Representation and Reinforcement Learning for Personalized Glycemic Control in Septic Patients

• Critically ill patients have the issue of poor glucose control, which includes the presence of dysglycemia and high glycemic variability. • Current clinical practice follows the guidelines suggested by the NICESugar trial to control the blood sugar level for cricital care. • However, there are overwhelming variations in clinical conditions and physiological states among patients under critical care. This limits clinicians’ ability to perform appropriate glycemic control. In addition, clinicians sometimes may not be able to consider the issue of glycemic control. • To help clinicians better address the challenge of managing patients’ glucose level, we need a personalized glycemic control strategy that can take into account the variations in patients’ physiological and pathological states. Reinforcement Learning (RL) in Clinical Domain