CN115274140A

CN115274140A - Diabetes digital health management system based on big data and artificial intelligence

Info

Publication number: CN115274140A
Application number: CN202210977933.0A
Authority: CN
Inventors: 李益非
Original assignee: Individual
Current assignee: Individual
Priority date: 2022-05-11
Filing date: 2022-05-11
Publication date: 2022-11-01
Also published as: CN114732402A; CN114732402B

Abstract

The invention discloses a digital health management system for diabetes based on big data and artificial intelligence, which comprises a sensor module, a blood sugar data preprocessing module, a blood sugar control module, a continuous blood sugar monitoring module and an alarm module, wherein the blood sugar data of a patient is obtained through the sensor module, the blood sugar data preprocessing module realizes cleaning and screening of abnormal data, the blood sugar control module calculates insulin to be infused by the patient through an algorithm and controls an insulin pump to infuse the insulin into the body of the patient, the continuous blood sugar monitoring module realizes real-time monitoring of the blood sugar data of the patient, and when the insulin pump begins to infuse or the amount of the insulin is insufficient, the alarm module can give an alarm to start working or work abnormally of the system, and the digital health management system has the advantages that: the system can monitor a patient non-invasively and continuously, help the patient to calculate and adjust the amount of the insulin injection accurately, and has the characteristics of wide coverage, low cost, high retention and high activity.

Description

Diabetes digital health management system based on big data and artificial intelligence

Technical Field

The invention relates to the field of medical health management, in particular to a digital health management system for diabetes based on big data and artificial intelligence.

Background

In the present society where economy develops rapidly, health problems have become an important issue for human beings. Diabetes is one of the challenges of human health, is a metabolic disease characterized by hyperglycemia, has a main clinical manifestation of hyperglycemia, and has no radical cure method at present. Diabetes belongs to chronic diseases. Chronic disease management refers to medical behaviors and processes of regular detection, continuous monitoring, assessment and comprehensive intervention management of chronic non-infectious diseases and risk factors thereof, and the main connotations comprise chronic disease early screening, chronic disease risk prediction, early warning and comprehensive intervention, comprehensive management of chronic disease crowds, chronic disease management effect assessment and the like. In fact, chronic disease management is the management of chronic patients and high risk groups, including management and intervention on aspects such as reasonable diet, behavior habits, health psychology and the like; and (5) propagating correct management concepts, knowledge and skills of the chronic diseases, and performing comprehensive prevention and treatment work of the chronic diseases.

The main clinical manifestation of diabetes, a typical chronic disease, is hyperglycemia, and no radical treatment method exists at present. In daily life, unhealthy living habits, family medical history, specific diseases and other factors are easy to induce diabetes. The prevention and treatment of diabetes is a major and remote task. The large population of diabetics requires extensive daily monitoring and ongoing treatment, which represents a heavy burden on governments, society and families. In addition, there is a large population of potential diabetic individuals who are at a high risk of developing diabetes, yet are not identified and intervened in a timely manner. Second, in recent years, the prevalence of diabetes has increased, and the prevalence trend is toward younger. However, the medical system is not sound, the diabetes treatment level is limited, the nursing conditions need to be improved urgently, and the medical resource distribution is seriously unbalanced, which brings serious burden to the corresponding diagnosis and treatment work. These problems have all severely limited the advancement of the diabetes medical health care industry. Finally, the residents have poor understanding of diabetes and low health awareness. Due to increasingly tense pace of life, continuously increased working pressure and over-lack of daily movement, the risk of people suffering from diabetes is greatly increased; factors such as environmental deterioration and poor dietary habits also create hidden troubles for the frequent occurrence of diabetes. People lack basic understanding and cognition of the hazards of diabetes, and the health literacy of most citizens needs to be improved.

Currently, the main assessment means of diabetes can be roughly divided into the following two categories: traditional diagnostics and autonomic health assessment. The traditional diabetes diagnosis mainly depends on clinical examination in hospitals, and related main examination indexes comprise items such as blood sugar, insulin, c-peptide level, glycosylated hemoglobin, blood fat, blood pressure and the like. The mode completely depends on the experience of doctors, and the result is accurate and reliable; but the cost is high, the autonomous participation of patients is lacked, only short-term disease information is reflected, and the flexibility is insufficient. The other mode is diabetes autonomous monitoring, which mainly depends on the participation and the leading of a patient, utilizes convenient monitoring equipment to perform autonomous evaluation of the condition of the patient anytime and anywhere, and provides support by a corresponding monitoring system in the background. The mode provides auxiliary diagnosis service for patients and families by mining and analyzing the existing disease information, has the characteristics of flexibility, convenience and simplicity in operation, and is suitable for daily health assessment and risk prediction. Although self-health management refers to a mode of evaluating the health condition and the disease risk of residents based on individuals, the self-health management analyzes and predicts the health condition and the disease risk of the residents and then takes the whole process of preventive measures, and the self-health management is a novel personal health monitoring mode. This increasingly popular form of health care is a product of human health needs and era development. The conventional hospital information system mainly provides convenience in medical treatment procedures, patient management, hospital business and the like, but is difficult to play a role in services such as auxiliary medical treatment, health consultation and the like. With the improvement of health consciousness of residents, people no longer meet the current situation of traditional medical treatment, but have higher requirements on disease diagnosis and treatment and health care. Ordinary residents, especially diabetics, want activities such as daily physical health care and disease diagnosis not to be limited to the scope of medical institutions, the health information of the patient can be grasped at any time, and the health decision can be made in time. This provides a direction for the development of a health assessment. By analyzing and mining the medical information, potential knowledge and rules can be searched, and auxiliary suggestions are provided for risk identification, disease diagnosis, medical treatment and the like, so that medical workers, patients and the like can be better guided to prevent and treat diseases, and the self-health management requirement is further met.

Disclosure of Invention

A digital health management system for diabetes based on big data and artificial intelligence comprises a sensor module, a blood sugar data preprocessing module, a blood sugar control module, a continuous blood sugar monitoring module and an alarm module, wherein the sensor module comprises a transmitting unit, a sensor unit and a receiving unit, the transmitting unit is used for transmitting a signal for acquiring data to the sensor unit, the sensor unit is used for monitoring the concentration of glucose in interstitial fluid through a blood sugar sensor embedded under the skin, the glucose in the interstitial fluid is contacted with the sensor and generates an oxidation reduction reaction, a chemical signal is converted into an electric signal and finally transmitted to the receiving unit, the glucose concentration data is transmitted to the blood sugar data preprocessing module, cleaning and screening of abnormal data are achieved, the blood sugar control module comprises a blood sugar concentration calculating unit and a blood sugar control unit, the blood sugar concentration calculating unit is used for transmitting the glucose concentration in the interstitial fluid monitored by the blood sugar sensor embedded under the skin into a blood sugar concentration by an algorithm, the blood sugar control unit is used for calculating insulin needed to be infused by a patient and controlling the insulin pump to infuse insulin into the patient, the insulin monitoring module, the continuous blood sugar data preprocessing module, the blood sugar control module is used for monitoring module and monitoring the insulin pump to give a real-time warning sign, the alarm to alarm and prompt that the patient that the insulin begins to infuse the insulin when the insulin.

Furthermore, the sensor module comprises a transmitting unit, a sensor unit and a receiving unit, wherein the transmitting unit transmits a signal for acquiring data to the sensor unit, the sensor unit monitors the glucose concentration in interstitial fluid through a subcutaneous blood glucose sensor, the glucose in the interstitial fluid is in contact with the sensor and generates an oxidation-reduction reaction, a chemical signal is converted into an electric signal, and the electric signal is finally transmitted to the receiving unit.

Further, the glucose concentration data are transmitted to the blood glucose data preprocessing module to realize cleaning and screening of abnormal data, the abnormal data comprise default and low-confidence data, the reasons for generating the abnormal data comprise insensitivity of a sensor probe and a motion state of a patient, if the abnormal data are generated due to the insensitivity of the sensor probe, the data can be continuously default and cannot be transmitted to a next module, so that the system cannot work normally, if the abnormal data are generated due to the motion state of the patient, the data can have a default or low-confidence value state within a certain time period, the blood glucose data preprocessing module can filter all the abnormal data, and when the patient guarantees a calm state, the system can restore to a normal working state.

Furthermore, the blood glucose data preprocessing module needs to update the blood glucose data transmitted by the sensor module at regular time, and comprises a blood glucose data updating unit, the blood glucose data in the sensor module needs to be called to train and test the support vector machine, when the blood glucose data updating unit trains the support vector machine through the blood glucose data called by the sensor module, the punishment factor and the kernel function parameter of the support vector machine are determined by adopting a firefly algorithm, in the process of optimizing the punishment factor and the kernel function parameter of the support vector machine by adopting the firefly algorithm, each firefly adopts a roulette rule to select to move towards an individual with higher fluorescence brightness than the firefly, the moving distance of each firefly is determined according to the attraction, on the basis, the firefly i is set to select to move towards the firefly j, and the (t + 1) th iterative update is finally realized, and the specific position updating formula is as follows:

in the above formula, x _i (t + 1) denotes the position of firefly i after the (t + 1) th iteration update, x _i (t) indicates the position of firefly i after the updating of the t-th iteration, x _j (t) represents the position of firefly j after the updating of the t-th iteration, β _ij (t) denotes the attraction of firefly j to firefly i after the t-th iteration update, α _ij (t) represents the random term coefficient of the firefly i moving randomly towards the firefly j after the t-th iteration updating, rand is a random coefficient obeying normal distribution, and rand belongs to [0,1 ∈ ]]。

Further, the attraction degree β of firefly j to firefly i after the t iteration is updated _ij The value of (t) is set to:

in the above-mentioned formula, the compound of formula,

represents the original attraction of firefly j to firefly i after the t-th iterative update, and

the values of (A) are:

wherein, beta ₀ Where r =0, the attraction degree of firefly, that is, the maximum attraction degree, γ is a light absorption coefficient, and represents a characteristic that firefly gradually decreases with increasing distance, and r _ij (t) is the Cartesian distance, ρ, between firefly i and firefly j after the tth iterative update _ij (t) is the historical adjustment coefficient of the attraction of firefly j to firefly i after the tth iterative update, and ρ _ij The value of (t) is:

μ (t) represents an iterative correction coefficient, and

wherein, T _max Is the maximum number of iterations, k _ij (t) represents the statistical coefficient of the history of the region between firefly i and firefly j after the t-th iteration update, k _ij The value of (t) is:

wherein omega _ij (t) denotes the position x _i (t) is the center, with r _ij (t) is a spherical region of radius, x _j (t) is the location of firefly j after the τ th iteration update, f (x) _j (τ),Ω _ij (t)) is for position x _j (τ) and region Ω _ij (t) a region between (t) and

m represents the number of fireflies in the population, and t represents the current iteration number.

Further, the random term coefficient alpha of the firefly i moving to the firefly j randomly after the t-th iteration updating _ij The value of (t) is set to:

representing a set consisting of the fireflies serving as the moving directions selected when the (t + 1) th iteration update is carried out on each firefly in the population as K (t), and only keeping one of the repeated fireflies when the repeated fireflies serving as the moving directions exist in the set K (t), and defining theta _K (t) represents a moving direction attribute value of firefly in the set K (t), and θ _K The value of (t) is:

in the above formula, M _K (t) represents the number of fireflies in the set K (t), M represents the number of fireflies in the population, y represents _K (t) represents the dominance value of the spatial distribution of fireflies in the set K (t), and y _K The value of (t) is:

wherein, the fireflies in the set K (t) are sequenced according to the absolute fluorescence brightness value from high to low to form a sequence L _K (t), then x' _l (t) represents the sequence L _K Location, x 'of the first firefly in (t) updated at the t iteration' _l+1 (t) represents the sequence L _K (l + 1) th firefly in (t) at the position updated at the t-th iteration, k is a given positive integer, and k is<M _K (t)，x _l,a (t) is distance position x 'in the population after the t iteration update' _l (t) the position of the firefly at a-th place,

is a sequence L _K (t) a spatial distribution comparison function of the first firefly, and

when theta is measured _K The value of (t) satisfies: theta _K (t)>1, then the random term coefficient alpha _ij The value of (t) is:

when theta is _K The value of (t) satisfies: theta _K When (t) is less than or equal to 1, the random term coefficient alpha is _ij The value of (t) is:

in the above formula, α ₀ Is given an initial random coefficient value, and alpha ₀ ∈[0,1]，ω _j (t) representsThe local spatial coefficient, omega, of firefly j in the set K (t) after the t-th iteration update _j The value of (t) is:

x _j (t) indicates the position of firefly j after the t-th iteration update, delta _j (t) represents the global optimization coefficient, δ, of firefly j in the set K (t) after the t-th iteration update _j The value of (t) is:

wherein the content of the first and second substances,

shows firefly j in sequence L after the t-th iterative update _K (ii) the ordering in (t),

represents the sequence L _K (t) A

The location of the individual fireflies after the t-th iteration of updating.

Further, the blood glucose concentration calculation unit converts the glucose concentration in interstitial fluid monitored by a blood glucose sensor embedded under the skin into the blood glucose concentration by utilizing an algorithm, the blood glucose is the glucose concentration in blood, the currently common unit is mmol/L, the normal value of the fasting blood glucose is 3.9-6.1mmol/L, the fasting blood glucose is more than or equal to 7.0mmol/L, diabetes mellitus should be considered, the blood glucose is more than or equal to 11.1mmol/L (200 mg/dL) in two hours of random blood glucose, the diabetes mellitus should also be considered, the conversion relation between the glucose and the blood glucose is deduced by adopting a fitting algorithm, glucose samples in the interstitial fluid are monitored and collected by the blood glucose sensor, and the glucose samples are recorded as x _i I =1,2, …, n, assuming the corresponding blood glucose output is noted as y _i I =1,2, …, n, using least squares to define

Is provided with

Where k is the slope of the fitted curve, b is the offset, x _i To collect a sample of glucose in interstitial fluid, y _i For the purpose of the corresponding blood glucose output,

for the output of the blood glucose to be fitted,

for the slope of the curve to be fitted,

offset of the fit, using least squares, of

Wherein L is a loss function, k needs to be searched, b minimizes L, and partial derivatives of k and b can be solved respectively, including:

further, can be obtained

Evaluating the goodness of the fit using a sum of squares of population, wherein the sum of squares of population includes sum of squared errors and sum of squares of regression, defining SST = SSE + SSR, wherein SST represents the sum of squares of population, SSE represents the sum of squared errors, SSR represents the sum of squares of regression, assuming goodness of fit is R ² Definition of

R ² The closer to 1, the closer to 0 the sum of squared errors is, and the smaller the error, the more pseudo-stationaryThe better that is fit, for a total sum of squares SST, there is

For the sum of squared errors SSE, there are

For regression sum of squares SSR, there are

Wherein

Represents the average of the fit output.

Furthermore, the blood sugar control unit deduces the required infusion dosage of the insulin pump according to the blood sugar concentration fitted by the blood sugar concentration calculation unit, and defines a visible variable v = {0,1} by utilizing a machine learning boltzmann machine algorithm ^D Where D is the dimension of the visible variable, defining the hidden variable h =0,1 ^P Where P is the dimension of the hidden variable, let L = [ L ] _ij ] _D×D Wherein L represents a weight between the visible variables v, and J = [ J = _ij ] _P×P Wherein J represents the weight between hidden variables h, let W = [ W = _ij ] _D×P Wherein W represents a weight between a visible variable and a hidden variable, defining a joint probability density function

Wherein z represents a hidden variable, p (v, h) represents a probability density function for the joint between the visible variable and the hidden variable, U (v, h) is an energy function,

wherein v is ^T Is the transposition of v, h ^T For the transposition of h, let p (v) be the marginal probability distribution of p (v, h), have p (v) = ∑ _h p (v, h), let θ be the parameter set θ = { W, L, J }Assuming that the total number of samples is N and V is a set of samples of V, with | | | V | = N, then the log-maximum likelihood estimate is:

in order to find the distribution parameter θ corresponding to the maximum of the log-maximum likelihood estimation, a gradient is found for the distribution parameter θ:

wherein the content of the first and second substances,

for the W gradient direction, there are:

therefore, the temperature of the molten metal is controlled,

wherein

Representation based on p _data Distribution lower pair variable vh ^T In the hope of expectation,

representation based on p _model Distribution lower pair variable vh ^T Expect that p is satisfied _data ＝ p _data (v)p _model (h|v)，p _model ＝p _model (v, h), then the three parameter distribution increments can be expressed as:

wherein α is the search step length of boltzmann machine, and after t iterations, the new three parameter set matrices can be expressed as:

wherein, W ^(t) Representing the weight between the visible and hidden variables after t iterations, L ^(t) Representing the weight between visible variables after t iterations, J ^(t) Representing the weight between hidden variables after t iterations, W ^(t+1) Represents the weight between the visible and hidden variables after t +1 iterations, L ^(t+1) Representing the weight between the visible variables after t +1 iterations, J ^(t+1) Representing the weights between hidden variables after iteration t +1, the transition probabilities can be written as:

wherein v is _-i Is { v-v _i Set of h _-i Is { h-h _i The set of (c) is (c),

meaning that k takes from 1 to D does not include the accumulation of i,

denotes the accumulation of m from 1 to D excluding j, σ is the sigmoid function, p (v) _i ＝1|h,v _-i ) Denotes given h and v _-i When, v _i Probability of =1, p (h) _j ＝1|v,h _-i ) Denotes given v and h _-i When h is present _j Probability of =1, for glycemic control unit, if p (v) _i ＝1|h,v _-i ) The larger the time, the more insulin needs to be infused into the patient, if p (h) _j ＝1|v,h _-i ) The larger the probability that insulin needs to be infused into the patient at that time.

Furthermore, the continuous blood sugar monitoring module monitors the work of the sensor module, the blood sugar data preprocessing module and the blood sugar control module in real time to ensure the vital signs of the patient, in the insulin treatment process, the non-intensive treatment patient needs to inject 2-3 times of medicaments every day, and the intensive treatment patient needs to inject 3-4 times of medicaments every day.

Further, alarm module includes volume key and megaphone, begin the infusion when the insulin pump, send the alarm sound that begins to infuse the insulin by alarm module, this alarm sound is the long ring sound, accessible volume key regulation alarm sound size, it is long for the insulin infusion for the alarm sound duration, suggestion patient's insulin pump is working, when the insulin volume is not enough, send the alarm sound that the insulin volume is not enough, this alarm sound is discontinuous short-time promotion, end alarm sound after supplementing the insulin, still can force the alarm sound of interrupt through double-click volume key.

The invention has the beneficial effects that: the system has the characteristics of intelligent equipment and intelligent decision, and mainly aims at solving the problem of how many medicines are to be taken by a patient, in fact, the time and the number of medicines are important, the patient forgets when and how much insulin is given in real life, and the system can help the family and the doctor of the patient to master the medication situation so as to remind and encourage the patient in time, or synthesize other life behavior data of the patient to adjust the management scheme, can self-adaptively complete the functions of signal transmission, data acquisition and the like through the sensor module, does not influence the normal life work of the patient in the working process, can realize the real-time monitoring for 24 hours, brings safety guarantee for the vital signs of the patient, a blood sugar data updating unit in the blood sugar data preprocessing module utilizes the firefly algorithm and is applied to the optimization of the parameters of the support vector machine, the inherent defects of the firefly algorithm are improved, so that the classification precision of a support vector machine can be effectively improved through parameters determined by the improved firefly algorithm, the blood sugar control module combines the fitting and the Boltzmann machine algorithm, the statistical probability is introduced in the state change of neurons, the equilibrium state of a network obeys Boltzmann distribution, a network operation mechanism is based on a simulated annealing algorithm, the Boltzmann machine combines the advantages of a multi-layer feedforward neural network and a discrete Hopfield network in the aspects of network structure, learning algorithm and dynamic operation mechanism, the method is established on the basis of the discrete Hopfield network, has learning capability, can seek the optimal solution through a simulated annealing process, a continuous blood sugar monitoring module can divide the disease risk according to the blood sugar concentration of a patient, and the risk level comprises low degree, medium degree, high degree and extremely high degree, the screening of the high risk group of diabetes can be realized, the alarm is given by the loudspeaker, and the working state of the system can be judged according to the alarm sound.

Drawings

The invention is further described with the aid of the accompanying drawings, in which, however, the embodiments do not constitute any limitation to the invention, and for a person skilled in the art, without inventive effort, further drawings may be derived from the following figures. FIG. 1 is a schematic diagram of the present invention.

Detailed Description

The invention is further described with reference to the following examples.

Referring to fig. 1, the present embodiment provides a digital health management system for diabetes based on big data and artificial intelligence, including a sensor module, a blood glucose data preprocessing module, a blood glucose control module, a continuous blood glucose monitoring module and an alarm module, where the sensor module includes a transmitting unit, a sensor unit and a receiving unit, where the transmitting unit transmits a signal for acquiring data to the sensor unit, the sensor unit monitors a glucose concentration in interstitial fluid through a subcutaneous blood glucose sensor, glucose in the interstitial fluid contacts with the sensor and undergoes an oxidation-reduction reaction, and converts a chemical signal into an electrical signal, and finally transmits the electrical signal to the receiving unit, and transmits the glucose concentration data to the blood glucose data preprocessing module, so as to clean and screen abnormal data, the blood glucose control module includes a blood glucose concentration calculating unit and a blood glucose control unit, the blood glucose concentration calculating unit converts the glucose concentration in the interstitial fluid monitored by the subcutaneous blood glucose sensor into a blood glucose concentration by using an algorithm, the blood glucose control unit calculates insulin that the patient needs to infuse insulin into the patient, and controls the insulin pump to infuse insulin into the patient, the continuous blood glucose data monitoring module, the blood glucose data preprocessing module reports a working state that the patient needs to infuse insulin, and reports a warning sound to alarm module to alarm when the patient starts to drink insulin.

Specifically, the sensor module comprises a transmitting unit, a sensor unit and a receiving unit, wherein the transmitting unit transmits a signal for acquiring data to the sensor unit, the sensor unit monitors the concentration of glucose in interstitial fluid through a subcutaneous blood glucose sensor, the glucose in the interstitial fluid is in contact with the sensor and generates an oxidation-reduction reaction, a chemical signal is converted into an electric signal, and the electric signal is finally transmitted to the receiving unit.

Specifically, glucose concentration data is transmitted to the blood glucose data preprocessing module, abnormal data are cleaned and screened, the abnormal data comprise default and low-confidence data, the reasons for generating the abnormal data comprise insensitive sensor probes and motion states of patients, if the abnormal data are generated due to the insensitive sensor probes, the data can be continuously default and cannot be transmitted to a next module, the system cannot work normally, if the abnormal data are generated due to the motion states of the patients, the data can be in default or low-confidence value states within a certain time period, the blood glucose data preprocessing module can filter all the abnormal data, and when the patients guarantee a calm state, the system can restore to the normal working state.

Specifically, the blood glucose data preprocessing module needs to update the blood glucose data transmitted by the sensor module at regular time, and includes a blood glucose data updating unit, the blood glucose data in the sensor module needs to be called to train and test the support vector machine, when the blood glucose data updating unit trains the support vector machine through the blood glucose data called by the sensor module, the penalty factor and the kernel function parameter of the support vector machine are determined by adopting a firefly algorithm, in the process of optimizing the penalty factor and the kernel function parameter of the support vector machine by adopting the firefly algorithm, each firefly adopts a roulette rule to select to move towards an individual with higher fluorescence brightness than the firefly, the moving distance of each firefly is determined according to the attraction force, on the basis, the firefly i is set to select to move towards the firefly j, and finally the (t + 1) th iterative update is realized, and the specific position updating formula is as follows:

in the above formula, x _i (t + 1) denotes the position of firefly i after the (t + 1) th iteration update, x _i (t) represents the position of firefly i after the updating of the t-th iteration, x _j (t) represents the position of firefly j after the updating of the t-th iteration, β _ij (t) denotes the attraction of firefly j to firefly i after the t-th iteration update, α _ij (t) represents the random term coefficient of the firefly i moving randomly towards the firefly j after the t-th iteration updating, rand is a random coefficient obeying normal distribution, and rand belongs to [0,1 ∈ ]]。

Specifically, the attraction degree β of firefly j to firefly i after the t-th iteration update _ij The value of (t) is set to:

in the above-mentioned formula, the compound of formula,

represents the original attraction of firefly j to firefly i after the t-th iteration update, and

the values of (A) are:

wherein beta is ₀ Where r =0, the attraction degree of firefly, that is, the maximum attraction degree, γ is a light absorption coefficient, and represents a characteristic that firefly gradually decreases with increasing distance, and r _ij (t) is the Cartesian distance, rho, between firefly i and firefly j after the tth iterative update _ij (t) is a historical adjustment coefficient of the attraction of firefly j to firefly i after the tth iteration update, and ρ is _ij The value of (t) is:

μ (t) represents an iterative correction coefficient, and

wherein omega _ij (t) denotes the position x _i (t) is the center, with r _ij (t) is a spherical region of radius, x _j (τ) is the location of firefly j after the τ -th iteration update, f (x) _j (τ),Ω _ij (t)) is for position x _j (τ) and region Ω _ij (t) a region between (t) and

Specifically, the random term coefficient α for the firefly i to move randomly toward the firefly j after the t-th iterative update _ij The value of (t) is set to:

selecting each firefly in the population when the (t + 1) th iteration update is carried outA set of fireflies in which the direction of movement is defined as K (t), and when there are fireflies in the set K (t) that are repeated in the direction of movement, only one of the repeated fireflies is retained, and θ is defined _K (t) represents a moving direction attribute value of fireflies in the set K (t), and θ _K The value of (t) is:

wherein the fireflies in the set K (t) are sequenced from high to low according to the absolute fluorescence brightness value to form a sequence L _K (t), then x' _l (t) represents the sequence L _K (t) location of the l firefly in the t iteration updated, x' _l+1 (t) represents the sequence L _K (t) the position of the (l + 1) th firefly in (t) after the updating of the t-th iteration, k is a given positive integer, and k is<M _K (t)，x _l,a (t) is distance position x 'in the population after the t iteration update' _l (t) the position of the firefly at a-th place,

when theta is _K The value of (t) satisfies: theta _K (t)>When 1, then the random term coefficient alpha _ij The value of (t) is:

in the above formula, α ₀ Is given an initial random coefficient value, and alpha ₀ ∈[0,1]，ω _j (t) represents the local spatial coefficients, ω, of firefly j in the set K (t) after the t-th iteration update _j The value of (t) is:

x _j (t) indicates the position of firefly j after the t-th iteration update, δ _j (t) represents the global optimization coefficient, δ, of firefly j in the set K (t) after the t-th iteration update _j The value of (t) is:

wherein the content of the first and second substances,

shows firefly j in sequence L after the t-th iteration update _K (ii) the ordering in (t),

represents the sequence L _K (t) A

The location of the individual fireflies after the t-th iteration of updating.

Specifically, the blood glucose concentration calculating unit is used for calculating the concentration of glucose in interstitial fluid monitored by a blood glucose sensor embedded under the skinThe blood glucose concentration is converted by an algorithm, the blood glucose is the concentration of glucose in blood, the currently common unit is mmol/L, the normal value of fasting blood glucose is 3.9-6.1mmol/L, the fasting blood glucose is more than or equal to 7.0mmol/L, diabetes is considered, the blood glucose of random blood glucose for two hours is more than or equal to 11.1mmol/L (200 mg/dL), the diabetes is also considered, the conversion relation between the glucose and the blood glucose is deduced by adopting a fitting algorithm, a glucose sample in interstitial fluid is monitored and collected by a blood glucose sensor and recorded as x _i I =1,2, …, n, assuming the corresponding blood glucose output is noted as y _i I =1,2, …, n, using least squares to define

Is provided with

Where k is the slope of the fitted curve, b is the offset, x _i For taking samples of glucose in interstitial fluid, y _i For the purpose of the corresponding blood glucose output,

for the fitted blood glucose output to be the result,

for the slope of the curve to be fitted,

offset of the fit, using least squares, of

Wherein L is a loss function, k needs to be searched, b minimizes L, partial derivatives can be respectively solved for k and b, and the method comprises the following steps:

the item transfer can be obtained as follows:

can obtain

R ² Closer to 1, the sum of the squared errors is indicated to be closer to 0, smaller errors indicate better fit, for the sum of the squared overall SST, there is

For the sum of squared errors SSE, there are

For regression sum of squares SSR, there are

Wherein

Represents the average of the fit output.

Specifically, the blood glucose control unit infers the required infusion dosage of the insulin pump according to the blood glucose concentration fitted by the blood glucose concentration calculation unit, and defines a visible variable v = {0,1} by using a machine learning boltzmann machine algorithm ^D Where D is the dimension of the visible variable, defining the hidden variable h =0,1 ^P Wherein P is hiddenDimension of hidden variable, let L = [ ] _ij ] _D×D Wherein L represents a weight between the visible variables v, and J = [ J = _ij ] _P×P Wherein J represents the weight between hidden variables h, let W = [ W = _ij ] _D×P Wherein W represents the weight between the visible variable and the hidden variable, defining a joint probability density function

wherein v is ^T Is the transposition of v, h ^T For the transposition of h, let p (v) be the marginal probability distribution of p (v, h), there is p (v) = ∑ _h p (V, h), let θ be the parameter set θ = { W, L, J }, assuming the total number of samples is N, and V is the sample set of V, with | | | V | = N, then the log-maximum likelihood estimate is:

wherein the content of the first and second substances,

for the W gradient direction, there are:

therefore, the temperature of the molten metal is controlled,

wherein

wherein v is _-i Is { v-v _i Set of h _-i Is { h-h _i The set of (c) is (c),

meaning that k takes from 1 to D does not include the accumulation of i,

denotes the accumulation of m from 1 to D excluding j, σ is the sigmoid function, p (v) _i ＝1|h,v _-i ) Denotes given h and v _-i When, v _i Probability of =1, p (h) _j ＝1|v,h _-i ) Denotes given v and h _-i When h is present _j Probability of =1, for glycemic control unit, if p (v) _i ＝1|h,v _-i ) The larger the time, the more insulin needs to be infused into the patient, if p (h) _j ＝1|v,h _-i ) The larger the probability that insulin needs to be infused into the patient at that moment.

Specifically, the continuous blood glucose monitoring module supervises the work of the sensor module, the blood glucose data preprocessing module and the blood glucose control module in real time to ensure the vital signs of the patient, in the insulin treatment process, the non-intensive treatment patient needs to inject 2-3 times a day, and the intensive treatment patient needs to inject 3-4 times a day.

Specifically, the continuous blood glucose monitoring module divides the risk of the disease according to the blood glucose concentration of the patient, the risk levels comprise low degree, moderate degree, high degree and extremely high degree, and screening of high risk groups of diabetes can be realized.

More specifically, when the fasting blood glucose value is less than 5.0mmol/L and the blood triglyceride value is less than 2.3mmol/L, the risk is determined to be low; when the fasting blood glucose value is between 5.6-6.1mmol/L and the triglyceride value in blood is less than 2.3mmol/L, the intermediate risk can be judged; when the fasting blood glucose value is between 6.1-7.0mmol/L and the triglyceride value in blood is less than 2.3mmol/L, the risk is determined to be high; when the fasting blood glucose value is between 6.1-7.0mmol/L and the blood triglyceride value is more than or equal to 2.3mmol/L, the risk is determined to be extremely high.

Preferably, the alarm module includes volume key and megaphone, when insulin pump begins the infusion, send the alarm sound that begins to infuse the insulin by alarm module, this alarm sound is long ring sound, accessible volume key regulation alarm sound size, it is long for insulin infusion duration to report to the police sound duration, indicate that patient's insulin pump is working, when insulin quantity is not enough, send the alarm sound that insulin quantity is not enough, this alarm sound is discontinuous short-time promotion sound, end alarm sound after supplementing insulin, still can force to break off alarm sound through double click volume key.

The invention has the beneficial effects that: the system has the characteristics of intelligent equipment and intelligent decision making, and mainly aims to solve the problem of ' how many medicines to be used ' of a patient, in fact, the time and how many medicines to be used ' of the patient are equally important, and the patient forgets when and how much insulin to be taken in real life, and can also help the family members, doctors and the like of the patient to master the medication situation so as to remind and encourage the patient in due time or synthesize other life behavior data of the patient to adjust the management scheme. The sensor module can be used for completing functions of signal transmission, data acquisition and the like in a self-adaptive manner, the working process of the sensor module does not influence normal life work and rest of a patient, real-time monitoring for 24 hours can be realized, safety guarantee is brought to vital signs of the patient, a blood glucose data updating unit in the blood glucose data preprocessing module utilizes a firefly algorithm and is applied to optimization of parameters of a support vector machine, inherent defects of the firefly algorithm are improved, the parameters determined through the improved firefly algorithm can effectively improve the classification precision of the support vector machine, a blood glucose control module combines fitting and a Boltzmann machine algorithm, statistical probability is introduced in neuron state change, balance state Boltzmann distribution of a network is achieved, a network operation mechanism is based on a simulated annealing algorithm, the Boltzmann machine combines advantages of a multi-layer feedforward neural network and a discrete Hofield network in terms of a network structure, a learning algorithm and a dynamic operation mechanism, the Boltzmann machine is established on a Hopfield network base, learning capability of learning, optimal solution can be achieved, continuous monitoring of a low-degree blood glucose concentration monitoring module, and high-risk of a diabetes alarm system can be screened according to high-degree risk of a high-risk alarm system.

Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the protection scope of the present invention, although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A digital health management system for diabetes based on big data and artificial intelligence comprises a sensor module, a blood sugar data preprocessing module, a blood sugar control module, a continuous blood sugar monitoring module and an alarm module, wherein the sensor module comprises a transmitting unit, a sensor unit and a receiving unit, the transmitting unit is used for transmitting a signal for acquiring data to the sensor unit, the sensor unit is used for monitoring the glucose concentration in interstitial fluid through a blood sugar sensor embedded under the skin, the glucose in the interstitial fluid is contacted with the sensor and generates an oxidation reduction reaction, a chemical signal is converted into an electric signal and finally transmitted to the receiving unit, the glucose concentration data is transmitted to the blood sugar data preprocessing module to realize cleaning and screening of abnormal data, the blood sugar control module comprises a blood sugar concentration calculating unit and a blood sugar control unit, the blood sugar concentration calculating unit is used for transmitting the glucose concentration in the interstitial fluid monitored by the blood sugar sensor embedded under the skin into a blood sugar concentration by an algorithm, the blood sugar control unit is used for calculating insulin needed to be infused by a patient and controlling an insulin pump to infuse insulin into the body of the patient by the continuous blood sugar data preprocessing module, the blood sugar data preprocessing module and the blood sugar control module, the blood sugar control module is used for monitoring the insulin pump to supervise the working sign of the insulin pump in real time, and the patient is alarmed by the continuous blood sugar data monitoring moduleThe module sends out an alarm sound for starting insulin infusion to prompt the patient that the insulin pump works, and when the insulin amount is insufficient, the module sends out an alarm sound for reminding the patient that the insulin amount is insufficient to supplement the insulin in time; the blood glucose concentration calculation unit converts the concentration of glucose in interstitial fluid monitored by a subcutaneous blood glucose sensor into the concentration of blood glucose by utilizing an algorithm, the blood glucose is the concentration of glucose in blood, the current common unit is mmol/L, the normal value of fasting blood glucose is 3.9-6.1mmol/L, the fasting blood glucose is more than or equal to 7.0mmol/L, diabetes is considered, the blood glucose of random blood glucose for two hours is more than or equal to 11.1mmol/L (200 mg/dL), the diabetes is also considered, the conversion relation of the glucose and the blood glucose is deduced by adopting a fitting algorithm, the glucose sample in the interstitial fluid is monitored and collected by the blood glucose sensor, and the sample is recorded as x _i I =1,2, …, n, assuming the corresponding blood glucose output is noted as y _i I =1,2, …, n, using least squares to define

Is provided with

for the output of the blood glucose to be fitted,

for the slope of the curve to be fitted,

offset of the fit, using least squares, of

Wherein Loss is a Loss function, k needs to be searched, b minimizes the Loss, and the partial derivatives of k and b can be respectively solved, which comprises:

can obtain

Evaluating how good the fit is using a global sum of squares, where global sum of squares includes sum of squared errors and sum of squared regressions, defining SST = SSE + SSR, where SST represents global sum of squares, SSE represents sum of squared errors, SSR represents sum of squared regressions, and assuming goodness of fit as R ² Definition of

For the sum of squared errors SSE, there are

For regression sum of squares SSR, there are

Wherein

Mean values representing the fit output;

the blood sugar control unit deduces the needed infusion dosage of the insulin pump according to the blood sugar concentration fitted by the blood sugar concentration calculation unit, and defines a visible variable v = {0,1 }by utilizing a machine learning Boltzmann machine algorithm ^D Where D is the dimension of the visible variable, the hidden variable h =0,1 is defined ^P Where P is the dimension of the hidden variable,let L = [ L = _ij ] _D×D Wherein L represents the weight between the visible variables v, and J = [ J = _ij ] _P×P Where J represents the weight between hidden variables h, let W = [ W = _ij ] _D×P Wherein W represents a weight between a visible variable and a hidden variable, defining a joint probability density function

Wherein z represents a hidden variable, p (v, h) represents a joint probability density function with respect to the visible variable and the hidden variable, U (v, h) is an energy function,

in order to find the distribution parameter θ corresponding to the maximum value of the log maximum likelihood estimation, the gradient is found for the distribution parameter θ:

wherein the content of the first and second substances,

for the W gradient direction, there are:

therefore, the temperature of the molten metal is controlled,

wherein

Representation based on p _data Distribution lower pair variable vh ^T In the expectation that,

representation based on p _model Distribution lower pair variable vh ^T Expect that p is satisfied _data ＝p _data (v)p _model (h|v)，p _model ＝p _model (v, h), then the three parameter distribution increments can be expressed as:

wherein, W ^(t) Represents the weight between the visible and hidden variables after t iterations, L ^(t) Representing the weight between visible variables after t iterations, J ^(t) Representing the weight between hidden variables after t iterations, W ^(t+1) Represents the weight between the visible and hidden variables after t +1 iterations, L ^(t+1) Representing the weight between the visible variables after t +1 iterations, J ^(t+1) Representing the weights between hidden variables after t +1 iterations, then the transition probabilities can be written as:

wherein v is _-i Is { v-v _i Set of h _-i Is { h-h _i The set of (c) is (c),

meaning that k takes from 1 to D does not include the accumulation of i,

denotes the accumulation of m from 1 to D excluding j, σ is the sigmoid function, p (v) _i ＝1|h,v _-i ) Denotes given h and v _-i When, v _i Probability of =1, p (h) _j ＝1|v,h _-i ) Denotes that v and h are given _-i When h is present _j Probability of =1, for glycemic control unit, if p (v) _i ＝1|h,v _-i ) The larger the time, the more necessary the infusion of insulin into the patient, if p (h) _j ＝1|v,h _-i ) The larger the probability that insulin needs to be infused into the patient at that moment.

2. The digital health management system for diabetes mellitus based on big data and artificial intelligence of claim 1, wherein the sensor module comprises a transmitting unit, a sensor unit and a receiving unit, wherein the transmitting unit transmits a signal for collecting data to the sensor unit, the sensor unit monitors the glucose concentration in interstitial fluid through a blood glucose sensor embedded under the skin, the glucose in interstitial fluid contacts with the sensor and generates redox reaction, chemical signals are converted into electric signals, and the electric signals are finally transmitted to the receiving unit.

3. The digital health management system for diabetes mellitus based on big data and artificial intelligence as claimed in claim 1, wherein glucose concentration data is transmitted to the blood glucose data preprocessing module to realize cleaning and screening of abnormal data, the abnormal data includes default and low confidence data, the reasons for generating abnormal data include insensitivity of sensor probe and unstable motion state of patient, if abnormal data is generated due to insensitivity of sensor probe, the data will be continuously default and cannot be transmitted to next module, which results in that the system cannot work normally, if abnormal data is generated due to motion state of patient, the data will have default or low confidence value state within a certain time period, the blood glucose data preprocessing module will filter out all abnormal data, and when the patient guarantees the calm state, the system will return to normal working state.

4. The digital diabetes health management system based on big data and artificial intelligence of claim 1, wherein the blood glucose data preprocessing module needs to update the blood glucose data sent from the sensor module at regular time, and comprises a blood glucose data updating unit, the blood glucose data in the sensor module needs to be called to train and test the support vector machine, when the blood glucose data called by the sensor module is used by the blood glucose data updating unit to train the support vector machine, the punishment factor and kernel function parameter of the support vector machine are determined by using firefly algorithm, during the optimization process of the punishment factor and kernel function parameter of the support vector machine by using firefly algorithm, each firefly uses roulette rule to select to move towards an individual with higher fluorescence brightness than self, and determines the moving distance of each firefly according to attraction, on the basis, firefly i is set to select to move towards firefly j, and finally realizes the (t + 1) th time of update, and the specific iterative position updating formula is as follows:

at the upper partIn, x _i (t + 1) denotes the position of firefly i after the (t + 1) th iteration update, x _i (t) indicates the position of firefly i after the updating of the t-th iteration, x _j (t) represents the position of firefly j after the updating of the t-th iteration, β _ij (t) represents the attraction of firefly j to firefly i after the t-th iterative update, α _ij (t) represents the random term coefficient of the firefly i moving randomly towards the firefly j after the t-th iteration updating, rand is a random coefficient obeying normal distribution, and rand belongs to [0,1 ∈ ]]。

5. The digital diabetes health management system based on big data and artificial intelligence of claim 4, wherein the attraction degree β of firefly j to firefly i after the t iteration is updated _ij The value of (t) is set to:

in the above-mentioned formula, the reaction mixture,

the values of (A) are:

wherein, beta ₀ Where r =0, the attraction degree of firefly, that is, the maximum attraction degree, γ is a light absorption coefficient, and represents a characteristic that firefly gradually decreases with increasing distance, and r _ij (t) is the Cartesian distance, rho, between firefly i and firefly j after the tth iterative update _ij (t) is a historical adjustment coefficient of the attraction of firefly j to firefly i after the tth iteration update, and ρ is _ij The value of (t) is:

μ (t) represents an iterative correction coefficient, and

wherein, T _max Is the maximum number of iterations, k _ij (t) represents the statistical coefficient of the history of the region between firefly i and firefly j after the t iteration update, k _ij The value of (t) is:

wherein omega _ij (t) is expressed as a position x _i (t) is the center, with r _ij (t) is a spherical region of radius, x _j (τ) is the position of firefly j after the τ -th iteration update, f (x) _j (τ),Ω _ij (t)) is for position x _j (τ) and region Ω _ij (t) a region between (t) and

m represents the number of fireflies in the population, and t represents the current number of iterations.

6. The digital health management system for diabetes mellitus based on big data and artificial intelligence of claim 4, wherein the coefficient of stochastic term α that will make firefly i move randomly towards firefly j after the t iteration is updated _ij The value of (t) is set to:

representing a set consisting of the fireflies which are selected as the moving directions when the (t + 1) th iteration update is carried out on all the fireflies in the population as K (t), and only keeping one of the repeated fireflies when the repeated fireflies exist in the set K (t) as the moving directions, and defining theta _K (t) represents a moving direction attribute value of firefly in the set K (t), and θ _K The value of (t) is:

in the above formula, M _K (t) represents the number of fireflies in the set K (t), M represents the number of fireflies in the population, y _K (t) represents the dominance of the spatial distribution of fireflies in the set K (t), and y _K The value of (t) is:

wherein, the fireflies in the set K (t) are sequenced according to the absolute fluorescence brightness value from high to low to form a sequence L _K (t), then x' _l (t) represents the sequence L _K Location, x 'of the first firefly in (t) updated at the t iteration' _l+1 (t) represents the sequence L _K (t) the position of the (l + 1) th firefly in (t) after the updating of the t-th iteration, k is a given positive integer, and k is<M _K (t)，x _l,a (t) is distance position x 'in the population after the t iteration update' _l (t) the position of the firefly at the a-th place,

is a sequence L _K (t) the spatial distribution of the first firefly in (t) is compared with the function, and

when theta is _K The value of (t) satisfies: theta.theta. _K (t)>1, then the random term coefficient alpha _ij The value of (t) is:

in the above formula, α ₀ Given an initial random coefficient value, and alpha ₀ ∈[0,1]，ω _j (t) represents the local spatial coefficients of firefly j in set K (t) after the t-th iterative update, ω _j The value of (t) is:

wherein, the first and the second end of the pipe are connected with each other,

represents the sequence L _K (t) A

The location of the individual fireflies after the t-th iteration of updating.

7. The digital diabetes health management system based on big data and artificial intelligence of claim 1, wherein the continuous blood sugar monitoring module monitors the sensor module, the blood sugar data preprocessing module and the blood sugar control module in real time to ensure the vital signs of the patient, during the insulin therapy, the non-intensive therapy patient needs to inject 2-3 times a day, and the intensive therapy patient needs to inject 3-4 times a day.

8. The digital health management system for diabetes mellitus based on big data and artificial intelligence of claim 1, wherein the alarm module comprises a volume key and a loudspeaker, when the insulin pump starts to infuse, the alarm module gives an alarm to start infusing insulin, the alarm is a long sound, the size of the alarm can be adjusted by the volume key, the duration of the alarm is the duration of the insulin infusion, the patient is reminded that the insulin pump is working, when the insulin is insufficient, the alarm is given out when the insulin is insufficient, the alarm is a short sound which is interrupted, the alarm is ended until the insulin is supplemented, and the alarm can be forcibly interrupted by double clicking the volume key.