Georgia Clinical & Translational Science Alliance (GaCTSA)

使用机器学习和 自适应采样 摘要 社区管理和应对COVID-19的一项关键任务是进行诊断和/或抗体检测， 识别患病个体。这些信息对于公共卫生官员估计流行率和传播至关重要， 并有效地规划所需的资源，如ICU病床，呼吸机，个人防护设备和医疗 工作人员此外，关于感染人数的信息可用于开发概率和统计模型 估计疾病的繁殖数量，并预测疾病爆发的可能空间和时间轨迹。 这为规划行动和制定社交距离、学校教育和社会服务的政策和指导方针提供了重要信息。 关闭，远程工作，社区封锁等，尽管诊断测试和识别的重要性， 由于检测试剂盒和资源数量有限，大规模检测是一项具有挑战性的任务。我们 建议的研究重点是开发基于机器学习的分配策略，以确定最佳的 COVID-19检测中心的位置，包括移动的和卫星中心，以最大限度地减少本地和全球预测 不确定性，最大限度地扩大地理覆盖范围，与时空爆发轨迹的预测相关联， 提高了疾病病例识别效率。

EFFECTIVE ALLOCATION OF TEST CENTERS FOR COVID-19 USING MACHINE LEARNING AND ADAPTIVE SAMPLING ABSTRACT A critical task in managing and dealing with COVID-19 in communities is to perform diagnostic and/or antibody tests to identify diseased individuals. This information is critical to public health officials to estimate prevalence and transmission, and to effectively plan for required resources such as ICU beds, ventilators, personal protective equipment, and medical staff. Additionally, information on the number of infected people can be used to develop probabilistic and statistical models to estimate the reproduction number of the disease, and to predict the likely spatial and temporal trajectories of the outbreak. This provides vital information for planning actions and preparing policies and guidelines for social-distancing, school closures, remote work, community lockdown, etc. Despite the importance of diagnostic testing and identification of the positive cases, broad-scale testing is a challenging task particularly due to the limited number of test kits and resources. Our proposed research focuses on the development machine learning-based allocation strategies for determining the optimal location of COVID-19 test centers, including mobile and satellite centers, to minimize the local and global prediction uncertainties, maximize geographic coverage, associated with projections of spatio-temporal outbreak trajectories, and to improve efficient identification of diseased cases.