METHOD AND SYSTEM FOR ASSESSING THE HEALTH OF THE HUMAN BODY BASED ON THE LARGE-VOLUME SLEEP DATA Russian patent published in 2021 - IPC G16H50/30 

FIELD: medicine.

SUBSTANCE: group of inventions relates to medicine, namely, to assessment of human health based on the large-volume sleep data. Proposed is a terminal apparatus containing a system for implementing the method, wherein the system comprises a sleep data receiving unit for obtaining various physiological data on the human body during sleep, wherein the data is supplied from a sensor installed on the electric bed or an intelligent bed containing an electrical control system; a sleep data storage unit containing a cloud server for storing all the sleep data collected by the sleep data receiving unit, wherein the cloud server is configured to interact with the database of the control system for receiving sleep data in real time; a unit for training using the data for pre-processing the sleep data and training using the data by means of an artificial intelligence learning module for obtaining the physiological assessment index; a unit for training a classifier model for training the classifier model on the sleep data obtained during training of the artificial intelligence learning model; and a report generating unit for generating a report on the analysis of the state of human health in accordance with the human physiological assessment index and the classifier model obtained based on the training using the data; wherein the unit for training using the data comprises: a pre-processor for data processing, intended to filter incomplete and incorrect sleep data stored on the cloud server and enter the correct data into the artificial intelligence learning model for training so that the artificial intelligence learning model could study the characteristics of diseases and conduct calculations in order to obtain the physiological assessment index; and an autoencoder for creating an autoencoder network and conducting iterative training using the data by calculating the losses of the network; wherein the correct data after pre-processing is divided into sleep data with the tag of the disease and sleep data without the tag of the disease, wherein the sleep data with the tag of the disease comes from the body of a person suffering from a disease of a known type while the sleep data without the tag of the disease comes from the body of a person with a condition corresponding to an unknown disease; wherein the process of entering the correct data into the artificial intelligence learning model for training includes: sending the sleep data without the tag of the disease after pre-processing to the autoencoder network to calculate and obtain the uncontrolled losses of the network, wherein the method for calculating the uncontrolled losses of the network includes: sending the sleep data without the tag of the disease after pre-processing to a deep autoencoder of the autoencoder network to obtain the root mean square difference between the input values and the output values and assign the difference to the uncontrolled losses of the network; sending the sleep data with the tag of the disease after pre-processing to an autoencoder network with the same parameters as the autoencoder network to obtain the output data of dimensional reduction of the intermediate layer and conduct the calculation required to obtain the controlled losses of the network, wherein the method for calculating the controlled losses of the network includes: dividing the output data of dimensional reduction of the intermediate layer into different categories in accordance with the tag of the disease, identifying the central point of each category, calculating the distance from the data of the same category to the central points of the same category and assigning the sum of the obtained distances to the controlled losses of the network; and adjusting the parameters of the network structure using the back propagation algorithm, conducting iterative training within the above two stages, and terminating the iteration after the obtained sum of the uncontrolled losses of the network and the controlled losses of the network stops changing or the iterations reach the maximum, wherein the obtained sleep data is then used as a physiological assessment index.

EFFECT: group of inventions provides an assessment of human health based on the large-volume sleep data.

7 cl, 4 dwg