Semi-Autonomous and autonomous cerebral physiologic artifact management platforms

汽车保险公司承担着为涉及机动车辆相关伤害案件的客户支付费用的角色。因此，准确估计和预测客户/费率付款人受伤后的临床轨迹是保险公司感兴趣的，因为它有助于预测潜在的年度政策支出，并为这些公司提供战略投资的能力。特异性损伤病例在轨迹预测方面比其他病例具有更大的不确定性，急性神经损伤（即神经损伤）。中度/重度创伤性脑损伤（TBI）是当前短期和长期结果和支持服务需求预测较差的典型情况。此外，这些人口承担了大量的短期和长期财政支助费用，加拿大每年的财政支助费用超过170亿加元。曼尼托巴公共保险公司（MPI; Crown Corporation）已经确定了对这一人群的客户轨迹建模的新方法的需求。最近的急性神经损伤的轨迹建模已经证明了添加急性期(即。前两周)从床边监测设备获得的大脑生理数据，MPI热衷于将其作为一种新方法来探索，以潜在地改进这些模型，从而帮助企业进行财务规划。然而，这样的大脑生理数据流往往是庞大的，复杂的，充满了人为的错误，限制了它们在大多数轨迹模型中的包含。人为误差包括患者运动噪声、护理程序噪声和一般计算机连接错误。目前，数据工件管理需要人工清理，这种方法不利于客户结果建模的有效数据使用，也不是保险公司的可行选择。利用全球数据科学、生理学专业知识和现有数据源，MPI将与Zeiler博士合作，通过开发和验证自主计算机驱动的脑生理学人工管理解决方案，弥合现有的知识差距，这将有助于自动生成干净的数据，并将其纳入未来的客户轨迹建模，帮助公司进行财务规划。该项目将培养2名硕士研究生和1名博士研究生，以及数名本科生。

Auto-insurers carry the role of disbursement for clients involved in motor vehicle related injury cases. As such, accurate estimation and prediction of client/rate-payer clinical trajectory post-injury is of interest to insurers, as it facilitates forecasting potential annual policy disbursements and provides the ability for such corporations to be strategic with investments. Specific injury cases carry more uncertainty than others in trajectory prediction, with acute neural injury (ie. moderate/severe traumatic brain injury (TBI)) being the exemplar situation where current short- and long-term outcome and support services requirement predication is poor. Furthermore, such populations carry substantial short- and long-term financial support costs, which are over 17 billion CAD annually in Canada. Manitoba Public Insurance (MPI; Crown Corporation) has identified the need for novel approaches to client trajectory modelling within this population. Recent trajectory modelling in acute neural injury has demonstrated the benefit of adding acute-phase (ie. first 2 weeks) brain physiology data from bedside monitoring devices, and MPI is keen to explore this as a novel means to potentially improve such models that will aid with corporate financial planning. However, such cerebral physiologic data streams are often large, complex and rife with artifactual errors, limiting their inclusion in most trajectory models. Artifact errors take the form of patient motion noise, nursing procedural noise and general computer connection errors. Currently, data artifact management requires manual cleaning, with such methods not conducive to efficient data usage for client outcome modelling, nor a viable option for an insurer. Leveraging the global data science, physiologic expertise and existing data sources, MPI will work with Dr. Zeiler to bridge the existing knowledge gap through the development and validation of autonomous computer-driven solutions for cerebral physiology artifact management, that would facilitate automated generation of clean data to be incorporated into future client trajectory modelling aiding with financial planning of the corporation. This project will provide training for 2x MSc and 1x PhD graduate students, and several undergraduate students.