stakes testing (Gilbert, 2020). As a result of these changes, teachers adjust their instructional approach—with some allowing these demands to replace their pedagogical knowledge and teaching experience (Papola-Ellis, 2014), whereas others strike a middle ground or completely resist the demands placed upon them (Eisenbach, 2012). This study investigated the effects of those demands on modern ELA teachers as well as teachers’ self-efficacy for writing and teaching writing. The needs assessment (mixed methods) revealed that teachers were torn when attempting to meet the demands of high-stakes testing, reporting that the test did not allow for teaching creativity or promote student engagement. Further, teachers expressed distrust with the state’s testing process and state-provided preparation materials. The most salient factor, however, was that teachers reported higher levels of self-efficacy for teaching writing (M=4.06; Likert scale of 1 to 5) but reported lower levels with self-efficacy for writing themselves (M=3.75). Following the needs assessment results, the next component of this study sought to investigate professional development aimed at improving teachers’ self-efficacy for writing and writing enjoyment. The National Writing Project (NWP) provided a gold standard for professional development in the area of writing, resting on the philosophy that successful teachers of writing must also be writers themselves (Andrews, 2008); however, due to the time commitment demands of NWP (five weeks) and questionable scalability (Meyer, 2003), the final two chapters aimed to develop a shorter and context-specific writing workshop for teachers. The writing workshop framework leaned on three core ideas: Darling-Hammond, Hyler, and Gardner’s (2017) Seven Elements of Effective Professional Development; Bandura’s (2002) work on self-efficacy, and the intervention literature by Locke, Whitehead, and Dix (2013) and Whitney (2008). At the conclusion of this study, a facilitator possesses the framework and ingredients to implement a one-week writing workshop to nourish two distinct identities: teachers of writing and teacher writers