CN-121971047-A - Method, system and equipment for adjusting circadian rhythm of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning

CN121971047ACN 121971047 ACN121971047 ACN 121971047ACN-121971047-A

Abstract

The application discloses a method, a system and equipment for adjusting circadian rhythms of conscious disturbance patients based on a Bayesian fusion algorithm and reinforcement learning, wherein the method comprises the steps of acquiring real-time physiological data of the patients; the real-time physiological data at least comprises heart rate, brain electrical signals, skin electrical response and body temperature, the real-time physiological data of a patient is input into a circadian rhythm regulation model, a corresponding intervention scheme is obtained, physiological data at the next moment is fed back, the circadian rhythm regulation model is updated, and the circadian rhythm regulation model is constructed based on a Bayesian fusion algorithm and a reinforcement learning model. The application realizes dynamic real-time intervention on the circadian rhythm of the patient with consciousness disturbance, improves the self-adaptive capacity of circadian rhythm regulation, and further improves the arousal efficiency of the patient with consciousness disturbance.

Inventors

LIN FA
LI RUNTING
CHEN XIAOLIN
YANG JUN
TANG YIQI
Pan Xiaozhuo
Luan Mengyin
ZHAO JIZONG

Assignees

首都医科大学附属北京天坛医院
上海君依悦远健康科技有限公司

Dates

Publication Date: 20260505
Application Date: 20250324

Claims (10)

1. The method for adjusting the circadian rhythm of the conscious disturbance patient based on the Bayesian fusion algorithm and the reinforcement learning is characterized by comprising the following steps of: acquiring real-time physiological data of a patient, wherein the real-time physiological data at least comprises heart rate, brain electrical signals, skin electrical response and body temperature; Inputting real-time physiological data of a patient into a circadian rhythm regulation model, acquiring a corresponding intervention scheme, feeding back physiological data of the next moment, and updating the circadian rhythm regulation model, wherein the circadian rhythm regulation model is constructed based on a Bayesian fusion algorithm and a reinforcement learning model; The training process of the circadian rhythm regulation model is as follows: based on the historical physiological data, carrying out consciousness state evaluation on the patient by using a Bayesian fusion algorithm to determine the historical consciousness state of the patient, wherein the types of the historical consciousness state at least comprise wakefulness, shallow sleep and deep sleep; The method comprises the steps of taking historical physiological data and historical consciousness states of a patient as a state space, taking an intervention scheme as an action space, taking consciousness state change of the intervention scheme adopted by the current consciousness states as a reward function, training a reinforcement learning model, and obtaining a trained reinforcement learning model, wherein the intervention scheme at least comprises illumination intervention, sound intervention and touch intervention.
2. The method for adjusting circadian rhythm of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning according to claim 1, wherein the method for evaluating conscious state of patient based on historical physiological data by using Bayesian fusion algorithm, and determining the historical conscious state of patient specifically comprises: Setting prior probability of the corresponding conscious state based on the historical physiological data; establishing likelihood functions according to the prior probability of each historical physiological data and the corresponding conscious state, Calculating the posterior probability of each consciousness state by using a Bayesian fusion algorithm according to the prior probability and likelihood function of the consciousness state, wherein the posterior probability of each consciousness state is the product of likelihood functions of historical physiological data under each consciousness state; And selecting the consciousness state corresponding to the maximum probability value in the posterior probability of each consciousness state as the historical consciousness state of the patient.
3. The method for circadian rhythm adjustment for conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning according to claim 1, wherein the prior probability of conscious state is set by medical professional.
4. The method for adjusting circadian rhythms of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning according to claim 3, wherein the calculation formula of posterior probability of each conscious state is as follows: P(r,e,s,t|C k )＝P(r|C k )P(e|C k )P(s|C k )P(t|C k ); Where P (C k r, e, s, t) is the posterior probability that the patient is in the conscious state C k under the observation of the heart rate r, the brain electrical signal e, the skin electrical response s and the body temperature t, P (r, e, s, t|C k ) is the prior probability that the patient is in the conscious state C k , P (C k ) is the prior probability that the patient is in the conscious state, P (r, e, s, t) is the probability that the heart rate r, the brain electrical signal e, the skin electrical response s and the body temperature t coexist, P (r|C k ) is the prior probability that the patient is in the conscious state corresponding to the heart rate r, P (e|C k ) is the prior probability that the patient is in the conscious state corresponding to the brain electrical signal e, P (s|C k ) is the prior probability that the patient is in the conscious state corresponding to the skin electrical response s, and P (t|C k ) is the prior probability that the patient is in the conscious state corresponding to the body temperature t.
5. The method for adjusting circadian rhythms of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning according to claim 1, wherein the calculation formula of the reward function is as follows: Where Q (s t ,a t ) is the change in conscious state of action a t taken in state s t , Q (s t ,a t ) is the change in conscious state of action a t taken in the last time step state s t , α is learning rate, R (s t ,a t ) is the instant reward of action a t taken in state s t , γ is discount factor, max a′ Q(s t+1 , a ') is the best Q value obtained by action a' reaching the next state s t+1 .
6. A circadian rhythm regulation system for conscious disturbance patients based on a Bayesian fusion algorithm and reinforcement learning is characterized by comprising a conscious state detection module, a control module and an intervention module; the consciousness state detection module is connected with the control module and is used for collecting real-time physiological data of a patient, wherein the real-time physiological data at least comprise heart rate, brain electrical signals, skin electrical response and body temperature; The control module is connected with the intervention module and is used for obtaining an intervention scheme by applying the method of any one of claims 1-4 and controlling the intervention module to intervene on a patient according to the intervention scheme.
7. The system for regulating the circadian rhythm of a patient with conscious disturbance based on the Bayesian fusion algorithm and the reinforcement learning according to claim 6, wherein the system for regulating the circadian rhythm of a patient with conscious disturbance based on the Bayesian fusion algorithm and the reinforcement learning further comprises a power supply module; The power module is connected with the control module.
8. The system for regulating circadian rhythm of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning according to claim 6, wherein the intervention module specifically comprises a haptic intervention module, an auditory intervention module and an illumination intervention module; The tactile intervention module, the auditory intervention module and the illumination intervention module are all connected with the control module, the tactile intervention module is used for performing tactile intervention on a patient based on the intervention scheme, the auditory intervention module is used for performing auditory intervention on the patient based on the intervention scheme, and the illumination intervention module is used for performing illumination intervention on the patient based on the intervention scheme.
9. The system for regulating circadian rhythm of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning according to claim 8, wherein the haptic intervention module at least comprises any one of an electromagnetically driven vibration sheet or a current stimulation module, and the illumination intervention module at least comprises an LED lamp.
10. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the bayesian fusion algorithm and reinforcement learning based circadian rhythm regulation system for conscious disturbance patients according to any of claims 1-5.

Description

Method, system and equipment for adjusting circadian rhythm of conscious disturbance patient based on Bayesian fusion algorithm and reinforcement learning Technical Field The application relates to the field of medical equipment, in particular to a method, a system and equipment for adjusting circadian rhythms of conscious disturbance patients based on Bayesian fusion algorithm and reinforcement learning. Background Studies have shown that the normal sleep-wake cycle is critical for the recovery of consciousness in conscious impaired patients, as sleep plays a critical role in brain repair, functional recovery and nerve regeneration. Current sleep treatment methods, including pharmaceutical and non-pharmaceutical interventions, have potential clinical value for improving the quality of sleep and regulation of sleep rhythms in these patients, but have not received sufficient attention. Currently, the main circadian therapies at present are as follows: 1) Phototherapy, which simulates natural circadian variation by adjusting illumination intensity and duration, helps to regulate the patient's biological clock, has been widely used to treat patients with sleep disorders and circadian rhythm disorders, especially those who are bedridden for a long period of time and who cannot be exposed to natural light. 2) Pharmaceutical intervention regimens such as the use of melatonin and its analogues may assist in modulating the patient's biological clock, melatonin is commonly used to treat circadian rhythm disorders, particularly due to its critical role in regulating the sleep-wake cycle. 3) Sensor technology, which is often combined with artificial intelligence or algorithms to provide real-time feedback, has also been proposed in the prior art to use biosensors to monitor physiological signals such as patient body temperature, activity level, heart rate, etc. to infer their circadian rhythm state, thereby providing personalized rhythm control advice. 4) Circadian rhythm modulation in cognition impaired patients prior art schemes involve intervention in circadian rhythms in mild cognitive impairment or Alzheimer's disease patients. Studies have shown that biological clock disorders in such patients are associated with deterioration of their cognitive function, and thus existing regimens may be regulated in a variety of ways, such as phototherapy, environmental control, and intervention in sleep-wake behavior. These prior art techniques above all provide, to some extent, a circadian rhythm modulation scheme for conscious impaired patients. However, in the circadian rhythm adjustment process of the patient with consciousness disturbance, all physiological parameters and consciousness states of the patient are changed in real time, and the adjustment mode is too lag only through unidirectional adjustment performed by a preset program or algorithm, and cannot be matched with the physical change direction of the patient with consciousness disturbance, so that the arousal efficiency of the patient with consciousness disturbance is low. Disclosure of Invention The application aims to provide a method, a system and equipment for adjusting the circadian rhythm of a patient with consciousness disturbance based on a Bayesian fusion algorithm and reinforcement learning, which can realize dynamic real-time intervention on the circadian rhythm of the patient with consciousness disturbance, improve the self-adaptive capacity of circadian rhythm adjustment and further improve the awakening efficiency of the patient with consciousness disturbance. In order to achieve the above object, the present application provides the following solutions: The application provides a method for adjusting circadian rhythms of conscious disturbance patients based on a Bayesian fusion algorithm and reinforcement learning, which comprises the steps of obtaining real-time physiological data of a patient, wherein the real-time physiological data at least comprise heart rate, brain electrical signals, skin electrical response and body temperature, inputting the real-time physiological data of the patient into a circadian rhythm adjustment model, obtaining a corresponding intervention scheme, feeding back physiological data at the next moment, and updating the circadian rhythm adjustment model, the circadian rhythm adjustment model is constructed based on the Bayesian fusion algorithm and the reinforcement learning model, the training process of the circadian rhythm adjustment model is as follows, the history physiological data is used for evaluating the consciousness state of the patient by using the Bayesian fusion algorithm, the type of the history consciousness state at least comprises a clear consciousness state, a shallow sleep state and a deep sleep state, taking the intervention scheme of the patient as a state space, training the reinforcement learning model by taking the consciousness state change of the intervention scheme taken by the current st