Systematic conservation prioritization with the prioritizr R package.

Plans for expanding protected area systems (prioritizations) need to fulfill conservation objectives. They also need to account for other factors, such as economic feasibility and anthropogenic land-use requirements. Although prioritizations are often generated with decision support tools, most tools have limitations that hinder their use for decision-making. We outlined how the prioritizr R package (https://prioritizr.net) can be used for systematic conservation prioritization. This decision support tool provides a flexible interface to build conservation planning problems. It can leverage a variety of commercial (e.g., Gurobi) and open-source (e.g., CBC and SYMPHONY) exact algorithm solvers to identify optimal solutions in a short period. It is also compatible with a variety of spatially explicit (e.g., ESRI Shapefile, GeoTIFF) and nonspatial tabular (e.g., Microsoft Excel Spreadsheet) data formats. Additionally, it provides functionality for evaluating prioritizations, such as assessing the relative importance of different places selected by a prioritization. To showcase the prioritizr R package, we applied it to a case study based in Washington state (United States) for which we developed a prioritization to improve protected area coverage of native avifauna. We accounted for land acquisition costs, existing protected areas, places that might not be suitable for protected area establishment, and spatial fragmentation. We also conducted a benchmark analysis to examine the performance of different solvers. The prioritization identified 12,400 km2 of priority areas for increasing the percentage of species' distributions covered by protected areas. Although open source and commercial solvers were able to quickly solve large-scale conservation planning problems, commercial solvers were required for complex, large-scale problems.. The prioritizr R package is available on the Comprehensive R Archive Network (CRAN). In addition to reserve selection, it can inform habitat restoration, connectivity enhancement, and ecosystem service provisioning. It has been used in numerous conservation planning exercises to inform best practices and aid real-world decision-making.

Priorización de la conservación sistemática con el paquete prioritizr R Resumen Los planes para expandir los sistemas de áreas protegidas (priorizaciones) necesitan cumplir con los objetivos de conservación. También necesitan considerar otros factores, como la viabilidad económica y los requerimientos para el uso antropogénico del suelo. Aunque con frecuencia las priorizaciones se generan con herramientas de apoyo para decidir, la mayoría de estas herramientas tienen limitantes que complican su uso en la toma de decisiones. Esbozamos cómo el paquete prioritizr R (https://prioritizr.net) puede usarse para la priorización de la conservación sistemática. Esta herramienta de apoyo para decidir proporciona una interfaz flexible para construir problemas de la planeación de la conservación. También puede sacar provecho de una variedad de solucionadores exactos de algoritmos comerciales (p. ej.: Gurobi) y de fuentes abiertas (p. ej.: CBC y SYMPHONY) para identificar soluciones óptimas en un periodo breve. La herramienta también es compatible con una variedad de formatos de datos tabulares con espacialidad explícita (p. ej.: ESRI Shapefile, GeoTIFF) y sin espacialidad (p. ej.: hojas de cálculo de Microsoft Excel). Además, proporciona la funcionalidad para evaluar las priorizaciones, como el análisis de la importancia relativa de los diferentes lugares seleccionados por una priorización. Para mostrar la funcionalidad del paquete prioritizr R, lo aplicamos a un estudio de caso en el Estado de Washington, Estado Unidos, para el cual desarrollamos una priorización para mejorar la cobertura del área protegida de la avifauna nativa. Consideramos los costos de adquisición de tierras, las áreas protegidas existentes y la fragmentación espacial. También realizamos un análisis comparativo para examinar el desempeño de los diferentes solucionadores. La priorización identificó 12,400 km2 de áreas prioritarias para incrementar el porcentaje de la distribución de especies cubiertas por las áreas protegidas. Aunque los solucionadores comerciales y de fuente abierta lograron resolver rápidamente los problemas de conservación a gran escala, sólo los comerciales fueron requeridos para los problemas complejos de gran escala. El paquete prioritizr R está disponible en el Comprehensive R Archive Network (CRAN). Además de seleccionar las reservas, el paquete puede informar la restauración de hábitat, la mejora de la conectividad y el suministro de servicios ambientales. El paquete se ha usado en varios ejercicios para informar las mejores prácticas y ayudar a la toma de decisiones en el mundo real.

A requisite role for induced regulatory T cells in tolerance based on expanding antigen receptor diversity.

Although both natural and induced regulatory T (nTreg and iTreg) cells can enforce tolerance, the mechanisms underlying their synergistic actions have not been established. We examined the functions of nTreg and iTreg cells by adoptive transfer immunotherapy of newborn Foxp3-deficient mice. As monotherapy, only nTreg cells prevented disease lethality, but did not suppress chronic inflammation and autoimmunity. Provision of Foxp3-sufficient conventional T cells with nTreg cells reconstituted the iTreg pool and established tolerance. In turn, acute depletion of iTreg cells in rescued mice resulted in weight loss and inflammation. Whereas the transcriptional signatures of nTreg and in vivo-derived iTreg cells were closely matched, there was minimal overlap in their T cell receptor (TCR) repertoires. Thus, iTreg cells are an essential nonredundant regulatory subset that supplements nTreg cells, in part by expanding TCR diversity within regulatory responses.

Privacy preservation for federated learning in health care.

Artificial intelligence (AI) shows potential to improve health care by leveraging data to build models that can inform clinical workflows. However, access to large quantities of diverse data is needed to develop robust generalizable models. Data sharing across institutions is not always feasible due to legal, security, and privacy concerns. Federated learning (FL) allows for multi-institutional training of AI models, obviating data sharing, albeit with different security and privacy concerns. Specifically, insights exchanged during FL can leak information about institutional data. In addition, FL can introduce issues when there is limited trust among the entities performing the compute. With the growing adoption of FL in health care, it is imperative to elucidate the potential risks. We thus summarize privacy-preserving FL literature in this work with special regard to health care. We draw attention to threats and review mitigation approaches. We anticipate this review to become a health-care researcher's guide to security and privacy in FL.

Sequential assessment of troponin in the diagnosis of myocardial infarction.

According to the American Heart Association, cardiovascular disease accounts for more than one third of all deaths in the United States. 1 The purpose of this retrospective case-control study was to determine which sample taken in a sequential draw was most important in diagnosing an acute myocardial infarction (AMI). One-hundred subjects were selected from a convenience sample. The "risk" of AMI diagnosis was modeled using binary multiple logistic regression. Overall, 78% (39 out of 50 cases) were diagnosed with an AMI at Tinitiai. Clearly, the initial cTnI assay is the most critical of the four sequential time points for the accurate assessment of the presence or absence of an AMI. Most importantly, sequential troponin testing increased the ability to diagnose AMI by 10-fold.

Effect of Head Trauma-Related Deaths on Corneal Endothelial Cell Loss in Eye Bank Donors.

PURPOSE: The purpose of this study was to examine the effect of head trauma-related deaths on corneal endothelial cell density (ECD) in eye bank donors. METHODS: This is a retrospective study of 287 corneas from donors with causes of death secondary to motor vehicle accident with sustained head trauma (n = 50), gunshot wound to the head (n = 138), fall with sustained head trauma (n = 2), and non-head-related traumatic causes of death (n = 97). Donors older than age 50 years were excluded due to concern for undiagnosed Fuchs endothelial dystrophy as a potential confounder for the cause of endothelial cell loss. Donor characteristics, ECD, and focal endothelial cell loss on specular microscopy were compared between the groups. Donors in the head trauma and nonhead trauma groups were matched by age; there were 42 age-matched donors in both groups. RESULTS: Age and ECD were negatively correlated (Pearson correlation coefficient = -0.57). Death-to-preservation time was not significantly different between the 2 groups ( P value = 0.59). The mean ECD in the head trauma group was 2859 ± 370 cells/mm 2 and 3041 ± 464 cells/mm 2 in the nonhead trauma group. The head trauma group had a lower ECD (178 ± 70 cells/mm 2 , P value = 0.013). After matching for age, the difference in ECD between the 2 groups was -94 ± 82 cells/mm 2 ( P value = 0.26). The adjusted odds of having focal endothelial cell loss was not statistically significant ( P value = 0.50) between the groups. CONCLUSIONS: After statistical adjustments, there were no differences between the head trauma and nonhead trauma groups.

Requirements for growth and IL-10 expression of highly purified human T regulatory cells.

Human regulatory T cells (T(R)) cells have potential for the treatment of a variety of immune mediated diseases but the anergic phenotype of these cells makes them difficult to expand in vitro. We have examined the requirements for growth and cytokine expression from highly purified human T(R) cells, and correlated these findings with the signal transduction events of these cells. We demonstrate that these cells do not proliferate or secrete IL-10 even in the presence of high doses of IL-2. Stimulation with a superagonistic anti-CD28 antibody (clone 9.3) and IL-2 partially reversed the proliferative defect, and this correlated with reversal of the defective calcium mobilization in these cells. Dendritic cells were effective at promoting T(R) cell proliferation, and under these conditions the proliferative capacity of T(R) cells was comparable to conventional CD4 lymphocytes. Blocking TGF-ß activity abrogated IL-10 expression from these cells, while addition of TGF-ß resulted in IL-10 production. These data demonstrate that highly purified populations of T(R) cells are anergic even in the presence of high doses of IL-2. Furthermore, antigen presenting cells provide proper co-stimulation to overcome the anergic phenotype of T(R) cells, and under these conditions they are highly sensitive to IL-2. In addition, these data demonstrate for the first time that TGF-ß is critical to enable human T(R) cells to express IL-10.

Investigating social connection as a protective factor against exam stress in college students.

Objective: To examine social connection as a protective factor against exam stress. Participants: 55 undergraduate students at two universities. Methods: Students were evaluated on an exam day for their hardest class and at baseline, a day in a week where they had no exams. Social connection, salivary cortisol, perceived stress, and cognitive control (measured with the Stroop test) were assessed. Exam scores were later reported. Results: Higher social connection was associated with lower perceived stress on exam day. At a small liberal arts school, higher levels of social connection were associated with higher Stroop scores. This correlation with cognitive control was not significant at a large public university. Conclusions: These findings indicate that social connection may be a protective factor in mitigating perceived stress and cognitive control capabilities may help facilitate reduced exam stress in some school environments.

Impact of Telestroke Implementation on Emergency Department Transfer Rate.

BACKGROUND AND OBJECTIVES: Telestroke networks are associated with improved outcomes from acute ischemic stroke (AIS) and facilitate greater access to care, particularly in underserved regions. These networks also have the potential to influence patient disposition through avoiding unnecessary interhospital transfers. This study examines the effect of implementation of the VA National Telestroke Program (NTSP) on interhospital transfer among Veterans. METHODS: We analyzed patients with AIS presenting to the emergency departments of 21 VA hospitals before and after telestroke implementation. Transfer rates were determined through review of administrative data and chart review and patient and facility-level characteristics were collected to identify predictors of transfer. Comparisons were made using t test, Wilcoxon rank sum, and χ 2 analysis. Multivariable logistic regression with sensitivity analysis was conducted to assess the influence of telestroke implementation on transfer rates. RESULTS: We analyzed 3,488 stroke encounters (1,056 pre-NTSP and 2,432 post-NTSP). Following implementation, we observed an absolute 14.4% decrease in transfers across all levels of stroke center designation. Younger age, higher stroke severity, and shorter duration from symptom onset were associated with transfer. At the facility level, hospitals with lower annual stroke volume were more likely to transfer; 1 hospital saw an increase in transfer rates following implementation. After adjusting for patient and facility characteristics, the implementation of VA NTSP resulted in a nearly 60% reduction in odds of transfer (odds ratio 0.39 [0.19, 0.77]). DISCUSSION: In addition to improving treatment in acute stroke, telestroke networks have the potential to positively affect the efficiency of interhospital networks through disposition optimization and the avoidance of unnecessary transfers.

The federated tumor segmentation (FeTS) tool: an open-source solution to further solid tumor research.

Objective.De-centralized data analysis becomes an increasingly preferred option in the healthcare domain, as it alleviates the need for sharing primary patient data across collaborating institutions. This highlights the need for consistent harmonized data curation, pre-processing, and identification of regions of interest based on uniform criteria.Approach.Towards this end, this manuscript describes theFederatedTumorSegmentation (FeTS) tool, in terms of software architecture and functionality.Main results.The primary aim of the FeTS tool is to facilitate this harmonized processing and the generation of gold standard reference labels for tumor sub-compartments on brain magnetic resonance imaging, and further enable federated training of a tumor sub-compartment delineation model across numerous sites distributed across the globe, without the need to share patient data.Significance.Building upon existing open-source tools such as the Insight Toolkit and Qt, the FeTS tool is designed to enable training deep learning models targeting tumor delineation in either centralized or federated settings. The target audience of the FeTS tool is primarily the computational researcher interested in developing federated learning models, and interested in joining a global federation towards this effort. The tool is open sourced athttps://github.com/FETS-AI/Front-End.

OpenFL: the open federated learning library.

Objective.Federated learning (FL) is a computational paradigm that enables organizations to collaborate on machine learning (ML) and deep learning (DL) projects without sharing sensitive data, such as patient records, financial data, or classified secrets.Approach.Open federated learning (OpenFL) framework is an open-source python-based tool for training ML/DL algorithms using the data-private collaborative learning paradigm of FL, irrespective of the use case. OpenFL works with training pipelines built with both TensorFlow and PyTorch, and can be easily extended to other ML and DL frameworks.Main results.In this manuscript, we present OpenFL and summarize its motivation and development characteristics, with the intention of facilitating its application to existing ML/DL model training in a production environment. We further provide recommendations to secure a federation using trusted execution environments to ensure explicit model security and integrity, as well as maintain data confidentiality. Finally, we describe the first real-world healthcare federations that use the OpenFL library, and highlight how it can be applied to other non-healthcare use cases.Significance.The OpenFL library is designed for real world scalability, trusted execution, and also prioritizes easy migration of centralized ML models into a federated training pipeline. Although OpenFL's initial use case was in healthcare, it is applicable beyond this domain and is now reaching wider adoption both in research and production settings. The tool is open-sourced atgithub.com/intel/openfl.

Federated learning enables big data for rare cancer boundary detection.

Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.

Affinity-based selection of regulatory T cells occurs independent of agonist-mediated induction of Foxp3 expression.

Natural regulatory T (nT(reg)) cells recognize self-peptides with high affinity, yet the understanding of how affinity influences their selection in the thymus is incomplete. We use altered peptide ligands in transgenic mice and in organ culture to create thymic environments spanning a broad range of ligand affinity. We demonstrate that the nT(reg) TCR repertoire is shaped by affinity-based selection, similar to conventional T cells. The effect of each ligand on the two populations is distinct, consistent with early nT(reg) cell lineage specification. Foxp3 expression is an independent process that does not rely on "high affinity" binding per se, but requires a high-potency agonistic interaction for its induction. The timing of ligand exposure, TGFbeta signaling, and the organization of the thymic architecture are also important. The development of nT(reg) cells is therefore a multistep process in which ligand affinity, potency, and timing of presentation all play a role in determining cell fate.

A central role for induced regulatory T cells in tolerance induction in experimental colitis.

In addition to thymus-derived or natural T regulatory (nT(reg)) cells, a second subset of induced T regulatory (iT(reg)) cells arises de novo from conventional CD4(+) T cells in the periphery. The function of iT(reg) cells in tolerance was examined in a CD45RB(high)CD4(+) T cell transfer model of colitis. In situ-generated iT(reg) cells were similar to nT(reg) cells in their capacity to suppress T cell proliferation in vitro and their absence in vivo accelerated bowel disease. Treatment with nT(reg) cells resolved the colitis, but only when iT(reg) cells were also present. Although iT(reg) cells required Foxp3 for suppressive activity and phenotypic stability, their gene expression profile was distinct from the established nT(reg) "genetic signature," indicative of developmental and possibly mechanistic differences. These results identified a functional role for iT(reg) cells in vivo and demonstrated that both iT(reg) and nT(reg) cells can act in concert to maintain tolerance.

Functional reprogramming of the primary immune response by T cell receptor antagonism.

The T cell receptor must translate modest, quantitative differences in ligand binding kinetics into the qualitatively distinct signals used to determine cell fate. Here, we use mice that express an endogenous T cell receptor (TCR) antagonist and an adoptive transfer system to examine the influence of TCR signal quality on the development of effector function. We show that activation of antigen-specific T cells in the presence of an antagonist results in a functional reprogramming of the primary immune response, marked by altered T cell homing, a failure to develop effector function, and ultimately clonal elimination by apoptosis. Importantly, antagonism does not block cell division, implying that the signals promoting clonal expansion and effector differentiation are distinct.

Impaired survival of peripheral T cells, disrupted NK/NKT cell development, and liver failure in mice lacking Gimap5.

The loss of Gimap5 (GTPase of the immune-associated protein 5) gene function is the underlying cause of lymphopenia and autoimmune diabetes in the BioBreeding (BB) rat. The in vivo function of murine gimap5 is largely unknown. We show that selective gene ablation of the mouse gimap5 gene impairs the final intrathymic maturation of CD8 and CD4 T cells and compromises the survival of postthymic CD4 and CD8 cells, replicating findings in the BB rat model. In addition, gimap5 deficiency imposes a block of natural killer (NK)- and NKT-cell differentiation. Development of NK/NKT cells is restored on transfer of gimap5(-/-) bone marrow into a wild-type environment. Mice lacking gimap5 have a median survival of 15 weeks, exhibit chronic hepatic hematopoiesis, and in later stages show pronounced hepatocyte apoptosis, leading to liver failure. This pathology persists in a Rag2-deficient background in the absence of mature B, T, or NK cells and cannot be adoptively transferred by transplanting gimap5(-/-) bone marrow into wild-type recipients. We conclude that mouse gimap5 is necessary for the survival of peripheral T cells, NK/NKT-cell development, and the maintenance of normal liver function. These functions involve cell-intrinsic as well as cell-extrinsic mechanisms.

Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data.

Several studies underscore the potential of deep learning in identifying complex patterns, leading to diagnostic and prognostic biomarkers. Identifying sufficiently large and diverse datasets, required for training, is a significant challenge in medicine and can rarely be found in individual institutions. Multi-institutional collaborations based on centrally-shared patient data face privacy and ownership challenges. Federated learning is a novel paradigm for data-private multi-institutional collaborations, where model-learning leverages all available data without sharing data between institutions, by distributing the model-training to the data-owners and aggregating their results. We show that federated learning among 10 institutions results in models reaching 99% of the model quality achieved with centralized data, and evaluate generalizability on data from institutions outside the federation. We further investigate the effects of data distribution across collaborating institutions on model quality and learning patterns, indicating that increased access to data through data private multi-institutional collaborations can benefit model quality more than the errors introduced by the collaborative method. Finally, we compare with other collaborative-learning approaches demonstrating the superiority of federated learning, and discuss practical implementation considerations. Clinical adoption of federated learning is expected to lead to models trained on datasets of unprecedented size, hence have a catalytic impact towards precision/personalized medicine.

The utility of joinpoint regression for estimating population parameters given changes in population structure.

The method of joinpoint regression has been used in numerous domains to assess changes in time series data, including such things as cancer mortality rates, motor vehicle collision mortalities, and disease risk. To help improve estimation of population parameters for use in ecological risk assessment and management, we present a simulation and analysis to describe the utility of this method for the ecological domain. We demonstrate how joinpoint regression can accurately identify if the population structure changes based on time series of abundance, as well as identify when this change occurs. In addition, we compare and contrast population parameter estimates derived through joinpoint and surplus production methods to those derived from standard surplus production methods alone. When considering a change point at 32 years (out of a 64 year simulation), the joinpoint regression model was able, on average, to estimate a joinpoint time of 32.31 years with a variance of 6.82 and 95% confidence interval for the mean relative bias of (0.0085, 0.0112). The model was able to consistently estimate population parameters, with variance of these estimations decreasing as the change in these population parameters increased. We conclude that joinpoint regression be added to the list of methods employed by those who assess ecological risk to allow for a more accurate and complete understanding of population dynamics.

Multi-Institutional Deep Learning Modeling Without Sharing Patient Data: A Feasibility Study on Brain Tumor Segmentation.

Deep learning models for semantic segmentation of images require large amounts of data. In the medical imaging domain, acquiring sufficient data is a significant challenge. Labeling medical image data requires expert knowledge. Collaboration between institutions could address this challenge, but sharing medical data to a centralized location faces various legal, privacy, technical, and data-ownership challenges, especially among international institutions. In this study, we introduce the first use of federated learning for multi-institutional collaboration, enabling deep learning modeling without sharing patient data. Our quantitative results demonstrate that the performance of federated semantic segmentation models (Dice=0.852) on multimodal brain scans is similar to that of models trained by sharing data (Dice=0.862). We compare federated learning with two alternative collaborative learning methods and find that they fail to match the performance of federated learning.