Human health monitoring, predicting and regulating method based on digital twinning

Abstract

The invention relates to a human health monitoring, predicting and regulating method based on digital twinning, which comprises the following steps: acquiring body surface information of a human body, constructing a human body digital twinning three-dimensional model, and acquiring the body surface information through a 3D scanner or a 3D reconstruction method based on a 2D image / video. And constructing a human body digital twinborn health monitoring model, constructing the human body surface information-based digital twinborn health monitoring model by using a machine learning method and taking the human body surface information as input and the internal function phenotype as output, and performing visual display. Building a human body digital twinborn health prediction and regulation model, exploring the influence of different regulation modes, and building an exterior phenotype characterization model and a health prediction model based on an interior phenotype; more accurate multi-dimensional health information is provided, the external phenotype and the internal function phenotype are combined, internal function evaluation based on the external phenotype is achieved, and finally the system integrating human body health monitoring, prediction and regulation is achieved.

Description

Technical Field

[0001] The present invention relates to a human health monitoring, prediction and regulation method based on digital twins. Background Art

[0002] Digital twin refers to the construction of a virtual model of a physical entity through digital technology, which enables all-round, dynamic tracking and simulation prediction of the entity. It can collect data through physiological sensors, videos, etc., and through modeling and simulation, build a digital model that is highly similar to the physical entity in the virtual space, and finally realize the monitoring, analysis and optimization of the physical entity[1]. At present, digital twins have certain applications in various industries, such as urban construction[2], intelligent manufacturing[3] and aerospace[4]; however, digital twins are still in the early stages of exploration in the field of health monitoring[5]. At present, the exploration of using digital twins in the field of health monitoring mainly includes: 1. Building patient-related digital twins, such as using wearable devices to monitor patients' vital signs data, and mapping and analyzing them in virtual models to achieve personalized health monitoring; 2. Building digital twin hospitals, such as creating a digital twin of a hospital, integrating registration, appointment, consultation, diagnosis, treatment and payment processes, and planning medical resources, so as to provide convenient and efficient medical services[6].

[0003] At present, the research related to the construction of patient digital twins mainly creates personalized virtual models based on the actual data of patients; most studies focus on the three-dimensional digital display of digital twins, which can realize the basic interactive functions of digital twins such as translation, rotation and scaling. However, related research still faces many challenges: 1. Many studies are still at the stage of model establishment and digital twin display, and the accuracy and comprehensiveness of digital twin information still need to be improved; 2. Existing digital twins often lack sufficient dynamic update capabilities and cannot reflect the dynamic health status changes of patients. 3. Existing digital twins often only consider the display of external information, and do not involve the representation of internal functions and the association between internal functional phenotypes and external phenotypes.

[0004] References: [1]RASHMI PANT JOSHI, SURBHI GULATI, ARPAN KUMAR KAR. Digital Twin for Industrial Applications - A Literature Review[C] / / Transfer, Diffusion andAdoption of Next-Generation Digital Technologies, Part 3: IFIP WG 8.6International Working Conference on Transfer and Diffusion of IT (TDIT 2023),December 15-16, 2023, Nagpur, India. 2024:92-103. [2] Zhang Ye, Cui Yu, Qu Cuicui, et al. Digital Twin Construction[J]. New Architecture, 2024(1):28-33. DOI:10.12069 / j.na.202401028. [3] Qi Tianhong, Huang Haifeng, Cao Le, et al. Research on conceptual model and key technologies of digital twins for intelligent manufacturing[J]. Information Technology and Standardization, 2024(11):45-50,60. DOI:10.3969 / j.issn.1671-539X.2024.11.019. [4] Su Bin, Jiang Zhao. A brief analysis of the application of digital twin technology in the aerospace field[J]. Modern Industrial Economy and Informatization, 2022, 12(9): 132-134. DOI: 10.16525 / j.cnki.14-1362 / n.2022.09.053. [5] HONGYU QIN, CHUANG WANG, YING LI, et al. Status Quo and Future Development of Digital Twins in Medical and Health Fields[C] / / Proceedings of 2023 International Conference on Display Technology. 2023:1-3. [6]EVANGELIA KATSOULAKIS, QI WANG, HUANMEI WU, et al. Digital twins for health: a scoping review[J]. npj Digital Medicine, 2024,7(1). DOI:10.1038 / s41746-024-01073-0. Summary of the invention

[0005] The purpose of the present invention is to provide an accurate human health monitoring, prediction and regulation method based on digital twins. The method of the present invention combines external phenotypes with intrinsic functional phenotypes to achieve human health monitoring, prediction and regulation based on digital twins. The method of the present invention obtains the body surface information and related phenotypes of the human body, and uses machine learning methods to construct a digital twin health monitoring model with the body surface information of the human body as input and the intrinsic functional phenotype as output; in addition, by exploring the effects of different regulation methods on the intrinsic phenotype (such as blood lipids, blood pressure, metabolism, etc.), and constructing a health prediction model with the intrinsic phenotype as input and the external phenotype as output; through this method, not only can the external phenotype and the intrinsic functional phenotype be associated, but also the intrinsic function and its changes can be intuitively displayed through the external phenotype, and finally the selection of personalized health regulation paradigm can be realized.

[0006] The present invention proposes a human health monitoring, prediction and regulation method based on digital twins, including data processing, building a health monitoring model of human digital twins and calculating intrinsic functional phenotypes. The specific steps are as follows: (1) Data processing Collecting body surface information data of a human body, wherein the body surface information data can be collected by a 3D scanner or obtained by a 3D reconstruction method after obtaining a 2D image / video; using a deep network or an end-to-end regression network or other methods to obtain data feature point locations, generate 3D point cloud data, and obtain body surface phenotypes based on the 3D point cloud data; wherein: the body surface information data of a human body includes one or more of head height, neck height, neck circumference, hip height, hip circumference, abdomen height, and abdomen circumference; (2) Constructing a health monitoring model of the human digital twin Taking the body surface information of the human body in step (1) as input and the intrinsic functional phenotype as output, a digital twin health monitoring model based on the body surface information of the human body is constructed by using a machine learning method. The intrinsic functional phenotype adopts a health monitoring module, and the health monitoring module includes lung function, heart function, metabolism-related functions and other individual health-related functions / information; (3) Calculation of intrinsic functional phenotype By exploring the impact of different regulatory methods on intrinsic functional phenotypes, and constructing an external phenotype representation model and a health prediction model based on the internal phenotype; based on this, the impact of digital twin internal phenotype disturbances on external phenotype representation and health can be analyzed, thereby guiding the selection of personalized health regulation paradigms; specifically: taking the human body surface phenotype data as the dependent variable, and the internal phenotype as the independent variable, a machine learning model is constructed to achieve external phenotype representation and health prediction based on the internal phenotype.

[0007] In the present invention, the lung function monitoring described in step (3) uses human lung function indicators such as lung capacity, maximum vital capacity, forced vital capacity, deep inspiration volume, forced expiratory volume in one second, forced expiratory volume in 30 seconds, tidal volume, residual volume, functional residual capacity, lung ventilation, etc. as dependent variables, and human body surface phenotypic data such as chest circumference, abdominal circumference, neck circumference, left and right arm circumference, lower limb circumference, waist-to-hip ratio, etc. as independent variables to construct a machine learning model, and ultimately realize the evaluation of lung function indicators based on human body surface information; in addition, the changes in the chest and abdomen of the user captured by the camera are used to extract and analyze the features of the video data to realize the estimation of the dynamic respiratory rate; wherein: the machine learning method in constructing the machine learning model includes but is not limited to any one of a multivariate linear regression model, a decision tree regression model or a random forest regression model.

[0008] In the present invention, the heart function described in step (3) utilizes the real-time video of the user's face or a certain part of the body (such as the wrist) captured by the camera to extract and analyze the video data, mining the features including the slight color changes of the facial skin and the blood pulse, and using the deep learning algorithm to realize the estimation of dynamic heart rate, as well as the calculation of the average heart rate, the maximum heart rate, the heart rate trend, etc.

[0009] In the present invention, the metabolism-related functions described in step (3) take the human resting metabolic rate as the dependent variable and the human body surface phenotype data as the independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdomen height, abdomen circumference, etc., to construct a machine learning model, wherein: the machine learning method in constructing the machine learning model includes but is not limited to any one of a multivariate linear regression model, a Lasso regression model, a ridge regression model, a decision tree regression model or a random forest regression model.

[0010] In the present invention, the other individual health-related functions / information described in step (3) include obesity-related information, biological age-related information, and gender-related information. ① Obesity-related information: taking body composition indicators related to fat content / muscle content related to obesity as dependent variables, and taking human body surface phenotype data as independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height, abdominal circumference, etc., to construct a machine learning model (including but not limited to ridge regression model, decision tree regression model, random forest regression model), and realize multi-dimensional evaluation based on human body surface phenotype data, such as total body fat mass, total body muscle mass, fat mass index, and ② Information related to biological age, taking actual age as the dependent variable and human body surface phenotypic data as independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height, abdominal circumference, etc., to build a machine learning model (including but not limited to multivariate linear regression model, Lasso regression model, ridge regression model, etc.) to realize age assessment based on human body surface phenotypic data; ③ Information related to gender, taking gender as the dependent variable and human body surface phenotypic data as the independent variable, to build a machine learning model (including but not limited to support vector machine, etc.) to realize gender classification based on human body surface phenotypic data.

[0011] In summary, compared with the prior art, the characteristics and beneficial effects of the present invention are: The present invention can provide real-time dynamic display of multi-dimensional and accurate health information (such as lung function, metabolism, obesity, age, gender, heart rate, etc.).

[0012] The present invention can not only realize the intrinsic function evaluation based on the external phenotype, but also realize the external phenotype characterization and health prediction based on the internal phenotype transformation, truly combining the internal function and external phenotype of the digital twin to guide the selection of personalized health regulation paradigm.

[0013] Starting from the human body's surface information, the present invention extracts surface phenotype data and combines the surface phenotype with the intrinsic function. It can not only realize the dynamic display of changes in the external information of the digital twin, but also provide an intrinsic function evaluation of the digital twin, giving the digital twin "life"; it can also realize external phenotype representation and health prediction based on internal phenotype transformation, and assist in the prediction and regulation of human health. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of human health monitoring, prediction and regulation based on digital twins.

[0015] Figure 2 Functional information related to digital twins. DETAILED DESCRIPTION

[0016] The present invention is further described below by way of embodiments in conjunction with the accompanying drawings.

[0017] Example 1: Step 1: Data collection Data type: Human body surface information data mainly includes head height, neck height, neck circumference, hip height, hip circumference, abdomen height, abdomen circumference, etc.

[0018] Data collection method: It can be acquired through a 3D scanner or by using a 3D reconstruction method after acquiring 2D images / videos.

[0019] Data feature point positioning and human body surface phenotype extraction: For example, using deep networks, end-to-end regression networks, etc. to generate 3D point cloud data, and obtain body surface phenotypes based on point cloud data.

[0020] Step 2: Build a health monitoring model for human digital twins (1) Using the human body surface phenotype as input and the intrinsic functional phenotype as output, a digital twin health monitoring model based on the human body surface information is constructed. The machine learning methods include but are not limited to multivariate linear regression model, random forest model, support vector machine model, etc.

[0021] (2) Health monitoring modules (including but not limited to the following modules) a) Pulmonary function - using human lung function indicators such as lung capacity, maximum vital capacity, forced vital capacity, deep inspiration volume, forced expiratory volume in one second, forced expiratory volume 30 times in one second, tidal volume, residual volume, functional residual capacity, lung ventilation, etc. as dependent variables, and human body surface phenotypic data such as chest circumference, abdominal circumference, neck circumference, left and right arm circumference, lower limb circumference, waist-to-hip ratio, etc. as independent variables, to build a machine learning model (machine learning methods include but are not limited to multivariate linear regression model, decision tree regression model, random forest regression model), and finally realize the evaluation of lung function indicators based on human body surface information; in addition, using the changes in the user's chest and abdomen captured by the camera, feature extraction and analysis of video data are performed to realize the estimation of dynamic respiratory rate.

[0022] b) Heart function - Using real-time video of the user's face or a part of the body (such as the wrist) captured by the camera, feature extraction and analysis are performed on the video data, including subtle color changes of the facial skin, blood pulses and other features. Deep learning algorithms are used to estimate dynamic heart rate, as well as calculate average heart rate, maximum heart rate, heart rate trend, etc.

[0023] c) Metabolic related functions - taking the human resting metabolic rate as the dependent variable and the human body surface phenotype data as the independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height, abdominal circumference, etc., to build a machine learning model (including but not limited to multiple linear regression model, Lasso regression model, ridge regression model, decision tree regression model, random forest regression model).

[0024] d) Other individual health-related functions / information - ① Obesity-related, taking body composition indicators related to obesity-related fat content / muscle content as dependent variables, and taking human body surface phenotype data as independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height, abdominal circumference, etc., to build machine learning models (including but not limited to ridge regression model, decision tree regression model, random forest regression model) to achieve multi-dimensional evaluation based on human body surface phenotype data, such as total body fat mass, total body muscle mass, fat mass index, Android to Gynoid ratio, visceral fat ② Biological age-related, taking actual age as the dependent variable and human body surface phenotypic data as the independent variable, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height, abdominal circumference, etc., to build a machine learning model (including but not limited to multivariate linear regression model, Lasso regression model, ridge regression model, etc.) to achieve age assessment based on human body surface phenotypic data; ③ Gender-related, taking gender as the dependent variable and human body surface phenotypic data as the independent variable, to build a machine learning model (including but not limited to support vector machine, etc.) to achieve gender classification based on human body surface phenotypic data.

[0025] Step 3: Build a health prediction and regulation module for human digital twins This module mainly explores the impact of different regulation methods (such as different weight loss methods - exercise, diet, medication, etc.) on internal phenotypes (such as blood lipids, blood pressure, metabolism, etc.), and constructs an external phenotype representation model and health prediction model based on internal phenotypes; based on this, it can analyze the impact of digital twin internal phenotype disturbances on external phenotype representation and health, thereby guiding the selection of personalized health regulation paradigms. Specifically, it includes: using the body's surface phenotype data as the dependent variable, and the internal phenotype (such as blood lipids, blood pressure, metabolism, etc.) as the independent variable, constructing a machine learning model (including but not limited to random forests, etc.), and realizing external phenotype representation and health prediction based on internal phenotypes.

Claims

1. A human health monitoring, prediction and regulation method based on digital twins, including data processing, building a health monitoring model of human digital twins and intrinsic functional phenotype calculation, characterized in that The specific steps are as follows: (1) Data processing Collecting body surface information data of a human body, wherein the body surface information data is collected by a 3D scanner or obtained by a 3D reconstruction method after obtaining a 2D image / video; using a deep network or an end-to-end regression network method to obtain data feature point positioning, generate 3D point cloud data, and obtain body surface phenotypes based on the 3D point cloud data; wherein: the body surface information data of a human body includes one or more of head height, neck height, neck circumference, hip height, hip circumference, abdomen height or abdomen circumference; (2) Constructing a health monitoring model of the human digital twin Taking the body surface information of the human body in step (1) as input and the intrinsic functional phenotype as output, a digital twin health monitoring model based on the body surface information of the human body is constructed by using a machine learning method. The intrinsic functional phenotype adopts a health monitoring module, and the health monitoring module includes lung function, heart function, metabolism-related functions and other individual health-related functions / information; (3) Calculation of intrinsic functional phenotype By exploring the impact of different regulatory methods on intrinsic functional phenotypes, and constructing an external phenotype representation model and a health prediction model based on the internal phenotype; based on this, the impact of digital twin internal phenotype disturbances on external phenotype representation and health can be analyzed, thereby guiding the selection of personalized health regulation paradigms; specifically: taking the human body surface phenotype data as the dependent variable, and the internal phenotype as the independent variable, a machine learning model is constructed to achieve external phenotype representation and health prediction based on the internal phenotype.

2. The method according to claim 1, characterized in that The pulmonary function monitoring described in step (3) uses human lung function indicators, specifically lung capacity, maximum vital capacity, forced vital capacity, deep inspiration volume, forced expiratory volume in one second, forced expiratory volume 30 times in one second, tidal volume, residual volume, functional residual capacity and lung ventilation as dependent variables, and human body surface phenotypic data, specifically chest circumference, abdominal circumference, neck circumference, left and right arm circumference, lower limb circumference, waist-to-hip ratio as independent variables, to construct a machine learning model, and ultimately achieve lung function indicator evaluation based on human body surface information; in addition, using the changes in the user's chest and abdomen captured by the camera, feature extraction and analysis of the video data are performed to achieve dynamic respiratory rate estimation; wherein: the machine learning method in constructing the machine learning model is any one of a multivariate linear regression model, a decision tree regression model or a random forest regression model.

3. The method according to claim 1, characterized in that The heart function described in step (3) utilizes the real-time video of the user's face or a certain part of the body captured by the camera to extract and analyze the features of the video data, including the slight color changes of the facial skin and the blood pulse characteristics, and utilizes the deep learning algorithm to realize the estimation of dynamic heart rate, as well as the calculation of average heart rate, maximum heart rate, and heart rate trend.

4. The method according to claim 1, characterized in that The metabolism-related functions described in step (3) take the resting metabolic rate of the human body as the dependent variable and the surface phenotype data of the human body as the independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height and abdominal circumference, to construct a machine learning model, wherein: the machine learning method in constructing the machine learning model includes but is not limited to any one of a multivariate linear regression model, a Lasso regression model, a ridge regression model, a decision tree regression model or a random forest regression model.

5. The method according to claim 1, characterized in that Other individual health-related functions / information described in step (3), including obesity-related information, biological age-related information, and gender-related information, ① Obesity-related information, taking body composition indicators related to fat content / muscle content related to obesity as dependent variables, and taking human body surface phenotype data as independent variables, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height and abdominal circumference, to construct a machine learning model, wherein the learning method includes but is not limited to a ridge regression model, a decision tree regression model, and a random forest regression model, to achieve a multi-dimensional evaluation based on human body surface phenotype data, specifically: total body fat mass, total body muscle mass, fat mass index, Android to Gynoid ratio and visceral fat mass; ② Biological age-related information, taking actual age as the dependent variable and body surface phenotype data of the human body as the independent variable, including head height, neck height, neck circumference, hip height, hip circumference, abdominal height, and abdominal circumference, to construct a machine learning model, wherein the learning method is any one of a multiple linear regression model, a Lasso regression model, or a ridge regression model, to achieve age assessment based on body surface phenotype data of the human body; ③ Gender-related information, taking gender as the dependent variable and the body surface phenotype data of the human body as the independent variable, constructing a machine learning model, wherein the learning method adopts a support vector machine to realize gender classification based on the body surface phenotype data of the human body.

Images