Diabetes patient blood glucose management method and system based on reinforcement learning, medium and terminal

Abstract

The invention provides a diabetes patient blood glucose management method and system based on reinforcement learning, a medium and a terminal. The method comprises the following steps: training a decision network model based on a reinforcement learning algorithm to obtain a trained decision network model; acquiring current state information of a diabetes patient; and sending the current state information to the trained decision network model, so that the trained decision network model determines the insulin injection amount corresponding to the diabetes patient at the next future moment based on the current state information, and blood glucose management of the diabetic patient is achieved. According to the method, the decision network model is trained by a reinforcement learning algorithm based on historical blood sugar data and future carbohydrate intake of the diabetes patient, so that the relationship between carbohydrate and blood sugar change is utilized from historical information of the diabetes patient, the purpose of better controlling blood sugar is achieved, and the conditions of irregular intake time and intake amount of carbohydrates can be better adapted.

Description

technical field [0001] The present invention relates to the field of physics, in particular to measurement technology, in particular to a method, system, medium and terminal for blood sugar management of diabetic patients based on reinforcement learning. Background technique [0002] Diabetes has become one of the chronic diseases that seriously endanger national health. Since diabetes itself is difficult to cure, and related complications will have a serious impact on patients, the treatment of diabetes has gradually become the focus of attention. [0003] Artificial pancreas is currently one of the best ways to achieve closed-loop control of blood sugar in diabetic patients. It consists of three parts, continuous blood glucose monitoring device (CGMS), insulin pump and control algorithm; among them, CGMS mainly collects relevant blood sugar data from patients , the control algorithm inputs these data into the preset algorithm according to the time series data obtained by C...

AI Technical Summary

Problems solved by technology

[0008] In the latest study, Fox I et al. proposed a blood glucose control method based on DQN, which is based on the design of artificial pancreas technology, and uses the blood glucose value measured by the continuous blood glucose monitoring device to design the patient's state. After the blood sugar level and insulin injection level, the insulin level to be injected at the next moment can be given. This method does not distinguish between basal insulin, which is used to maintain the body’s blood sugar stability, and bolus, which is used to offset the rise in blood sugar caused by intake. Insulin dosage; Sun Q et al. proposed a blood glucose control model based on the Actor-Critic structure, using the blood glucose information of the patient on the previous day to predict the next day's basal insulin dosage (basal rate) and the insulin carbohydrate ratio of three meals (insulin-to -carbohydrate ratio, CIR), the blood glucose data required by this method can not only be the blood glucose value measured by the continuous blood glucose monitoring device, but also a small amount of blood glucose data obtained by manual monitoring, but this method only uses the high blood glucose of the previous day Glycemic and hypoglycemic features that do not take into account correlations between patient glycemic sequences

[0009] Moreover, the above-mentioned reinforcement learning method does not take carbohydrate intake into consideration, but only judges the appropriate amount of insulin injection based on changes in blood sugar. In reality, it is usually known that the next amount of carbohydrates consumed

Images