Injection scheme optimization method based on safe Bayesian optimization and intelligent medicine box

Abstract

The invention provides an injection scheme optimization method based on safe Bayesian optimization and an intelligent medicine box, relates to the technical field of data processing, and is applied to the intelligent medicine box, a pre-filled injection pen is stored in the intelligent medicine box, and the intelligent medicine box is used for optimizing injection time and injection dosage and reminding a user to use the pre-filled injection pen for injection. The injection scheme optimization method comprises the following steps: acquiring user data; based on the user data, determining safety constraint conditions of the user in the injection scheme optimization process; the injection cost of the user in the injection scheme optimization process is set; modeling injection cost and safety constraint conditions by adopting an independent Gaussian process, and constructing a Gaussian process model; constructing an acquisition function for selecting a next injection scheme; based on the acquisition function, solving the Gaussian process model, and determining an optimal injection scheme for next injection; according to the optimal injection scheme, prompt information is sent to a user, and the user is reminded to use the pre-filled injection pen for injection.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an injection scheme optimization method based on safe Bayesian optimization and an intelligent medicine box. Background Art

[0002] A disposable injection pen is a pen-shaped syringe pre-filled with medicine, also known as a prefilled injection pen. According to the latest definition of disposable medical devices by the International Medical Device Industry Association, "medical devices that are only used on one patient (including multiple uses) and are discarded after use are all disposable medical devices", that is, disposable injection pens are disposable medical device products. Currently, such products on the market are mostly used in the field of chronic diseases such as diabetes, mainly including insulin, peptide drugs and various high-value-added biological preparations.

[0003] Prefilled injection pens do not require patients to switch drugs between medicine bottles and syringes when self-injecting. The desired dose can be set directly through the dose setting component and injected. Disposable injection pens have outstanding advantages in terms of convenience, safety, and drug utilization during the drug injection process. They can help ordinary patients master the operating steps more quickly and further improve the self-management of patients in the field of chronic diseases.

[0004] In recent years, the number of drug developments using prefilled injection pens has increased, but currently the injection devices for this type of drug only have basic dosage adjustments and visual estimates of approximate drug dosages, without accurate patient injection dosage records, drug remaining, injection reminders, and other information. For busy office workers and the elderly, there are risks of missed injections, incorrect injections, and insufficient dosages. Currently, many reminders are completed through handwritten marks on the medicine box. If the medicine box is lost, it will bring great difficulties to the injection at the correct time and dosage. Missed injections and wrong injections have adverse effects on patients' regular and standardized treatment. Many patients may experience adverse reactions or recurrences of the disease, and even worsen the disease, seriously affecting the patient's prognosis. Summary of the invention

[0005] In order to solve the technical problem that currently dosage reminders are handwritten on medicine boxes, if the medicine box is lost, it will bring great difficulties to the injection of the correct time and dosage, and missed or wrong injections will have adverse effects on the patient's regular and standardized treatment. Many patients may experience adverse reactions or recurrence of the disease, and even cause the disease to worsen and seriously affect the patient's prognosis, the present invention provides an injection scheme optimization method based on safe Bayesian optimization and a smart medicine box.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect:

[0008] An injection scheme optimization method based on safe Bayesian optimization provided by an embodiment of the present invention is applied to a smart medicine box, wherein the smart medicine box stores a prefilled injection pen, and the smart medicine box is used to optimize the injection time and injection dose, and remind the user to use the prefilled injection pen for injection. The injection scheme optimization method includes:

[0009] S1: Get user data;

[0010] S2: Based on the user data, determining safety constraints of the user during the injection scheme optimization process;

[0011] S3: Set the injection cost of the user during the injection plan optimization process;

[0012] S4: using independent Gaussian processes to model the injection cost and the safety constraint conditions, and constructing a Gaussian process model;

[0013] S5: constructing an acquisition function for selecting the next injection plan;

[0014] S6: solving the Gaussian process model based on the acquisition function to determine the best injection plan for the next injection;

[0015] S7: Based on the optimal injection plan, a prompt message is sent to the user to remind the user to use the prefilled injection pen for injection.

[0016] Second aspect:

[0017] An embodiment of the present invention provides a smart medicine box, comprising:

[0018] processor;

[0019] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the injection scheme optimization method based on safe Bayesian optimization as described in the first aspect is implemented.

[0020] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0021] In the present invention, the smart medicine box can optimize the next injection plan based on a safe Bayesian optimization method to ensure that the drug exerts the maximum therapeutic effect at the most suitable time point and dosage range, and then send reminders to patients according to the optimized injection plan to help patients inject on time and in the right amount to avoid missing or wrong injections. There is no need to make handwritten marks on the medicine box, which improves patients' medication experience and compliance, reduces the burden on patients and medical staff, and is an important tool for self-management and precision treatment of chronic diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 A schematic diagram of a flow chart of an injection scheme optimization method based on safe Bayesian optimization provided by an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of the structure of a smart medicine box provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0026] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0027] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0028] Reference Manual Attached Figure 1 , showing a flow chart of an injection scheme optimization method based on safe Bayesian optimization provided by an embodiment of the present invention.

[0029] The embodiment of the present invention provides an injection scheme optimization method based on safe Bayesian optimization, which is applied to a smart medicine box, wherein the smart medicine box stores a prefilled injection pen, and the smart medicine box is used to optimize the injection time and injection dose, and remind the user to use the prefilled injection pen for injection. The injection scheme optimization method based on safe Bayesian optimization may include the following steps:

[0030] S1: Get user data.

[0031] Among them, user data includes: blood sugar level, body temperature, heart rate, blood pressure, age, gender, weight, height, hypoglycemia records, allergic reaction records, medication data, daily routine and dietary patterns, etc.

[0032] S2: Based on user data, determine the user's safety constraints during the injection plan optimization process.

[0033] Optionally, the safety constraints specifically include: blood sugar constraints, drug dosage constraints, injection interval constraints, drug stability constraints, dietary constraints, and individual patient constraints.

[0034] Optionally, blood sugar constraints are used to prevent hypoglycemia and ensure that the patient's blood sugar level is safe and stable after injection:

[0035] f 1 (x) = A(x) - A s

[0036] Among them, f 1 (x) represents the blood glucose constraint, x represents the injection scheme, A(x) represents the predicted blood glucose value when the injection scheme x is adopted, and A s Indicates the blood sugar safety threshold.

[0037] Optionally, drug dose constraints are used to limit the single injection dose to avoid drug toxicity or side effects caused by overdose:

[0038] f 2 (x) = g m -g

[0039] Among them, f 2 (x) represents the drug dosage constraint, g m represents the maximum safe dose, and g represents the injection dose.

[0040] Optionally, injection interval constraints are used to ensure that the injection interval is sufficient to prevent excessive accumulation of drugs in the body and affect efficacy and safety:

[0041] f 3 (x) = Δt-t s

[0042] Among them, f 3 (x) represents the injection interval constraint, Δt represents the current injection interval, t s Indicates safe injection interval.

[0043] Optionally, drug stability constraints are used to ensure that the drug is stored within an appropriate temperature range to maintain drug efficacy and chemical stability:

[0044] f 4 (x) = min(C(x) - C min ,C max -C(x)

[0045] Among them, f 4 (x) represents the drug stability constraint, min represents the smaller value, C(x) represents the temperature prediction value when injection scheme x is adopted, and C minIndicates the lower limit of the drug's suitable temperature, C max Indicates the upper limit of suitable temperature for drugs.

[0046] Optionally, dietary constraints are used to optimize the injection time to improve drug efficacy or reduce side effects based on the drug's dietary dependency requirements:

[0047] When the dietary restriction is injection with meals, the dietary restriction is as follows:

[0048] f 5 (x) = 1-|tt meal |

[0049] Among them, f 5 (x) represents dietary restriction, t represents injection time, t meal Indicates meal time.

[0050] When the dietary restriction is fasting injection, the dietary restriction is as follows:

[0051] f 5 (x)=|tt meal |.

[0052] Optionally, individual patient constraints are used to limit the dose according to the patient's tolerance, avoid individual adverse reactions, and achieve personalized medication:

[0053] f 6 (x) = g s -g

[0054] Among them, f 6 (x) represents the patient's individual constraint, g s represents the patient tolerated dose, and g represents the injected dose.

[0055] S3: Set the injection cost of the user during the injection plan optimization process.

[0056] Optionally, the injection regimen includes: injection time and corresponding injection dose.

[0057] Optionally, the injection cost includes a pain score, a side effect probability value, and a blood sugar fluctuation value:

[0058] f 0 (x) = ω 1 ×S t +ω 2 ×S f +ω 3 ×S x

[0059] Among them, f 0 (x) represents the injection cost, S t represents pain score, S fRepresents the probability value of side effects, S x Indicates blood sugar fluctuation value, ω 1 represents the weight coefficient of pain score, ω 2 The weight coefficient representing the probability value of side effects, ω 3 Represents the weight coefficient of blood sugar fluctuation value.

[0060] Among them, those skilled in the art can set the weight coefficient ω of the pain score according to actual conditions. 1 , the weight coefficient of the side effect probability value ω 2 And the weight coefficient of blood sugar fluctuation value ω 3 The present invention does not limit the size.

[0061] In the present invention, the injection cost function not only focuses on the therapeutic effect when optimizing the injection plan, but also considers the comfort and safety of the patient to achieve all-round optimization.

[0062] S4: Use independent Gaussian processes to model injection costs and safety constraints and construct a Gaussian process model.

[0063] Among them, Gaussian process is a flexible non-parametric Bayesian model, which is often used in function modeling and uncertainty quantification, and is widely used in tasks such as regression, classification, and optimization.

[0064] Optionally, the Gaussian process model is specifically:

[0065] f i (x)~GP[μ i (x),k i (x,x′)]

[0066] Among them, f i (x) represents the Gaussian process function of the i-th safety constraint condition, x represents the injection plan, which includes the injection time and the corresponding injection dose, GP represents the Gaussian process, μ i (x) represents the mean function of the i-th safety constraint condition, k i (x,x′) represents the covariance function of the i-th safety constraint.

[0067] In this invention, the injection cost and safety constraints are modeled through Gaussian process, which can achieve flexible, accurate and efficient injection scheme optimization, while providing patients with a safer and more personalized medication experience. Its advantages in uncertainty quantification, dynamic adjustment and optimization efficiency make it a core tool in injection scheme optimization.

[0068] S5: Construct an acquisition function for selecting the next injection plan.

[0069] Optionally, the acquisition function is the key to Bayesian optimization and is used to select the optimal injection plan for the next step. The acquisition function is specifically:

[0070]

[0071] Among them, x n represents the injection scheme at the nth iteration, arg min represents the minimum value of the function, α 0 (x) represents the unconstrained acquisition function, τ i represents the constraint weight of the i-th safety constraint, B i (x) represents the obstacle value of the i-th safety constraint, and m represents the total number of safety constraints.

[0072] In the present invention, the acquisition function can balance the optimization goal and safety constraints, improve the safety and effect of the injection plan, and support dynamic adjustment and exploration to meet the personalized needs of patients. Through the introduction of confidence intervals and logarithmic barrier values, the acquisition function can generate the optimal injection plan under complex conditions, which is a safe and efficient intelligent optimization method.

[0073] Optionally, the unconstrained acquisition function specifically adopts an induced unconstrained acquisition function:

[0074]

[0075] in, represents the posterior mean based on the Gaussian process at the n-1th iteration, represents the posterior variance based on the Gaussian process at the n-1th iteration, Represents the confidence level scaling parameter of the Gaussian process at the nth iteration.

[0076] It should be noted that the unconstrained acquisition function focuses on minimizing the injection cost (such as pain score, side effect probability, blood sugar fluctuation). By adopting the induced unconstrained acquisition function, by combining the posterior mean and variance of the Gaussian process, dynamically balancing exploration and utilization, it is possible to quantify uncertainty in the optimization process, improve the safety and robustness of the model, and support dynamic adjustment of the confidence level to adapt to different optimization stages, simplify the objective function to improve computational efficiency, and make the optimization solution more accurate and efficient and meet personalized needs.

[0077] Optionally, the barrier value is:

[0078]

[0079] Among them, ln represents the logarithmic function with a natural constant as the base, represents the mean function of the ith safety constraint at the n-1th iteration, represents the variance function of the ith safety constraint at the n-1th iteration, represents the confidence level scaling parameter of the ith safety constraint at the nth iteration.

[0080] Among them, the confidence level scaling parameter is a key parameter in the Gaussian process, which is used to control the range of prediction uncertainty, thereby affecting the handling and exploration of safety constraints during the optimization process. It plays an important role in the calculation of acquisition functions and barrier values.

[0081] It should be noted that the barrier value is used to quantify the degree of violation of each safety constraint, and can dynamically quantify the degree of satisfaction of the safety constraint, ensuring that the optimization process always runs within a safe range and preventing the violation of the constraint, thereby improving the safety and reliability of the optimization, and can control the importance of each constraint in the optimization through weights.

[0082] S6: Based on the acquisition function, the Gaussian process model is solved to determine the best injection plan for the next injection.

[0083] In a possible implementation, a Bayesian optimization iterative mechanism is used to solve the problem. S6 specifically includes sub-steps S601 to S607:

[0084] S601: Calculate an unconstrained acquisition function based on current user data and a Gaussian process model.

[0085] S602: Calculate the barrier value of the safety constraint condition by using the confidence level scaling parameter.

[0086] S603: Select the next injection plan with the goal of minimizing the acquisition function.

[0087] S604: Perform injection according to the selected injection scheme, and observe the actual values ​​of the injection cost and safety constraints.

[0088] S605: Update user data according to this injection plan.

[0089] S606: Update the Gaussian process model according to the actual values ​​of the injection cost and the safety constraint.

[0090] S607: Determine whether the termination condition is met. If so, terminate the iteration and output the currently determined injection scheme as the best injection scheme. Otherwise, return to S601 to continue the iteration.

[0091] In the present invention, the iterative mechanism of Bayesian optimization is adopted to realize dynamic, personalized and safe optimization of injection schemes, which can not only improve the treatment effect, but also reduce the patient risk and operation complexity. It is efficient and robust at the same time, and is an important technical support for the smart medicine box to realize precision medicine.

[0092] S7: Based on the optimal injection plan, a prompt message is sent to the user to remind the user to use the prefilled injection pen for injection.

[0093] Optionally, a preset audio may be played for prompting, or a corresponding optimal injection scheme may be displayed on a display screen for prompting.

[0094] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0095] In the present invention, the smart medicine box can optimize the next injection plan based on a safe Bayesian optimization method to ensure that the drug exerts the maximum therapeutic effect at the most suitable time point and dosage range, and then send reminders to patients according to the optimized injection plan to help patients inject on time and in the right amount to avoid missing or wrong injections. There is no need to make handwritten marks on the medicine box, which improves patients' medication experience and compliance, reduces the burden on patients and medical staff, and is an important tool for self-management and precision treatment of chronic diseases.

[0096] Reference Manual Attached Figure 2 , showing a schematic structural diagram of a smart medicine box provided by the present invention.

[0097] The present invention also provides a smart medicine box 20, comprising:

[0098] Processor 201;

[0099] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201 , the injection scheme optimization method based on safe Bayesian optimization as in the method embodiment is implemented.

[0100] The smart medicine box 20 provided by the present invention can execute the above-mentioned injection scheme optimization method based on safe Bayesian optimization and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0101] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0102] In the present invention, the smart medicine box can optimize the next injection plan based on a safe Bayesian optimization method to ensure that the drug exerts the maximum therapeutic effect at the most suitable time point and dosage range, and then send reminders to patients according to the optimized injection plan to help patients inject on time and in the right amount to avoid missing or wrong injections. There is no need to make handwritten marks on the medicine box, which improves patients' medication experience and compliance, reduces the burden on patients and medical staff, and is an important tool for self-management and precision treatment of chronic diseases.

[0103] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0104] There are a few points to note:

[0105] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0106] (2) For the sake of clarity, the thickness of a film or region in the drawings used to describe the embodiments of the present invention is enlarged or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a film, a region or a substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0107] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0108] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for optimizing injection scheme based on safe Bayesian optimization, characterized in that: Applied to a smart medicine box, the smart medicine box stores a prefilled injection pen, the smart medicine box is used to optimize the injection time and injection dose, and remind the user to use the prefilled injection pen for injection. The injection scheme optimization method includes: S1: Get user data; S2: Based on the user data, determining safety constraints of the user during the injection scheme optimization process; S3: Set the injection cost of the user during the injection plan optimization process; S4: using independent Gaussian processes to model the injection cost and the safety constraint conditions, and constructing a Gaussian process model; S5: constructing an acquisition function for selecting the next injection plan; S6: solving the Gaussian process model based on the acquisition function to determine the best injection plan for the next injection; S7: Based on the optimal injection plan, a prompt message is sent to the user to remind the user to use the prefilled injection pen for injection.

2. The injection scheme optimization method based on safe Bayesian optimization according to claim 1, characterized in that: The injection scheme includes: injection time and corresponding injection dosage.

3. The injection scheme optimization method based on safe Bayesian optimization according to claim 1, characterized in that: The safety constraints specifically include: blood sugar constraints, drug dosage constraints, injection interval constraints, drug stability constraints, dietary constraints, and individual patient constraints; The blood sugar constraint is used to prevent hypoglycemia and ensure the safety and stability of the patient's blood sugar level after injection; The drug dosage constraint is used to limit the single injection dosage to avoid drug toxicity or side effects caused by excessive injection; The injection interval constraint is used to ensure that the injection interval is sufficient to prevent excessive accumulation of drugs in the body and affect efficacy and safety; The drug stability constraint is used to ensure that the drug is stored within an appropriate temperature range to maintain the drug's efficacy and chemical stability; The dietary restriction is used to optimize the injection time to improve drug efficacy or reduce side effects according to the dietary dependence requirements of the drug; The individual patient constraints are used to limit the dosage according to the patient's tolerance, avoid individual adverse reactions, and achieve personalized medication.

4. The injection scheme optimization method based on safe Bayesian optimization according to claim 1, characterized in that: The injection cost includes a pain score, a side effect probability value, and a blood sugar fluctuation value.

5. The injection scheme optimization method based on safe Bayesian optimization according to claim 1, characterized in that: The Gaussian process model is specifically: F i (x)~GP[μ i (x),k i (x,x′)] Among them, f i (x) represents the Gaussian process function of the i-th safety constraint condition, x represents the injection scheme, which includes the injection time and the corresponding injection dose, GP represents the Gaussian process, μ i (x) represents the mean function of the i-th safety constraint condition, k i (x,x′) represents the covariance function of the i-th safety constraint.

6. The injection scheme optimization method based on safe Bayesian optimization according to claim 1, characterized in that: The acquisition function is specifically: Among them, x n represents the injection scheme at the nth iteration, arg min represents the minimum value of the function, α0(x) represents the unconstrained acquisition function, τ i represents the constraint weight of the i-th safety constraint, B i (x) represents the obstacle value of the i-th safety constraint, and m represents the total number of safety constraints.

7. The injection scheme optimization method based on safe Bayesian optimization according to claim 6, characterized in that: The unconstrained acquisition function specifically adopts the induced unconstrained acquisition function: in, represents the posterior mean based on the Gaussian process at the n-1th iteration, represents the posterior variance based on the Gaussian process at the n-1th iteration, Represents the confidence level scaling parameter of the Gaussian process at the nth iteration.

8. The injection scheme optimization method based on safe Bayesian optimization according to claim 6, characterized in that: The barrier value is specifically: Among them, ln represents the logarithmic function with a natural constant as the base, represents the mean function of the ith safety constraint at the n-1th iteration, represents the variance function of the ith safety constraint at the n-1th iteration, represents the confidence level scaling parameter of the ith safety constraint at the nth iteration.

9. The injection scheme optimization method based on safe Bayesian optimization according to claim 6, characterized in that: The S6 specifically includes: S601: Calculate an unconstrained acquisition function based on current user data and a Gaussian process model; S602: Calculate the barrier value of the safety constraint condition by using the confidence level scaling parameter; S603: Selecting the next injection plan with the goal of minimizing the acquisition function; S604: Perform injection according to the selected injection scheme, and observe the actual values ​​of the injection cost and the safety constraint; S605: updating user data according to this injection plan; S606: updating the Gaussian process model according to the actual values ​​of the injection cost and the safety constraint; S607: Determine whether the termination condition is met; if so, terminate the iteration and output the currently determined injection scheme as the optimal injection scheme; otherwise, return to S601 to continue iteration.

10. A smart medicine box, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the injection scheme optimization method based on safe Bayesian optimization as described in any one of claims 1 to 7 is implemented.

Images