Highly Scalable Personalized Treatment Recommendation in Critical Care

Highly Scalable Personalized Treatment Recommendation in Critical Care Chandra Prasetyo Utomo Department of Informatics Universitas YARSI Jakarta, Indonesia [email protected] Kohei Ichikawa Division of Information Science Nara Institute of Science and Technology Nara, Japan ichikaw[email protected] Soratouch Pornmaneerattanatri Division of Information Science Nara Institute of Science and Technology Nara, Japan [email protected] Kundjanasith Thonglek Applied Information Systems Research Division Osaka University Osaka, Japan [email protected] Nashuha Insani Department of Informatics Universitas YARSI Jakarta, Indonesia [email protected] Martha Riskiaty Department of Informatics Universitas YARSI Jakarta, Indonesia [email protected] Abstract—Universitas YARSI and Nara Institute of Science and Technology (NAIST) are collaborating on a project focused on developing machine learning models for medical healthcare. In an era where personalized medicine is becoming increasingly important, the project aims to establish new treatment methods that effectively utilize both edge computing and cloud computing technologies in machine learning. A key focus of this collaboration is optimizing patient care in Intensive Care Units (ICUs). The project’s goal is to build a reinforcement learning model that maximizes patient survival probability. This model will be based on the analysis of timeseries vital data and medical treatment records obtained from ICU patients. One of the challenges of the project is the need to finely cluster a vast number of patients based on their vital data and responses to medical treatments. Subsequently, the optimal medical treatment for each cluster needs to be determined. To achieve this, extensive parallel computing is essential for clustering based on numerous parameters and for the optimization of the corresponding reinforcement learning models. In addressing these challenges, YARSI is primarily responsible for developing the algorithms for the reinforcement learning model construction. Meanwhile, NAIST is focusing on optimizing these processes on distributed systems. Index Terms—reinforcement learning, parallel computing, distributed systems, critical care, personalized medicine