Automated Computational Modeling and Adaptive Control for Critical Patient Resuscitation

医生面临着巨大的挑战，要根据快速变化的生命体征做出决定，抢救和稳定危重和受伤的患者。该项目将研究生理测量的自动化处理，使用高速计算模型提供近乎即时的治疗建议，以稳定血压、心输出量和肾功能。此外，将开发反馈控制算法，能够根据患者的特定生理状态自动调节药物和液体的管理，以优化危重护理。在持续监测和调整个性化治疗方面的这一进展预计将使遭受创伤、烧伤、感染和休克的患者受益。由此产生的决策辅助系统和自适应闭环式药物和液体输送系统将极大地提高重症监护治疗的准确性和可靠性，从而提高存活率和改善治疗结果。该项目将在注射血管活性药物和液体疗法的情况下，开发心血管和液体反应的适应性模型。相应的药物和液体给药问题是由变化的生理动态反应和由于药物吸收引起的反应的显著时延所挑战的。将调查多模型观察者，为医生提供即时剂量建议，以实现血压、心输出量和尿量的目标值。这些模型将计算患者对各种药物和液体的反应性，并将自我适应随着时间的推移(患者内和患者间的变异性)不同患者对治疗的不同反应。将开发检测算法，以识别患者生理反应的潜在突然变化，如内出血的存在，并向医生发出警报。此外，还将开发基于模型的自适应和稳健的闭环药物输注算法，以自动化给药过程，以优化患者复苏。研究小组将与医学专家合作，提供动物实验的生理数据，并协助评估开发的模型和决策支持算法。

Physicians face enormous challenges in making decisions to resuscitate and stabilize critically ill and injured patients based on rapidly changing vital signs. This project will investigate automated processing of physiological measurements, using high-speed computational models to provide near-instantaneous treatment recommendations to stabilize blood pressure, cardiac output and renal function. Additionally, feedback control algorithms will be developed capable of automatically regulating the administration of drugs and fluids to optimize critical care, subject to the specific physiological state of the patient. This advance in continually monitored and adjusted personalized treatment is expected to benefit patients suffering from trauma, burn, infection, and shock. The resulting decision assistance system and adaptive closed-loop drug and fluid delivery system will greatly benefit the accuracy and reliability of critical care treatments, resulting in increased survival rates and improved therapeutic outcomes. The project will develop adaptive models of cardio-vascular and fluid response, subject to the injection of vasoactive drugs and fluid therapy. The corresponding drug and fluid administration problem is challenged by a changing physiological dynamic response and a significant time-delay in the response due to drug absorption. Multi-model observers will be investigated to provide instantaneous dosage recommendations to doctors to achieve targeted values of blood pressure, cardiac output and urinary output. The models will compute the patient's responsiveness to various drugs and fluids, and will self-adapt to varying responses to treatment from patient-to-patient and within a single patient over time (intra-patient and inter-patient variability). Detection algorithms will be developed to identify potential sudden changes in the patient's physiological response, such as the presence of an internal hemorrhage, and alert the doctors. Additionally, model-based adaptive and robust closed-loop drug infusion algorithms will be developed to automate the drug administration process for optimized patient resuscitation. The research team will collaborate with medical experts that will provide physiological data from animal experiments and will assist in the evaluation of the developed models and decision support algorithms.