Implanted sensor regulation and control data analysis method and system based on machine learning

Abstract

The invention provides an implanted sensor regulation and control data analysis method and system based on machine learning, relates to the technical field of implantation regulation and control application, and aims to enlarge a parameter search space, explore and apply a potential optimization scheme neglected in the past and effectively reduce the loss of a regulation and control effect by introducing an unbalanced associated reference optimization parameter group. Specifically, by comprehensively considering various balanced and unbalanced reference optimization parameter combinations, personalized regulation and control can be carried out for physiological characteristics and disease states of different patients, the treatment effect is remarkably improved, and a result of regulation and control optimization each time is taken as a training sample of machine learning, so that online learning and updating of machine learning are realized; according to the method, the prediction precision and generalization ability are continuously improved, so that the method can better adapt to the change of the physiological state of the patient, thereby providing a scientific and quantitative regulation and control optimization basis, improving the automation and intelligence level of the whole optimization process, and developing a new possibility for the application of implantable medical equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of implant regulation and control, in particular to an implant sensor regulation and control data analysis method and system based on machine learning. BACKGROUND

[0002] With the rapid development of medical technology, implantable medical devices are increasingly widely used in clinical medicine, especially in the field of neural regulation, such as the treatment of Parkinson's disease, epilepsy and other nervous system diseases. As a key component of these devices, implantable sensors can monitor patients' physiological indicators in real time, providing important diagnostic and treatment basis for doctors. However, how to effectively regulate these implantable sensors to achieve the best treatment effect is an important challenge faced by the medical community and the engineering technology field.

[0003] Traditional implantable sensor regulation methods mainly rely on doctors' experience or pre-set fixed parameters, and these methods often have unstable regulation effects, poor adaptability, and difficulty in coping with individual differences. SUMMARY

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of the embodiments of the present application is to provide an implant sensor regulation and control data analysis method and system based on machine learning.

[0005] According to one aspect of the present application, an implant sensor regulation and control data analysis method based on machine learning is provided, the method comprising: obtaining implant sensor regulation signal data to be optimized, and obtaining X reference optimization parameter groups; there are at least one pair of reference optimization parameters in the X reference optimization parameter groups that present an unbalanced correlation; X is a positive integer greater than 1; determining the regulation effect loss required for regulating and optimizing the target implant sensor regulation signal based on the X reference optimization parameter groups under the target implantable neural regulation state label, generating X regulation effect losses; the target implant sensor regulation signal belongs to the implant sensor regulation signal data; based on the X regulation effect losses, extracting a target optimization parameter group of the implant sensor regulation signal data under the target implantable neural regulation state label from the X reference optimization parameter groups; based on the target optimization parameter group under the target implantable neural regulation state label, regulating and optimizing the implant sensor regulation signal in the implant sensor regulation signal data, generating a target regulation and optimization result of the implant sensor regulation signal data.

[0006] In a possible implementation of the first aspect, the X reference optimization parameter groups include a first reference optimization parameter group and a second reference optimization parameter group, reference optimization parameter pairs in the first reference optimization parameter group exhibit unbalanced correlation; wherein the reference optimization parameter pairs in the first reference optimization parameter group are determined based on different reference optimization parameters in a same reference optimization parameter range; or the reference optimization parameter pairs in the first reference optimization parameter group are determined based on reference optimization parameters in different reference optimization parameter ranges, the optimization parameters in the different reference optimization parameter ranges are different; reference optimization parameter pairs in the second reference optimization parameter group exhibit balanced correlation, wherein the reference optimization parameter pairs in the second reference optimization parameter group are determined based on a same reference optimization parameter in a same reference optimization parameter range.

[0007] In a possible implementation of the first aspect, the X reference optimization parameter groups are obtained by: determining a control effect loss required for control optimization of the target implanted sensor control signal under the target implanted neuromodulation state label according to a balanced optimization parameter group Ya, the balanced optimization parameter group Ya belongs to Y balanced optimization parameter groups, the balanced optimization parameter groups in the Y balanced optimization parameter groups cover reference optimization parameter pairs that exhibit balanced correlation, Y is a positive integer greater than 1, and a is a positive integer not less than Y; if the Y control effect losses corresponding to the Y balanced optimization parameter groups are obtained, extracting a basic optimization parameter group of the target implanted sensor control signal under the target implanted neuromodulation state label from the Y balanced optimization parameter groups based on the Y control effect losses; updating the basic optimization parameter group based on the variation amplitudes in the variation amplitude sequence to generate the X reference optimization parameter groups.

[0008] In a possible implementation of the first aspect, the X reference optimization parameter groups include a third reference optimization parameter group; reference optimization parameter pairs in the third reference optimization parameter group exhibit unbalanced correlation; one reference optimization parameter in the third reference optimization parameter group is obtained by updating a first basic optimization parameter in the basic optimization parameter group based on a first variation amplitude in the variation amplitude sequence; another reference optimization parameter in the third reference optimization parameter group is obtained by updating a second basic optimization parameter in the basic optimization parameter group based on a second variation amplitude in the variation amplitude sequence.

[0009] In a possible implementation of the first aspect, the X reference optimization parameter groups further include a fourth reference optimization parameter group; reference optimization parameters in the fourth reference optimization parameter group are in balanced correlation with each other; One reference optimization parameter in the fourth reference optimization parameter group is obtained by updating the first basic optimization parameter based on a third variation amplitude in the variation amplitude sequence; Another reference optimization parameter in the fourth reference optimization parameter group is obtained by updating the second basic optimization parameter based on an inverse of the third variation amplitude.

[0010] In a possible implementation of the first aspect, the determining of the control effect loss required for the control optimization of the target implant sensor control signal based on the X reference optimization parameter groups under the target implantable neural control state label includes: Obtaining a target control signal node in the target implant sensor control signal, determining a candidate signal node of the target control signal node under the target implantable neural control state label from the target implant sensor control signal; Determining a control deviation degree between a control parameter of the target control signal node and a control parameter of the candidate signal node; Comparing the control deviation degree with the X reference optimization parameter groups respectively to generate comparison information corresponding to the X reference optimization parameter groups respectively; Determining a corresponding control optimization knowledge point of the target control signal node under a reference optimization parameter group Yb based on the comparison information corresponding to the reference optimization parameter group Yb; the reference optimization parameter group Yb belongs to the X reference optimization parameter groups, and b is a positive integer not greater than X, until the corresponding control optimization knowledge points of the target control signal node under the X reference optimization parameter groups are obtained respectively; Based on the X control optimization knowledge points obtained, the control parameters of the target control signal node are processed for control optimization respectively to generate X control optimization results; Based on the X control optimization results, the control effect loss required for the control optimization of the target implant sensor control signal based on the X reference optimization parameter groups under the target implantable neural control state label is determined to generate X control effect losses.

[0011] In a possible implementation of the first aspect, the candidate signal nodes include a first candidate signal node and a second candidate signal node, and the regulation deviation includes a first regulation deviation and a second regulation deviation, the first regulation deviation being a regulation deviation between a regulation parameter of the first candidate signal node and a regulation parameter of the target regulation signal node, and the second regulation deviation being a regulation deviation between a regulation parameter of the second candidate signal node and the regulation parameter of the target regulation signal node. The comparing the regulation deviations with the X reference optimization parameter groups respectively generates comparison information corresponding to the X reference optimization parameter groups respectively, and includes: The number of regulation optimization knowledge points under the target implantable neural regulation state label is obtained, and based on the number of regulation optimization knowledge points under the target implantable neural regulation state label, the first comparison information between the first regulation deviation and the reference optimization parameter group Yb is obtained. Based on the number of regulation optimization knowledge points under the target implantable neural regulation state label, the second comparison information between the second regulation deviation and the reference optimization parameter group Yb is obtained. The first comparison information and the second comparison information are taken as the comparison information corresponding to the reference optimization parameter group Yb, and until the comparison information corresponding to the X reference optimization parameter groups respectively is obtained. In a possible implementation of the first aspect, the determining, based on the comparison information corresponding to the reference optimization parameter group Yb, of the regulation optimization knowledge point corresponding to the target regulation signal node under the reference optimization parameter group Yb includes: Based on the first comparison information, a first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb is determined. Based on the second comparison information, a second knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb is determined. Based on the first knowledge point mapping value and the second knowledge point mapping value, the regulation optimization knowledge point of the target regulation signal node under the reference optimization parameter group Yb is determined.

[0012] In a possible implementation of the first aspect, the reference optimization parameter group Yb includes a first reference optimization parameter and a second reference optimization parameter, the first reference optimization parameter being smaller than the second reference optimization parameter, and the regulation optimization knowledge point including four knowledge points. The determining, based on the first comparison information, of the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb includes: if the first comparison information represents that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is less than a first reference optimization parameter, a first mapping value is output as a first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; if the first comparison information represents that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is not less than the first reference optimization parameter, a second mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; if the first comparison information represents that the first regulation deviation degree is not a decrease deviation, and the first regulation deviation degree is less than a second reference optimization parameter, a third mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; if the first comparison information represents that the first regulation deviation degree is an increase deviation, and the first regulation deviation degree is not less than the second reference optimization parameter, a fourth mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; the first mapping value, the second mapping value, the third mapping value, and the fourth mapping value have mapping relationship with four regulation optimization knowledge points respectively.

[0013] In a possible implementation of the first aspect, the reference optimization parameter group Yb includes a first reference optimization parameter and a second reference optimization parameter, the first reference optimization parameter is less than the second reference optimization parameter, and the regulation optimization knowledge points include three knowledge points. The first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb is determined based on the first comparison information, and includes: if the first comparison information represents that the first regulation deviation degree is less than the first reference optimization parameter, a fifth mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; if the first comparison information represents that the first regulation deviation degree is greater than the second reference optimization parameter, a sixth mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; if the first comparison information represents that the first regulation deviation degree is equal to the first reference optimization parameter, or the first regulation deviation degree is equal to the second reference optimization parameter, a seventh mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; If the first control information represents that the first regulation deviation is greater than the first reference optimization parameter, and the first regulation deviation is less than the second reference optimization parameter, a seventh mapping value is output as a first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group Yb; the fifth mapping value, the sixth mapping value, and the seventh mapping value are in mapping relationship with three kinds of regulation optimization knowledge points, respectively.

[0014] For example, in a possible implementation of the first aspect, the regulation optimization processing on the regulation parameters of the target regulation signal node based on the X regulation optimization knowledge points includes: querying the X regulation optimization quantities corresponding to the X regulation optimization knowledge points from the regulation optimization quantity template to generate X regulation optimization quantities; performing regulation optimization processing on the regulation parameters of the target regulation signal node based on the X regulation optimization quantities to generate X regulation optimization results.

[0015] For example, in a possible implementation of the first aspect, the determination of the regulation effect loss required for the regulation optimization of the target implanted sensor regulation signal based on the X reference optimization parameter groups under the target implanted neural regulation state label includes: determining the regulation error rate of the target implanted sensor regulation signal under the reference optimization parameter group Yb based on the regulation optimization result Yb covering the regulation optimized target implanted sensor regulation signal and the initial regulation signal corresponding to the target implanted sensor regulation signal; the reference optimization parameter group Yb belongs to the X reference optimization parameter groups, b is a positive integer not greater than X, the regulation optimization result Yb is a regulation optimization result obtained by performing regulation optimization processing on the target implanted sensor regulation signal based on the reference optimization parameter group Yb, and the target implanted sensor regulation signal is obtained by reconstructing the encoding features of the initial regulation signal; obtaining the performance evaluation index of the reference optimization parameter group Yb, determining the regulation effect loss of the target implanted sensor regulation signal under the reference optimization parameter group Yb based on the regulation error rate of the target implanted sensor regulation signal under the reference optimization parameter group Yb and the performance evaluation index of the reference optimization parameter group Yb, and generating X regulation effect losses until the X regulation effect losses corresponding to the target implanted sensor regulation signal under the X reference optimization parameter groups are obtained.

[0016] For example, in a possible implementation form of the first aspect, the extracting, based on the X regulation effect losses, the target optimization parameter set of the implanted sensor regulation signal data under the target implanted neural regulation state label from the X reference optimization parameter sets, comprises: extracting, based on the X regulation effect losses, a reference optimization parameter set with a minimum regulation effect loss from the X reference optimization parameter sets; outputting the extracted reference optimization parameter set as the target optimization parameter set of the implanted sensor regulation signal data under the target implanted neural regulation state label.

[0017] For example, in a possible implementation form of the first aspect, the number of the target implanted neural regulation state labels is Z, and Z is a positive integer; The method further comprises: performing regulation optimization on a regulation parameter of the implanted sensor regulation signal of the implanted sensor regulation signal data based on the target optimization parameter set under the target implanted neural regulation state label Kc, to generate a regulation optimization result of the implanted sensor regulation signal of the implanted sensor regulation signal data under the target implanted neural regulation state label Kc; the target implanted neural regulation state label Kc belongs to Z target implanted neural regulation state labels, and c is a positive integer not greater than Z; determining a regulation effect loss of the implanted sensor regulation signal of the implanted sensor regulation signal data under the target implanted neural regulation state label Kc based on the regulation optimization result of the implanted sensor regulation signal of the implanted sensor regulation signal data under the target implanted neural regulation state label Kc; If the Z regulation effect losses of the implanted sensor regulation signal of the implanted sensor regulation signal data under the Z target implanted neural regulation state labels are obtained, a regulation optimization result corresponding to a minimum regulation effect loss in the Z regulation effect losses is output as the target regulation optimization result of the implanted sensor regulation signal of the implanted sensor regulation signal data.

[0018] For example, in a possible implementation form of the first aspect, the method further comprises: loading the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label to the modulator, the modulator being configured to determine the target modulation optimization result of the implant sensor modulation signal data based on the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label.

[0019] For example, in a possible implementation manner of the first aspect, the loading the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label to the modulator comprises: obtaining a first parameter label code of the first evaluation index of the first optimization parameter in a first reference optimization parameter range, and a second parameter label code of the second evaluation index of the second optimization parameter in a second reference optimization parameter range; loading the first parameter label code of the first optimization parameter group and the second parameter label code of the second optimization parameter group to the modulator.

[0020] For example, in a possible implementation manner of the first aspect, the loading the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label to the modulator comprises: obtaining a first parameter label code of the first evaluation index of the first optimization parameter in a first reference optimization parameter range, and a second parameter label code of the second evaluation index of the second optimization parameter in a second reference optimization parameter range; obtaining an encoding deviation value between the first parameter label code and the second parameter label code; loading the first parameter label code and the encoding deviation value to the modulator, or loading the second parameter label code and the encoding deviation value to the modulator.

[0021] For example, in a possible implementation manner of the first aspect, the loading the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label to the modulator comprises: generating a first representation attribute corresponding to the first evaluation index of the first optimization parameter; the first representation attribute is used to represent that the first evaluation index is the evaluation index of the first optimization parameter; generating a second representation attribute corresponding to the second evaluation index of the second optimization parameter; the second representation attribute is used to represent that the second evaluation index is the evaluation index of the second optimization parameter; loading the first evaluation index, the first characterization attribute, the second evaluation index and the second characterization attribute to the controller.

[0022] For example, in a possible implementation of the first aspect, the loading the first optimization parameter and the second optimization parameter in the target optimization parameter under the target implantable neuromodulation state label to the controller comprises: generating a first characterization attribute of a first evaluation index corresponding to the first optimization parameter group; the first characterization attribute is used to characterize the first evaluation index as the evaluation index of the first optimization parameter; generating a second characterization attribute of a second evaluation index corresponding to the second optimization parameter group; the second characterization attribute is used to characterize the second evaluation index as the evaluation index of the second optimization parameter; obtaining a target loss value between the first evaluation index and the second evaluation index; loading the first evaluation index, the first characterization attribute and the target loss value to the controller; or loading the second evaluation index, the second characterization attribute and the target loss value to the controller.

[0023] According to an aspect of the embodiments of the present application, a deep learning system is provided, which comprises a processor and a machine readable storage medium, the machine readable storage medium stores machine executable instructions, the machine executable instructions are loaded and executed by the processor to implement the machine learning based implantable sensor control data analysis method in any of the possible implementation manners.

[0024] According to an aspect of the embodiments of the present application, a computer program product or computer program is provided, which comprises computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in the various optional implementation manners of the above three aspects.

[0025] In the technical solutions provided by some embodiments of the present application, the precision and efficiency of implant sensor regulation optimization are significantly improved. By introducing an unbalanced reference optimization parameter group, the parameter search space is expanded, and potential optimization schemes that have been ignored in the past are successfully explored and applied, thereby effectively reducing the regulation effect loss. Specifically, by comprehensively considering a variety of balanced and unbalanced reference optimization parameter combinations, individualized regulation can be performed according to the physiological characteristics and disease states of different patients, the treatment effect is significantly improved, and the results of each regulation optimization are used as training samples for machine learning, online learning and updating of machine learning are realized, the prediction accuracy and generalization ability are continuously improved, and machine learning can better adapt to changes in the physiological state of the patient, thereby providing a scientific and quantitative basis for regulation optimization and improving the automation and intelligence level of the entire optimization process, thereby opening up new possibilities for the application of implantable medical devices. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be enabled in the embodiments will be briefly introduced below, and it should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be extracted in combination with these drawings without creative labor.

[0027] Figure 1 A flowchart of a machine learning-based implant sensor regulation data analysis method provided by an embodiment of the present application is shown in FIG. 1. Figure 2 An architecture schematic block diagram of a deep learning system for implementing the machine learning-based implant sensor regulation data analysis method described above is shown in FIG. 2. DETAILED DESCRIPTION

[0028] The following description is provided to enable any person skilled in the art to implement and combine the present application, and is provided in the context of specific application scenarios and their requirements. It is obvious for those skilled in the art that various changes can be made to the disclosed embodiments, and the general principles defined in the present application can be applied to other embodiments and application scenarios without departing from the principles and scope of the present application. Therefore, the present application is not limited to the described embodiments, but should be given the broadest scope consistent with the claims.

[0029] Figure 1 A flowchart of a machine learning-based implant sensor regulation data analysis method provided by an embodiment of the present application is shown in FIG. 1.

[0030] Step S110, obtaining the implant sensor regulation signal data to be optimized, and obtaining X reference optimization parameter groups. There are at least one pair of reference optimization parameters in the X reference optimization parameter groups that present an unbalanced correlation. X is a positive integer greater than 1.

[0031] In detail, the implant sensor regulation signal data is collected from the sensor implanted in the patient's body, reflecting the patient's physiological state or pathological condition. The reference optimization parameter group is a series of parameter combinations set in advance to try different regulation strategies. Each reference optimization parameter group contains multiple parameters such as stimulation frequency, stimulation intensity, etc., which jointly act on the implanted sensor to achieve a specific therapeutic effect. Among them, if the ratio or relationship between certain parameters in the reference optimization parameter group is not regular or linear, it is considered to present an unbalanced correlation. For example, a combination of high stimulation frequency and low stimulation intensity is uncommon or special compared to the conventional treatment range.

[0032] In this embodiment, it is assumed that an implanted sensor system for monitoring and adjusting the brain waves of a Parkinson's patient is being processed. The server first receives regulation signal data from the implanted sensor, which reflects the patient's brain wave activity pattern in a specific time period.

[0033] At the same time, the server obtains 5 reference optimization parameter groups (X = 5). Taking two of the reference optimization parameter groups as examples: Reference optimization parameter group 1: stimulation frequency: 50Hz, stimulation intensity: 2mA In this group, the parameter pair of stimulation frequency and stimulation intensity presents a balanced correlation, and they are both in the common treatment range and the ratio is relatively stable.

[0034] Reference optimization parameter group 2: stimulation frequency: 100Hz, stimulation intensity: 0.5mA In this group, the stimulation frequency is high and the stimulation intensity is low, and the parameter pair presents an unbalanced correlation. This imbalance may be based on specific experimental research or an attempt to set up for certain special symptoms.

[0035] Step S120, determining the regulation effect loss required for regulation optimization of the target implant sensor regulation signal based on the X reference optimization parameter groups under the target implantable neural regulation state label, and generating X regulation effect losses. The target implant sensor regulation signal belongs to the implant sensor regulation signal data.

[0036] In detail, the target implantable neural regulation state label is a specific treatment target or desired treatment effect, which is used to guide the direction of regulation optimization. For example, "relieving Parkinson's symptoms - hand tremor" is a specific treatment target.

[0037] The loss of regulation effect is the difference between the actual treatment effect and the ideal effect when trying to regulate using a certain reference optimization parameter set. This index is used to quantify the effectiveness of the regulation strategy. By comparing the regulation effect loss of different parameter sets, the effect of each parameter set in actual application can be evaluated, and the optimal parameter combination can be selected.

[0038] That is, assuming that the target implantable neural regulation state label is "relieve Parkinson's symptoms-hand tremor". The server will simulate the regulation and optimization of the target implant sensor signal according to the five reference optimization parameter sets.

[0039] For reference optimization parameter set 1, the server processes the target implant sensor signal according to a stimulation frequency of 50Hz and a stimulation intensity of 2mA. Then, by comparing with the preset ideal effect, the regulation effect loss under this parameter set is evaluated. For example, it may be found that although the hand tremor is reduced to some extent, there is still obvious shaking, and after complex calculation and analysis, it is concluded that the regulation effect loss is 30%.

[0040] For reference optimization parameter set 2, the processing is carried out at a stimulation frequency of 100Hz and a stimulation intensity of 0.5mA. It may be found that this parameter combination has little effect on reducing hand tremor, and even other symptoms of the patient appear. Calculation shows that the regulation effect loss is 60%.

[0041] Similarly, the five reference optimization parameter sets are processed and evaluated, and finally five regulation effect losses are generated.

[0042] Step S130, based on the X regulation effect losses, extracting the target optimization parameter set of the implant sensor regulation signal data under the target implantable neural regulation state label from the X reference optimization parameter sets.

[0043] In this embodiment, the server compares the five generated regulation effect losses. Assuming that the regulation effect losses are: reference optimization parameter set 1-30%, reference optimization parameter set 2-60%, reference optimization parameter set 3-25%, reference optimization parameter set 4-40%, and reference optimization parameter set 5-18%.

[0044] Since the goal is to find the parameter set with the smallest loss, it can be seen that the regulation effect loss of reference optimization parameter set 5 is the smallest. Therefore, the server extracts reference optimization parameter set 5 as the target optimization parameter set under the target implantable neural regulation state label "relieve Parkinson's symptoms-hand tremor".

[0045] Step S140, based on the target optimization parameter group under the target implantable neural regulation state label, the implanted sensor control signal in the implanted sensor control signal data is regulated and optimized to generate a target regulation optimization result of the implanted sensor control signal data, and the target regulation optimization result of the implanted sensor control signal data is used as a training sample for machine learning.

[0046] In this embodiment, after the server obtains the target optimization parameter group (assuming: stimulation frequency 80 Hz, stimulation intensity 1.5 mA), it applies these parameters to the actual implant sensor control signal data.

[0047] After a period of optimization and control, the server again collected the patient's brain wave data and physiological indicators related to hand tremors. Comparing and analyzing the data with the previous status, it was found that the hand tremors had been significantly alleviated, almost to normal.

[0048] Ultimately, the server determines that this control result is the target control optimization result of the implanted sensor control signal data, and records the relevant data and results for subsequent tracking and evaluation, while providing a reference for possible further optimization.

[0049] The above is just a simple scenario example. In actual applications, the implanted sensor control signal data and reference optimization parameter group will be more complex, involving more physiological indicators and parameter settings. The server needs to perform a lot of calculations and analysis to achieve precise control optimization.

[0050] Assume in another scenario that an implanted sensor is used to adjust the operating parameters of a pacemaker.

[0051] In step S110, the server obtains implanted sensor control signal data to be optimized, including information such as heart rate, intensity, and rhythm. The eight reference optimization parameter groups (X = 8) obtained cover different pacing frequency ranges (e.g., 60 to 120 beats / minute) and pacing intensity settings (e.g., 1 to 5 volts).

[0052] For example, reference optimized parameter group 1 may have a pacing rate of 70 beats / minute and a pacing intensity of 2 volts, while reference optimized parameter group 2 may have a pacing rate of 100 beats / minute and a pacing intensity of 4 volts. The pacing rate and pacing intensity combinations of certain parameter groups may exhibit an unbalanced correlation, such as reference optimized parameter group 3 having a pacing rate of 60 beats / minute and a pacing intensity of 5 volts.

[0053] In step S120, the target implantable neuromodulation state label is "improving heart failure symptoms". The server simulates the target implantable sensor modulation signal according to the eight reference optimization parameter groups respectively. For each parameter group, the server analyzes its improvement effect on heart function, for example, by monitoring indicators such as cardiac output, blood pressure, myocardial oxygen consumption, etc. to evaluate the modulation effect loss.

[0054] Taking reference optimization parameter group 1 as an example, after the server applies a pacing frequency of 70 times / minute and a pacing intensity of 2 volts for modulation, it is found that the cardiac output increases, but the blood pressure does not improve significantly, and the myocardial oxygen consumption slightly increases. The comprehensive calculation obtains a modulation effect loss of 25%.

[0055] For reference optimization parameter group 3, after applying a pacing frequency of 60 times / minute and a pacing intensity of 5 volts for modulation, myocardial overstimulation may occur, and the heart function indicators may even deteriorate. The calculation obtains a modulation effect loss of 55%.

[0056] After evaluating the eight reference optimization parameter groups, eight modulation effect losses are generated.

[0057] In step S130, the server compares the eight modulation effect losses. Suppose the modulation effect losses are: reference optimization parameter group 1-25%, reference optimization parameter group 2-30%, reference optimization parameter group 3-55%, reference optimization parameter group 4-20%, reference optimization parameter group 5-28%, reference optimization parameter group 6-35%, reference optimization parameter group 7-18%, and reference optimization parameter group 8-22%. Obviously, the modulation effect loss of reference optimization parameter group 7 is the smallest.

[0058] In step S140, the server uses the parameters of reference optimization parameter group 7 (supposed to be a pacing frequency of 85 times / minute and a pacing intensity of 3 volts) to actually modulate and optimize the implantable sensor modulation signal. After a period of time, through detection of the patient's heart function indicators, such as a significant increase in ejection fraction, the heart failure symptoms are significantly reduced, and it is determined that this modulation result is the target modulation optimization result.

[0059] Based on the above steps, the embodiments of the present application significantly improve the accuracy and efficiency of implant sensor regulation optimization. By introducing the unbalanced associated reference optimization parameter group, the parameter search space is expanded, and the potential optimization scheme previously ignored is successfully explored and applied, thereby effectively reducing the regulation effect loss. Specifically, by considering a variety of balanced and unbalanced reference optimization parameter combinations, individualized regulation can be performed according to the physiological characteristics and disease states of different patients, significantly improving the treatment effect. The results of each regulation optimization are used as training samples for machine learning, enabling online learning and updating of machine learning, continuously improving its prediction accuracy and generalization ability, so that it can better adapt to changes in the physiological state of the patient, thereby providing a scientific and quantitative basis for regulation optimization and improving the automation and intelligence level of the entire optimization process, opening up new possibilities for the application of implantable medical devices.

[0060] In a possible implementation, the X reference optimization parameter groups include a first reference optimization parameter group and a second reference optimization parameter group, and the reference optimization parameters in the first reference optimization parameter group are in unbalanced association. The reference optimization parameter pair in the first reference optimization parameter group is determined based on different reference optimization parameters in the same reference optimization parameter range. Alternatively, the reference optimization parameter pair in the first reference optimization parameter group is determined based on reference optimization parameters in different reference optimization parameter ranges, and the optimization parameters in different reference optimization parameter ranges are different.

[0061] The reference optimization parameter pair in the second reference optimization parameter group is in balanced association, and the reference optimization parameter pair in the second reference optimization parameter group is determined based on the same reference optimization parameter in the same reference optimization parameter range.

[0062] In this embodiment, the scenario of the implant sensor system processing the brain waves of Parkinson's patients is taken as an example.

[0063] The first reference optimization parameter group obtained by the server is: a stimulation frequency of 120 Hz and a stimulation intensity of 0.8 mA. In this group, the stimulation frequency is high and the stimulation intensity is low, and the optimization parameter pair is in unbalanced association. This imbalance may be determined based on different reference optimization parameter ranges. For example, in order to target the hand tremor symptoms of some Parkinson's patients in a certain time period, a stimulation frequency of 120 Hz is selected from the high frequency stimulation range, and a stimulation intensity of 0.8 mA is selected from the lower stimulation intensity range, expecting to more effectively suppress the abnormal hand tremor through this unbalanced combination.

[0064] The second reference optimization parameter set is: stimulation frequency 70 Hz, stimulation intensity 1.5 mA. In this set, the parameter pair of stimulation frequency and stimulation intensity presents a balanced correlation. They are determined based on the common values in the same reference optimization parameter range. For example, for most Parkinson's patients in the relatively stable stage, this balanced parameter setting can more smoothly regulate brain waves and reduce the symptoms of hand tremors, while reducing possible side effects.

[0065] In another scenario about the adjustment of the working parameters of a cardiac pacemaker.

[0066] The first reference optimization parameter set is: pacing frequency 110 beats per minute, pacing intensity 1.5 volts. The parameter pair presents an unbalanced correlation, which may be based on a specific experimental study. For patients with more serious heart failure symptoms but limited myocardial tolerance, 110 beats per minute is selected from a higher pacing frequency range, and 1.5 volts is selected from a lower pacing intensity range, in an attempt to enhance cardiac pumping function without excessive stimulation of the myocardium.

[0067] The second reference optimization parameter set is: pacing frequency 80 beats per minute, pacing intensity 3 volts. The parameter pair presents a balanced correlation, which is based on typical values in the common parameter range of a cardiac pacemaker. It is suitable for general heart failure patients, providing sufficient pacing support to improve cardiac function while ensuring normal cardiac rhythm.

[0068] Through the above examples in specific scenarios, it can be more intuitively understood how the unbalanced correlation of the first reference optimization parameter set and the balanced correlation of the second reference optimization parameter set are determined according to different conditions and treatment needs.

[0069] In a possible implementation, step S110 includes: Step S111, determining the control effect loss required for control optimization of the target implanted sensor control signal according to the balanced optimization parameter set Ya under the target implanted neuromodulation state label. The balanced optimization parameter set Ya belongs to Y balanced optimization parameter sets, the balanced optimization parameter sets in the Y balanced optimization parameter sets cover the parameter pairs that present a balanced correlation, Y is a positive integer greater than 1, and a is a positive integer not less than Y.

[0070] Step S112, if the Y control effect losses corresponding to the Y balanced optimization parameter sets are obtained, extracting the basic optimization parameter set of the target implanted sensor control signal under the target implanted neuromodulation state label from the Y balanced optimization parameter sets based on the Y control effect losses.

[0071] Step S113, based on the variation amplitude in the variation amplitude sequence, updating the basic optimization parameter set to generate the X reference optimization parameter sets.

[0072] Still taking the implanted sensor system for processing the brain waves of Parkinson's patients as an example.

[0073] The server first obtains 3 balanced optimization parameter sets (Y=3), which are: Balanced optimization parameter set 1: stimulation frequency 60Hz, stimulation intensity 1.5mA Balanced optimization parameter set 2: stimulation frequency 70Hz, stimulation intensity 2mA Balanced optimization parameter set 3: stimulation frequency 80Hz, stimulation intensity 2.5mA Under the target implanted neural regulation state label of "relieving Parkinson's symptoms-hand tremor", the server regulates and optimizes the target implanted sensor regulation signal according to the 3 balanced optimization parameter sets.

[0074] For balanced optimization parameter set 1, the server processes the target implanted sensor regulation signal according to the stimulation frequency of 60Hz and the stimulation intensity of 1.5mA. Through comparison with the preset ideal effect and complex calculation and analysis, it is concluded that the regulation effect loss under this parameter set is 40%.

[0075] For balanced optimization parameter set 2, process with stimulation frequency of 70Hz and stimulation intensity of 2mA, calculate the regulation effect loss as 30%.

[0076] For balanced optimization parameter set 3, process according to the stimulation frequency of 80Hz and the stimulation intensity of 2.5mA, and obtain the regulation effect loss of 25%.

[0077] After the server obtains the 3 regulation effect losses, it compares them. Since the regulation effect loss of balanced optimization parameter set 3 is the smallest, balanced optimization parameter set 3 is extracted from the 3 balanced optimization parameter sets as the basic optimization parameter set of the target implanted sensor regulation signal under the target implanted neural regulation state label of "relieving Parkinson's symptoms-hand tremor".

[0078] Suppose the variation amplitude sequence is: the stimulation frequency variation amplitude is ±10Hz, and the stimulation intensity variation amplitude is ±0.5mA.

[0079] Based on this variation amplitude, the basic optimization parameter set (stimulation frequency 80Hz, stimulation intensity 2.5mA) is updated to obtain a new reference optimization parameter set.

[0080] For example, the generated new reference optimization parameter set 1 is: stimulation frequency 90 Hz, stimulation intensity 2 mA; the new reference optimization parameter set 2 is: stimulation frequency 70 Hz, stimulation intensity 3 mA; the new reference optimization parameter set 3 is: stimulation frequency 80 Hz, stimulation intensity 2 mA, and so on, and finally X reference optimization parameter sets are generated.

[0081] In the scene of heart pacemaker working parameter adjustment: The server obtains 4 balanced optimization parameter sets (Y = 4): Balanced optimization parameter set 1: pacing frequency 75 times / minute, pacing strength 2.5 volts Balanced optimization parameter set 2: pacing frequency 85 times / minute, pacing strength 3 volts Balanced optimization parameter set 3: pacing frequency 95 times / minute, pacing strength 3.5 volts Balanced optimization parameter set 4: pacing frequency 105 times / minute, pacing strength 4 volts Under the target implantable nerve regulation state label of "improving heart failure symptoms", the server performs regulation optimization processing and obtains the corresponding regulation effect loss.

[0082] For example, the regulation effect loss of the balanced optimization parameter set 1 is 35%, the regulation effect loss of the balanced optimization parameter set 2 is 28%, the regulation effect loss of the balanced optimization parameter set 3 is 22%, and the regulation effect loss of the balanced optimization parameter set 4 is 25%.

[0083] After comparison, the balanced optimization parameter set 3 with the smallest regulation effect loss is extracted as the basic optimization parameter set.

[0084] Suppose the variation range is: the pacing frequency variation range is ±5 times / minute, and the pacing strength variation range is ±0.5 volts.

[0085] Based on this variation range, the basic optimization parameter set (pacing frequency 95 times / minute, pacing strength 3.5 volts) is updated to generate new reference optimization parameter sets, such as pacing frequency 100 times / minute, pacing strength 3 volts; pacing frequency 90 times / minute, pacing strength 4 volts, and so on, thereby obtaining X reference optimization parameter sets.

[0086] In one possible implementation, the X reference optimization parameter sets include a third reference optimization parameter set. The reference optimization parameters in the third reference optimization parameter set are unevenly associated. One reference optimization parameter in the third reference optimization parameter set is obtained by updating a first basic optimization parameter in the basic optimization parameter set based on a first variation range in the variation range sequence. Another reference optimization parameter in the third reference optimization parameter set is obtained by updating a second basic optimization parameter in the basic optimization parameter set based on a second variation range in the variation range sequence.

[0087] Still taking the implanted sensor system for processing brain waves of Parkinson's patients as an example.

[0088] Suppose the basic optimization parameter group has been determined through the foregoing steps as: stimulation frequency 80 Hz, stimulation intensity 2.5 mA. The variation range sequence is: stimulation frequency variation range is ±10 Hz, stimulation intensity variation range is ±0.5 mA.

[0089] Among the X generated reference optimization parameter groups, one is a third reference optimization parameter group. For example, the third reference optimization parameter group is: stimulation frequency 95 Hz, stimulation intensity 2 mA.

[0090] In this third reference optimization parameter group, the stimulation frequency 95 Hz is obtained by updating the first basic optimization parameter (stimulation frequency 80 Hz) in the basic optimization parameter group based on the first variation range (+15 Hz) in the variation range sequence; the stimulation intensity 2 mA is obtained by updating the second basic optimization parameter (stimulation intensity 2.5 mA) in the basic optimization parameter group based on the second variation range (-0.5 mA) in the variation range sequence.

[0091] In the scenario of heart pacemaker working parameter adjustment.

[0092] The basic optimization parameter group is: pacing frequency 90 beats / minute, pacing intensity 3 volts. The variation range is: pacing frequency variation range is ±5 beats / minute, pacing intensity variation range is ±0.5 volts.

[0093] The third reference optimization parameter group among the X generated reference optimization parameter groups can be: pacing frequency 85 beats / minute, pacing intensity 3.5 volts.

[0094] Among them, the pacing frequency 85 beats / minute is obtained by updating the first basic optimization parameter (pacing frequency 90 beats / minute) in the basic optimization parameter group based on the first variation range (-5 beats / minute) in the variation range sequence; the pacing intensity 3.5 volts is obtained by updating the second basic optimization parameter (pacing intensity 3 volts) in the basic optimization parameter group based on the second variation range (+0.5 volts) in the variation range sequence.

[0095] In a possible implementation, the X reference optimization parameter groups further include a fourth reference optimization parameter group. The reference optimization parameters in the fourth reference optimization parameter group present balanced correlation. One reference optimization parameter in the fourth reference optimization parameter group is obtained by updating the first basic optimization parameter based on a third variation range in the variation range sequence. Another reference optimization parameter in the fourth reference optimization parameter group is obtained by updating the second basic optimization parameter based on the opposite number of the third variation range.

[0096] In the scenario of processing the brain waves of Parkinson's patients by the implanted sensor system: Suppose the basic optimization parameter group is: stimulation frequency 80Hz, stimulation intensity 2.5mA, and the variable amplitude sequence is: stimulation frequency variable amplitude is ±10Hz, and stimulation intensity variable amplitude is ±0.5mA.

[0097] The fourth reference optimization parameter group among the generated X reference optimization parameter groups is, for example: stimulation frequency 90Hz, stimulation intensity 2mA.

[0098] In this fourth reference optimization parameter group, the stimulation frequency 90Hz is obtained by updating the first basic optimization parameter (stimulation frequency 80Hz) in the basic optimization parameter group based on the third variable amplitude (+10Hz) in the variable amplitude sequence; and the stimulation intensity 2mA is obtained by updating the second basic optimization parameter (stimulation intensity 2.5mA) in the basic optimization parameter group based on the opposite number (-0.5mA corresponding to -10Hz) of the third variable amplitude (+10Hz).

[0099] In the scenario of adjusting the working parameters of a cardiac pacemaker: The basic optimization parameter group is: pacing frequency 90 times / minute, pacing intensity 3 volts, and the variable amplitude is: pacing frequency variable amplitude is ±5 times / minute, and pacing intensity variable amplitude is ±0.5 volts.

[0100] The fourth reference optimization parameter group among the generated X reference optimization parameter groups can be: pacing frequency 95 times / minute, pacing intensity 2.5 volts.

[0101] Among them, the pacing frequency 95 times / minute is obtained by updating the first basic optimization parameter (pacing frequency 90 times / minute) in the basic optimization parameter group based on the third variable amplitude (+5 times / minute) in the variable amplitude sequence; and the pacing intensity 2.5 volts is obtained by updating the second basic optimization parameter (pacing intensity 3 volts) in the basic optimization parameter group based on the opposite number (-0.5 volts corresponding to -5 times / minute) of the third variable amplitude (+5 times / minute).

[0102] In a possible implementation, the step S120 includes: Step S121, acquiring a target regulation signal node in the target implanted sensor regulation signal, and determining a candidate signal node of the target regulation signal node under the target implanted neural regulation state label from the target implanted sensor regulation signal.

[0103] Step S122, determining the regulation deviation degree between the regulation parameter of the target regulation signal node and the regulation parameter of the candidate signal node.

[0104] Step S123, the control deviation is compared with the X reference optimization parameter groups respectively, and the X reference optimization parameter groups correspondingly generated control information respectively.

[0105] Step S124, based on the control information corresponding to the reference optimization parameter group Yb, the target control signal node corresponding to the control optimization knowledge point under the reference optimization parameter group Yb is determined. The reference optimization parameter group Yb belongs to the X reference optimization parameter groups, and b is a positive integer not greater than X, until the target control signal node corresponding to the control optimization knowledge point under the X reference optimization parameter groups is obtained.

[0106] Step S125, based on the X control optimization knowledge points obtained, the control parameters of the target control signal node are respectively controlled and optimized to generate X control optimization results.

[0107] Step S126, based on the X control optimization results, the control effect loss required for the target implanted sensor control signal to be controlled and optimized according to the X reference optimization parameter groups under the target implanted neural control state label is determined, and X control effect losses are generated.

[0108] Still taking the implanted sensor system for processing the brain waves of Parkinson's patients as an example. The target implanted neural control state label is "relieve Parkinson's symptoms-hand tremor".

[0109] The server obtains the target control signal node in the target implanted sensor control signal, such as the brain wave frequency peak value in a certain time period. From the target implanted sensor control signal, the candidate signal node of this target control signal node under the target implanted neural control state label is determined, such as the brain wave frequency peak value in other similar time periods under the same symptom label.

[0110] The control deviation between the control parameters (such as frequency value) of the target control signal node (brain wave frequency peak value in a certain time period) and the control parameters of the candidate signal node (brain wave frequency peak value in other similar time periods) is determined. Assuming that the frequency of the target control signal node is 10Hz, the frequency of the candidate signal node is 8Hz, and the control deviation is 2Hz.

[0111] The control deviation is compared with the X reference optimization parameter groups respectively, and the X reference optimization parameter groups correspondingly generated control information respectively.

[0112] For the reference optimization parameter group 1, the corresponding control information is generated by comparing the parameter range and the setting standard.

[0113] Based on the reference optimization parameter group 1 corresponding to the control information, the target control signal node corresponding to the control optimization knowledge point under the reference optimization parameter group 1 is determined. For example, the ideal frequency range set by the reference optimization parameter group 1 is 8-12Hz, and the current control deviation is 2Hz. Within the range, the corresponding control optimization knowledge point may be to maintain the current stimulation parameters.

[0114] In the same way, the target control signal node corresponding to the control optimization knowledge point under other reference optimization parameter groups is determined respectively, until all control optimization knowledge points of the target control signal node under X reference optimization parameter groups are obtained.

[0115] Based on the X control optimization knowledge points obtained, the control parameters of the target control signal node are respectively processed for control optimization. For example, the control optimization knowledge point corresponding to the reference optimization parameter group 1 is to keep unchanged, so no adjustment is made; the control optimization knowledge point corresponding to the reference optimization parameter group 2 is to increase the stimulation intensity by 0.5mA, so the corresponding parameter adjustment is made to generate X control optimization results.

[0116] Based on the X control optimization results, the control effect loss required for the target implanted sensor control signal to be controlled and optimized under the target implanted neural control state label of "relieving Parkinson's symptoms-hand tremor" according to the X reference optimization parameter groups is determined. For example, a control optimization result makes the hand tremor reduce by 80%, but there is still slight shaking. After complex calculation, it is concluded that the control effect loss of this reference optimization parameter group is 20%. In turn, X control effect losses are calculated.

[0117] In the scene of heart pacemaker working parameter adjustment, the target implanted neural control state label is "improve heart failure symptoms". The server obtains the target control signal node, such as the heart beat interval time at a certain time. The candidate signal nodes under this state label are determined, such as the heart beat interval time at other similar times. The control deviation of the target control signal node and the candidate signal nodes is calculated. The control deviation is compared with each reference optimization parameter group to generate control information, and the corresponding control optimization knowledge point under each reference optimization parameter group is determined. Based on the control optimization knowledge point, the control optimization processing is performed to generate the control optimization result. Finally, the control effect loss of each reference optimization parameter group is determined according to the control optimization result, and X control effect losses are generated.

[0118] In a possible implementation, the candidate signal nodes include a first candidate signal node and a second candidate signal node, and the regulation deviation includes a first regulation deviation and a second regulation deviation. The first regulation deviation is a regulation deviation between a regulation parameter of the first candidate signal node and a regulation parameter of the target regulation signal node, and the second regulation deviation is a regulation deviation between a regulation parameter of the second candidate signal node and the regulation parameter of the target regulation signal node.

[0119] Step S123 includes: Step S1231: Obtain the number of regulation optimization knowledge points under the target implantable neural regulation state label, and obtain first comparison information between the first regulation deviation and the reference optimization parameter group Yb based on the number of regulation optimization knowledge points under the target implantable neural regulation state label.

[0120] Step S1232: Obtain second comparison information between the second regulation deviation and the reference optimization parameter group Yb based on the number of regulation optimization knowledge points under the target implantable neural regulation state label.

[0121] Step S1233: Take the first comparison information and the second comparison information as comparison information corresponding to the reference optimization parameter group Yb, and obtain comparison information corresponding to the X reference optimization parameter groups.

[0122] Still taking the implanted sensor system of the Parkinson's patient's brain wave as an example, the target implantable neural regulation state label is "relieve Parkinson's symptoms-hand tremor".

[0123] The target regulation signal node obtained by the server is the brain wave frequency peak at a certain moment, the first candidate signal node is the brain wave frequency peak at the same moment the day before, and the second candidate signal node is the brain wave frequency peak at the same moment last week.

[0124] Suppose the frequency of the target regulation signal node is 12 Hz, the frequency of the first candidate signal node is 10 Hz, and the frequency of the second candidate signal node is 15 Hz. Then the first regulation deviation is 2 Hz (12 Hz-10 Hz), and the second regulation deviation is -3 Hz (12 Hz-15 Hz).

[0125] The server obtains the number of regulation optimization knowledge points under the target implantable neural regulation state label "relieve Parkinson's symptoms-hand tremor", for example, there are 5.

[0126] Based on this number, the first control information between the first control deviation and the reference optimization parameter set 1 is obtained. Assuming that the acceptable deviation range specified by the reference optimization parameter set 1 is ±1 Hz, since the first control deviation is 2 Hz, which exceeds the range, the first control information may indicate that the deviation is too large.

[0127] Similarly, based on the number of control optimization knowledge points, the second control deviation and the reference optimization parameter set 1 are obtained. Since the second control deviation is -3 Hz, which also exceeds the acceptable deviation range, the second control information also indicates that the deviation is too large.

[0128] The first control information and the second control information are taken as the corresponding control information of the reference optimization parameter set 1.

[0129] In the same way, the first control deviation and the second control deviation are obtained, and the corresponding control information of the other reference optimization parameter sets (such as reference optimization parameter sets 2, 3, etc.) is obtained, until the corresponding control information of X reference optimization parameter sets is obtained.

[0130] In the scene of adjusting the working parameters of the cardiac pacemaker, the target implantable neural control state label is "improve heart failure symptoms".

[0131] The target control signal node obtained by the server is the cardiac pacing interval time at a certain moment, the first candidate signal node is the cardiac pacing interval time at the same moment one hour ago, and the second candidate signal node is the cardiac pacing interval time at the same moment two days ago.

[0132] Assuming that the pacing interval time of the target control signal node is 0.8 seconds, the first candidate signal node is 0.9 seconds, and the second candidate signal node is 0.7 seconds. Then the first control deviation is -0.1 seconds, and the second control deviation is 0.1 seconds.

[0133] The server obtains the number of control optimization knowledge points under the target implantable neural control state label "improve heart failure symptoms", which is 4, for example.

[0134] Based on this number, the first control information between the first control deviation and the reference optimization parameter set 1 is obtained. Assuming that the acceptable deviation range specified by the reference optimization parameter set 1 is ±0.05 seconds, the first control deviation exceeds the range, and the first control information shows that the deviation is too large.

[0135] The second control information between the second control deviation and the reference optimization parameter set 1 is obtained, and since it is within the range, the second control information shows that it is within the acceptable range.

[0136] The two control information is taken as the corresponding control information of the reference optimization parameter set 1, and then the control information of other reference optimization parameter sets is obtained.

[0137] In the implantation of a sensor for regulating sleep apnea patients, the target implantable neuroregulation state label is "improve sleep apnea symptoms".

[0138] The target regulation signal node obtained by the server is the number of apneas in a certain time period, the first candidate signal node is the number of apneas in the same time period last week, and the second candidate signal node is the number of apneas in the same time period last month.

[0139] Suppose the number of apneas of the target regulation signal node is 8, the first candidate signal node is 10, and the second candidate signal node is 12. Then the first regulation deviation is -2, and the second regulation deviation is -4.

[0140] The server obtains the number of regulation optimization knowledge points under the target implantable neuroregulation state label "improve sleep apnea symptoms", for example, there are 6.

[0141] Based on this number, the first control information between the first regulation deviation and the first reference optimization parameter set is obtained. Suppose the acceptable deviation range specified by the first reference optimization parameter set is ±3, and since the first regulation deviation is -2, it is within the range, so the first control information may indicate that the deviation is within the acceptable range.

[0142] Similarly, based on the number of regulation optimization knowledge points, the second control information between the second regulation deviation and the first reference optimization parameter set is obtained. Because the second regulation deviation is -4, it exceeds the acceptable deviation range, so the second control information indicates that the deviation is too large.

[0143] The first control information and the second control information are used as the control information corresponding to the first reference optimization parameter set.

[0144] In the same way, the control information corresponding to the first regulation deviation and the second regulation deviation and other reference optimization parameter sets respectively is obtained, until the control information corresponding to X reference optimization parameter sets is obtained.

[0145] In one possible implementation, step S124 includes: Step S1241, based on the first control information, determining the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0146] Step S1242, based on the second control information, determining the second knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0147] Step S1243: Determine the control optimization knowledge point of the target control signal node under the reference optimization parameter group Yb based on the first knowledge point mapping value and the second knowledge point mapping value.

[0148] In this embodiment, taking the implantable sensor system of Parkinson's patients' brain waves as an example, the target implantable neural regulation state label is "relieving Parkinson's symptoms - hand tremor".

[0149] Assuming reference optimization parameter set 1 (Yb=1), the first comparison information indicates that the deviation between the target control signal node and the first candidate signal node is within an acceptable range. Based on this, the first knowledge point mapping value is determined to be 1. The second comparison information indicates that the deviation between the target control signal node and the second candidate signal node is too large. The second knowledge point mapping value is determined to be 0.

[0150] Then, based on the first knowledge point mapping value 1 and the second knowledge point mapping value 0, the control optimization knowledge point of the target control signal node under the reference optimization parameter group 1 is comprehensively determined to be "appropriately fine-tune the stimulation frequency."

[0151] In the scenario of pacemaker operating parameter adjustment, the target implantable neural regulation state label is "improvement of heart failure symptoms."

[0152] For reference optimization parameter set 2 (Yb=2), the first comparison information shows that the deviation between the target control signal node and the first candidate signal node is too large, and the first knowledge point mapping value is determined to be 0. The second comparison information shows that the deviation is within an acceptable range, and the second knowledge point mapping value is determined to be 1.

[0153] Based on the first knowledge point mapping value 0 and the second knowledge point mapping value 1, it is determined that the control optimization knowledge point of the target control signal node under the reference optimization parameter group 2 is "maintaining the pacing intensity unchanged."

[0154] In the implanted sensor regulation scenario for sleep apnea patients, the target implantable neural regulation state label is "improvement of sleep apnea symptoms."

[0155] For reference optimization parameter group 3 (Yb=3), the first comparison information indicates that the deviation is within the acceptable range, and the first knowledge point mapping value is determined to be 1. The second comparison information also indicates that the deviation is within the acceptable range, and the second knowledge point mapping value is determined to be 1.

[0156] Based on the first knowledge point mapping value 1 and the second knowledge point mapping value 1, it is determined that the control optimization knowledge point of the target control signal node under the reference optimization parameter group 3 is "maintaining the current respiratory stimulation parameters."

[0157] In a possible implementation, the reference optimization parameter set Yb includes a first reference optimization parameter and a second reference optimization parameter, the first reference optimization parameter is smaller than the second reference optimization parameter, and the regulation optimization knowledge point includes four knowledge points.

[0158] Step S1241 includes: If the first comparison information indicates that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is smaller than the first reference optimization parameter, a first mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0159] If the first comparison information indicates that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is not smaller than the first reference optimization parameter, a second mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0160] If the first comparison information indicates that the first regulation deviation degree is not a decrease deviation, and the first regulation deviation degree is smaller than the second reference optimization parameter, a third mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0161] If the first comparison information indicates that the first regulation deviation degree is an increase deviation, and the first regulation deviation degree is not smaller than the second reference optimization parameter, a fourth mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb. The first mapping value, the second mapping value, the third mapping value, and the fourth mapping value are in mapping relationship with the four regulation optimization knowledge points respectively.

[0162] Taking an implanted sensor system of brain waves of a Parkinson's patient as an example, a target implanted neural regulation state label is "relieving Parkinson's symptoms-hand tremor".

[0163] Suppose that the first reference optimization parameter of the reference optimization parameter set 4 (Yb=4) is 5 Hz, and the second reference optimization parameter is 15 Hz.

[0164] The first comparison information indicates that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is 3 Hz which is smaller than the first reference optimization parameter 5 Hz. At this time, a first mapping value (supposed to be 0.1) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set 4.

[0165] If the first comparison information represents that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is not less than 7 Hz and is less than 5 Hz, a second mapping value (assuming 0.2) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 4.

[0166] If the first comparison information represents that the first regulation deviation degree is not a decrease deviation, and the first regulation deviation degree is less than 12 Hz and is greater than 15 Hz, a third mapping value (assuming 0.3) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 4.

[0167] If the first comparison information represents that the first regulation deviation degree is an increase deviation, and the first regulation deviation degree is not less than 18 Hz and is greater than 15 Hz, a fourth mapping value (assuming 0.4) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 4.

[0168] In the scene of heart pacemaker working parameter adjustment, the target implanted neural regulation state label is "improve heart failure symptoms".

[0169] It is assumed that the first reference optimization parameter of the reference optimization parameter group 5 (Yb=5) is 60 times / minute, and the second reference optimization parameter is 90 times / minute.

[0170] The first comparison information represents that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is less than 50 times / minute and is less than 60 times / minute. At this time, a first mapping value (assuming 0.5) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 5.

[0171] If the first comparison information represents that the first regulation deviation degree is a decrease deviation, and the first regulation deviation degree is not less than 65 times / minute and is less than 60 times / minute, a second mapping value (assuming 0.6) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 5.

[0172] If the first comparison information represents that the first regulation deviation degree is not a decrease deviation, and the first regulation deviation degree is less than 80 times / minute and is less than 90 times / minute, a third mapping value (assuming 0.7) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 5.

[0173] If the first comparison information represents that the first regulation deviation degree is an increase deviation, and the first regulation deviation degree is not less than 100 times / minute and is greater than 90 times / minute, a fourth mapping value (assuming 0.8) is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter group 5.

[0174] In a possible implementation, the reference optimization parameter set Yb includes a first reference optimization parameter and a second reference optimization parameter, the first reference optimization parameter is smaller than the second reference optimization parameter, and the regulation optimization knowledge point includes three knowledge points.

[0175] Step S1241 includes: If the first comparison information indicates that the first regulation deviation degree is smaller than the first reference optimization parameter, a fifth mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0176] If the first comparison information indicates that the first regulation deviation degree is greater than the second reference optimization parameter, a sixth mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0177] If the first comparison information indicates that the first regulation deviation degree is equal to the first reference optimization parameter or the first regulation deviation degree is equal to the second reference optimization parameter, a seventh mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb.

[0178] If the first comparison information indicates that the first regulation deviation degree is greater than the first reference optimization parameter and smaller than the second reference optimization parameter, a seventh mapping value is output as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set Yb. The fifth mapping value, the sixth mapping value, and the seventh mapping value are in mapping relationship with the three regulation optimization knowledge points respectively.

[0179] In this embodiment, it is assumed that in the scenario of processing brain waves of Parkinson's patients, the first reference optimization parameter of the reference optimization parameter set 6 (Yb=6) is 10 Hz, and the second reference optimization parameter is 18 Hz.

[0180] The first comparison information indicates that the first regulation deviation degree is 8 Hz, which is smaller than the first reference optimization parameter 10 Hz. At this time, the server outputs the fifth mapping value (assuming 0.5) as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set 6.

[0181] If the first comparison information indicates that the first regulation deviation degree is 20 Hz, which is greater than the second reference optimization parameter 18 Hz, the server outputs the sixth mapping value (assuming 0.7) as the first knowledge point mapping value of the target regulation signal node under the reference optimization parameter set 6.

[0182] If the first comparison information indicates that the first control deviation degree is equal to 10 Hz (i.e., the first reference optimization parameter) or equal to 18 Hz (i.e., the second reference optimization parameter), the server outputs the seventh mapping value (assuming 0.6) as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group 6.

[0183] If the first comparison information indicates that the first control deviation degree is 15 Hz, greater than the first reference optimization parameter 10 Hz and less than the second reference optimization parameter 18 Hz, the server also outputs the seventh mapping value 0.6 as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group 6.

[0184] For example, in a possible implementation, the step S125 includes: Step S1251, querying from the control optimization amount template, the control optimization amount corresponding to each of the X control optimization knowledge points, to generate X control optimization amounts.

[0185] Step S1252, performing control optimization processing on the control parameters of the target control signal node according to the X control optimization amounts, to generate X control optimization results.

[0186] The server has obtained 5 control optimization knowledge points, each corresponding to a reference optimization parameter group. Now, based on these control optimization knowledge points, the control optimization results are to be generated.

[0187] First, the server queries from the control optimization amount template the control optimization amount corresponding to each of the 5 control optimization knowledge points.

[0188] The control optimization knowledge point 1 indicates that the stimulation frequency needs to be increased by 5 Hz, and the server finds in the control optimization amount template that the corresponding control optimization amount is to increase the stimulation frequency by 5 Hz.

[0189] The control optimization knowledge point 2 indicates that the stimulation intensity needs to be reduced by 0.5 mA, and the server finds in the template that the corresponding control optimization amount is to reduce the stimulation intensity by 0.5 mA.

[0190] The control optimization knowledge point 3 indicates that the stimulation frequency needs to be increased by 10 Hz and the stimulation intensity needs to be increased by 0.8 mA, and the server obtains from the template that the control optimization amount is to adjust according to these specific values.

[0191] The control optimization knowledge point 4 indicates that the stimulation frequency needs to remain unchanged, but the stimulation intensity needs to be increased by 1 mA, and the corresponding control optimization amount is to increase the stimulation intensity by 1 mA and keep the stimulation frequency unchanged.

[0192] The fifth regulation optimization knowledge point proposes to reduce the stimulation frequency by 8 Hz and the stimulation intensity by 0.3 mA. The regulation optimization amount found by the server from the template is to make corresponding adjustments according to these values.

[0193] Next, the server performs regulation optimization processing on the regulation parameters of the target regulation signal node according to the five regulation optimization amounts, generating five regulation optimization results.

[0194] For the first regulation optimization amount, the server increases the stimulation frequency of the current target regulation signal node by 5 Hz based on the original value, obtaining the first regulation optimization result. Assuming that the original stimulation frequency is 60 Hz, it becomes 65 Hz after increasing by 5 Hz.

[0195] For the second regulation optimization amount, the server reduces the stimulation intensity of the target regulation signal node by 0.5 mA. If the original stimulation intensity is 2 mA, it becomes 1.5 mA after reduction, thereby generating the second regulation optimization result.

[0196] For the third regulation optimization amount, the server adjusts both the stimulation frequency and the stimulation intensity. For example, the original stimulation frequency is 70 Hz, which becomes 80 Hz after increasing by 10 Hz, and the original stimulation intensity is 1.2 mA, which becomes 2 mA after increasing by 0.8 mA, forming the third regulation optimization result.

[0197] For the fourth regulation optimization amount, the server keeps the stimulation frequency of the target regulation signal node unchanged at 85 Hz, but increases the stimulation intensity by 1 mA. Assuming that the original stimulation intensity is 1.8 mA, it becomes 2.8 mA after increasing, thereby generating the fourth regulation optimization result.

[0198] For the fifth regulation optimization amount, the server reduces the stimulation frequency of the target regulation signal node by 8 Hz and the stimulation intensity by 0.3 mA. If the original stimulation frequency is 90 Hz, it becomes 82 Hz after reduction, and the original stimulation intensity is 2.5 mA, which becomes 2.2 mA after reduction, finally generating the fifth regulation optimization result.

[0199] In this way, the server completes the regulation optimization processing of the regulation parameters of the target regulation signal node based on the five regulation optimization knowledge points and the corresponding regulation optimization amounts, successfully generating five regulation optimization results.

[0200] In this process, the server strictly follows the settings in the regulation optimization amount template and accurately adjusts the regulation parameters in order to achieve the target implantable neural regulation state of "relieving Parkinson's symptoms - hand tremor".

[0201] The server will also further evaluate and analyze these five regulation optimization results. For example, it will observe whether the frequency and amplitude of the patient's hand tremors have decreased, and whether the brain wave activity pattern has become more normal after these regulation optimizations. If a regulation optimization result fails to achieve the expected effect, the server may re-examine the corresponding regulation optimization knowledge points and regulation optimization amounts, or even re-query more appropriate adjustment schemes from the regulation optimization amount template, and perform regulation optimization processing again until the regulation optimization result that can most effectively alleviate the symptoms of Parkinson's disease is found.

[0202] At the same time, the server will also consider the individual differences and real-time physiological responses of the patient. For example, if the patient experiences discomfort or side effects after implementing a certain regulation optimization result, the server will immediately stop the current adjustment and re-evaluate all reference optimization parameter groups and regulation optimization knowledge points to ensure the safety and effectiveness of the regulation optimization process.

[0203] In addition, the server will record each regulation optimization process and result in detail, forming a rich database. This not only helps to provide reference for the subsequent treatment of the current patient, but also accumulates valuable experience and data for the treatment of other Parkinson's patients. Through continuous learning and improvement, the server can provide more and more accurate and efficient implant sensor regulation optimization services for patients.

[0204] Suppose that in another regulation optimization, the server obtains five different regulation optimization knowledge points.

[0205] Regulation optimization knowledge point 1 indicates that the stimulation frequency needs to be significantly increased, such as by 20 Hz. The server queries the corresponding specific increase amount from the regulation optimization amount template, which is 20 Hz.

[0206] Regulation optimization knowledge point 2 suggests a small decrease in stimulation intensity, for example, by 0.2 mA. The server confirms in the template that the corresponding regulation optimization amount for this knowledge point is a decrease of 0.2 mA.

[0207] Regulation optimization knowledge point 3 proposes to significantly increase both the stimulation frequency and the stimulation intensity. The server queries the specific regulation optimization amounts as follows: stimulation frequency increase of 30 Hz and stimulation intensity increase of 1.5 mA.

[0208] Regulation optimization knowledge point 4 indicates that only the stimulation frequency needs to be fine-tuned, with a decrease of 3 Hz. The server obtains the regulation optimization amount from the template as exactly 3 Hz decrease.

[0209] Regulation optimization knowledge point 5 advocates a significant decrease in stimulation intensity, for example, by 1 mA. The server finds the corresponding decrease amount in the regulation optimization amount template as 1 mA.

[0210] Then, the server processes the regulation parameters of the target regulation signal node according to the regulation optimization amount.

[0211] If the initial stimulation frequency is 50Hz, for regulation optimization knowledge point 1, the server increases it by 20Hz to get 70Hz, and generates the corresponding regulation optimization result.

[0212] For regulation optimization knowledge point 2, assuming the original stimulation intensity is 2.5mA, the server reduces it by 0.2mA to get 2.3mA, and forms a new regulation optimization result.

[0213] For regulation optimization knowledge point 3, if the initial stimulation frequency is 60Hz and the stimulation intensity is 1mA, the server increases them respectively to make the stimulation frequency 90Hz and the stimulation intensity 2.5mA, and generates a new result.

[0214] For regulation optimization knowledge point 4, if the original stimulation frequency is 80Hz, the server reduces it by 3Hz to adjust it to 77Hz, and gets the corresponding regulation optimization result.

[0215] For regulation optimization knowledge point 5, if the initial stimulation intensity is 3mA, the server reduces it by 1mA to get 2mA, and completes the regulation optimization processing.

[0216] Through the above series of operations, the server generates 5 different regulation optimization results again, and is ready to perform subsequent effect evaluation and analysis to determine which result can most effectively achieve the goal of "relieving Parkinson's symptoms-hand tremor".

[0217] Such regulation optimization process is a dynamic process of continuous trial and optimization. The server will continuously adjust and improve the regulation strategy according to the actual treatment effect and patient feedback to provide the most suitable treatment plan for the patient.

[0218] Thus, in the regulation optimization of the implanted sensor system for processing the brain waves of Parkinson's patients, the regulation parameters of the target regulation signal node are processed according to the regulation optimization amount by accurately querying the regulation optimization amount template, so as to continuously explore the best regulation scheme for the treatment of patients.

[0219] For example, in one possible implementation, step S126 includes: Step S1261, based on the target implanted sensor control signal covered by the regulation optimization result Yb after regulation optimization, and the initial control signal corresponding to the target implanted sensor control signal, determine the regulation error rate of the target implanted sensor control signal under the reference optimization parameter set Yb. The reference optimization parameter set Yb belongs to the X reference optimization parameter sets, b is a positive integer not greater than X, the regulation optimization result Yb is the regulation optimization result obtained by regulating and optimizing the target implanted sensor control signal according to the reference optimization parameter set Yb, and the target implanted sensor control signal is obtained by reconstructing the encoding features of the initial control signal.

[0220] Step S1262, obtain the performance evaluation index of the reference optimization parameter set Yb, and determine the regulation effect loss of the target implanted sensor control signal under the reference optimization parameter set Yb according to the regulation error rate of the target implanted sensor control signal under the reference optimization parameter set Yb and the performance evaluation index of the reference optimization parameter set Yb, until the target implanted sensor control signal under the X reference optimization parameter sets respectively corresponding to the regulation effect loss is obtained, and X regulation effect losses are generated.

[0221] Suppose in an implanted sensor system for monitoring and adjusting the brain waves of Parkinson's patients, 3 regulation optimization results have been obtained, corresponding to 3 reference optimization parameter sets respectively.

[0222] First, for the reference optimization parameter set 1. The obtained regulation optimization result 1 covers the target implanted sensor control signal after regulation optimization. The initial control signal is the brain wave signal of the patient before any optimization processing. By analyzing and comparing the target implanted sensor control signal after regulation optimization and the initial control signal in detail.

[0223] Suppose in the initial control signal, the abnormal frequency fluctuation of the brain waves of Parkinson's patients is relatively obvious, and in the regulation optimization result 1, the abnormal frequency fluctuation is improved to a certain extent, but there is still instability in some period. After complex calculation and comparison, it is determined that the regulation error rate of the target implanted sensor control signal under the reference optimization parameter set 1 is 25%.

[0224] Next, the performance evaluation index of the reference optimization parameter set 1 is obtained. This performance evaluation index is obtained based on a large number of previous experiments and clinical data, and is used to measure the proportional relationship between the expected effect and the actual effect of the parameter set under similar conditions. Suppose the performance evaluation index of the reference optimization parameter set 1 is 0.8.

[0225] Then, according to the regulation error rate of 25% of the target implanted sensor regulation signal under the reference optimization parameter group 1 and the performance evaluation index of 0.8 of the group, the regulation effect loss of the target implanted sensor regulation signal under the reference optimization parameter group 1 is determined by a specific algorithm (for example: regulation effect loss = regulation error rate / performance evaluation index) to be 31.25%.

[0226] Next, for the reference optimization parameter group 2. Similarly, based on the target implanted sensor regulation signal covered by the regulation optimization result 2 and the initial regulation signal, it is found that the stability of the brain wave has been greatly improved, but there are still some small abnormal fluctuations. After accurate calculation, it is determined that the regulation error rate of the target implanted sensor regulation signal under the reference optimization parameter group 2 is 18%.

[0227] The performance evaluation index of the reference optimization parameter group 2 is assumed to be 0.7. According to the regulation error rate of 18% and the performance evaluation index of 0.7, it is calculated that the regulation effect loss of the target implanted sensor regulation signal under the reference optimization parameter group 2 is about 25.71%.

[0228] Finally, for the reference optimization parameter group 3. Analyzing the target implanted sensor regulation signal in the regulation optimization result 3 and the initial regulation signal, the brain wave almost reaches the ideal state of stability, and only occasionally appears very short abnormality. After calculation, the regulation error rate is 8%.

[0229] The performance evaluation index of the reference optimization parameter group 3 is assumed to be 0.9. Therefore, it is calculated that the regulation effect loss of the target implanted sensor regulation signal under the reference optimization parameter group 3 is about 8.89%.

[0230] Through the above steps, the server obtains the regulation effect loss of the target implanted sensor regulation signal under the three reference optimization parameter groups respectively, and generates three regulation effect losses, which are 31.25%, 25.71% and 8.89% respectively. The server will evaluate the effect of each reference optimization parameter group according to these loss values, and select the most suitable optimization scheme for the patient to achieve the target implanted neural regulation state of "relieving Parkinson's symptoms-hand tremor".

[0231] For example, in one possible implementation, step S130 includes: Step S131, based on the X regulation effect losses, extracting a reference optimization parameter group with the smallest regulation effect loss from the X reference optimization parameter groups.

[0232] Step S132, outputting the extracted reference optimization parameter group as the target optimization parameter group of the implanted sensor regulation signal data under the target implanted neural regulation state label.

[0233] Suppose that in an implanted sensor system for monitoring and adjusting the brain waves of a Parkinson's patient, a total of 5 reference optimization parameter groups are obtained, which are reference optimization parameter groups 1 to 5 respectively, and their corresponding regulation effect losses have been calculated.

[0234] After calculation, the regulation effect loss of the reference optimization parameter group 1 is 28%, the regulation effect loss of the reference optimization parameter group 2 is 35%, the regulation effect loss of the reference optimization parameter group 3 is 20%, the regulation effect loss of the reference optimization parameter group 4 is 25%, and the regulation effect loss of the reference optimization parameter group 5 is 15%.

[0235] The server first compares the 5 regulation effect losses. By comparing one by one, the server finds that the regulation effect loss 15% of the reference optimization parameter group 5 is the smallest among the 5 values.

[0236] Then, the server extracts the reference optimization parameter group 5.

[0237] Finally, the server outputs the extracted reference optimization parameter group 5 as the target optimization parameter group of the implanted sensor regulation signal data under the target implanted neural regulation state label of "relieving Parkinson's symptoms - hand tremor".

[0238] This means that in the current evaluation and calculation, the reference optimization parameter group 5 is considered to be the best parameter combination in achieving the "relieving Parkinson's symptoms - hand tremor" target, and subsequent actual regulation optimization operations on the implanted sensor regulation signal data will be based on this target optimization parameter group.

[0239] For example, in a possible implementation, the number of target implanted neural regulation state labels is Z, and Z is a positive integer.

[0240] Step S140 includes: Step S141, based on the target optimization parameter group under the target implanted neural regulation state label Kc, the regulation parameters of the implanted sensor regulation signal of the implanted sensor regulation signal data are regulated and optimized to generate the regulation optimization result of the implanted sensor regulation signal of the implanted sensor regulation signal data under the target implanted neural regulation state label Kc. The target implanted neural regulation state label Kc belongs to Z target implanted neural regulation state labels, and c is a positive integer not greater than Z.

[0241] Step S142, based on the regulation optimization result of the implanted sensor regulation signal of the implanted sensor regulation signal data under the target implanted neural regulation state label Kc, the regulation effect loss of the implanted sensor regulation signal of the implanted sensor regulation signal data under the target implanted neural regulation state label Kc is determined.

[0242] Step S143, if the Z target implantable neuromodulation state labels correspond to Z modulation effect losses of the implant sensor modulation signal data of the implant sensor modulation signal, the minimum modulation effect loss in the Z modulation effect losses corresponds to the modulation optimization result, and the target modulation optimization result of the implant sensor modulation signal data of the implant sensor modulation signal is output.

[0243] Suppose in an implant sensor system for monitoring and adjusting the brain waves of Parkinson's patients, the number of target implantable neuromodulation state labels is 3, which are "relieving Parkinson's symptoms-hand tremor", "improving Parkinson's symptoms-movement retardation", and "reducing Parkinson's symptoms-muscle stiffness".

[0244] First, for the target implantable neuromodulation state label "relieving Parkinson's symptoms-hand tremor". The server performs modulation optimization processing on the modulation parameters of the implant sensor modulation signal data based on the target optimization parameter group under this label, such as a stimulation frequency of 70Hz and a stimulation intensity of 1.8mA. After processing, the modulation optimization result under the "relieving Parkinson's symptoms-hand tremor" state label is generated. Then, the server calculates the modulation effect loss under this state label by comparing the relevant data before and after optimization, such as the frequency and amplitude changes of hand tremor, which is assumed to be 20%.

[0245] Then, for the target implantable neuromodulation state label "improving Parkinson's symptoms-movement retardation". The server performs modulation optimization processing on the implant sensor modulation signal according to its corresponding target optimization parameter group, such as a stimulation frequency of 80Hz and a stimulation intensity of 2mA, to obtain the modulation optimization result under this state label. Then, the server also determines the modulation effect loss under the "improving Parkinson's symptoms-movement retardation" state label through a series of evaluations and calculations, which is assumed to be 15%.

[0246] Finally, for the target implantable neuromodulation state label "reducing Parkinson's symptoms-muscle stiffness". The server processes the implant sensor modulation signal according to its target optimization parameter group, such as a stimulation frequency of 65Hz and a stimulation intensity of 1.5mA, and generates the corresponding modulation optimization result. The server further analyzes and calculates to obtain the modulation effect loss under the "reducing Parkinson's symptoms-muscle stiffness" state label, which is assumed to be 18%.

[0247] The server obtains three control effect losses of the implant sensor control signal data of the implant sensor control signal corresponding to the three target implantable neuromodulation state labels, which are 20%, 15%, and 18% respectively. After comparison, the server finds that 15% is the smallest among the three control effect losses. Therefore, the server outputs the control optimization result of the implant sensor control signal data of the implant sensor control signal as the target control optimization result of the implant sensor control signal under the state label of "improving Parkinson's symptoms - action slowness". This means that in this round of control optimization, the optimization effect for "improving Parkinson's symptoms - action slowness" is the most ideal, and subsequent treatment and adjustment may be more based on this result.

[0248] For example, in a possible implementation, the method further comprises: Step A110, loading the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label to the controller, and the controller is used to determine the target control optimization result of the implant sensor control signal data based on the first optimization parameter and the second optimization parameter in the target optimization parameter group under the target implantable neuromodulation state label.

[0249] For example, in a possible implementation, step A110 comprises: Obtaining the first parameter label code of the first evaluation index of the first optimization parameter in the first reference optimization parameter range, and the second parameter label code of the second evaluation index of the second optimization parameter in the second reference optimization parameter range.

[0250] Loading the first parameter label code corresponding to the first optimization parameter group and the second parameter label code corresponding to the second optimization parameter group to the controller.

[0251] For example, in another possible implementation, step A110 can further comprise: Obtaining the first parameter label code of the first evaluation index of the first optimization parameter in the first reference optimization parameter range, and the second parameter label code of the second evaluation index of the second optimization parameter in the second reference optimization parameter range.

[0252] Obtaining the coding deviation value between the first parameter label code and the second parameter label code.

[0253] Loading the first parameter label code and the coding deviation value to the controller, or loading the second parameter label code and the coding deviation value to the controller.

[0254] For example, in another possible implementation, step A110 can further comprise: A first characterization attribute corresponding to the first evaluation index of the first optimization parameter is generated. The first characterization attribute is used to characterize that the first evaluation index is the evaluation index of the first optimization parameter.

[0255] A second characterization attribute corresponding to the second evaluation index of the second optimization parameter is generated. The second characterization attribute is used to characterize that the second evaluation index is the evaluation index of the second optimization parameter.

[0256] The first evaluation index, the first characterization attribute, the second evaluation index and the second characterization attribute are loaded to the controller.

[0257] For example, in another possible implementation, step A110 can further include: A first characterization attribute corresponding to the first evaluation index of the first optimization parameter group is generated. The first characterization attribute is used to characterize that the first evaluation index is the evaluation index of the first optimization parameter.

[0258] A second characterization attribute corresponding to the second evaluation index of the second optimization parameter group is generated. The second characterization attribute is used to characterize that the second evaluation index is the evaluation index of the second optimization parameter.

[0259] A target loss value between the first evaluation index and the second evaluation index is obtained.

[0260] The first evaluation index, the first characterization attribute and the target loss value are loaded to the controller. Alternatively, the second evaluation index, the second characterization attribute and the target loss value are loaded to the controller.

[0261] In an implanted sensor system for monitoring and adjusting the brain waves of Parkinson's patients, it is assumed that a target optimization parameter group under the target implanted neural regulation state label of "relieving Parkinson's symptoms - hand tremor" has been determined, in which the first optimization parameter is the stimulation frequency 80 Hz, and the second optimization parameter is the stimulation intensity 2.5 mA.

[0262] Firstly, the server obtains the first parameter label code of the first evaluation index (for example, the evaluation of the effect of relieving hand tremor) of the first optimization parameter (stimulation frequency 80 Hz) in the first reference optimization parameter range (for example, the common stimulation frequency range). It is assumed that this code is "0101". At the same time, the second parameter label code of the second evaluation index (for example, the evaluation of the effect of relieving hand tremor) of the second optimization parameter (stimulation intensity 2.5 mA) in the second reference optimization parameter range (for example, the common stimulation intensity range) is obtained, which is assumed to be "1010". Then, "0101" and "1010" are loaded to the controller.

[0263] Alternatively, the server obtains the first parameter tag code "0101" and the second parameter tag code "1010" described above, and calculates the encoding deviation value between the two. Assuming that the encoding deviation value is "0011". Next, "0101" and "0011" are loaded into the controller, or "1010" and "0011" are loaded into the controller.

[0264] Alternatively, the server generates a first characterization attribute corresponding to the first evaluation index of the first optimization parameter, such as "this evaluation index is dedicated to the evaluation of the stimulation frequency 80Hz". At the same time, a second characterization attribute corresponding to the second evaluation index of the second optimization parameter is generated, such as "this evaluation index is dedicated to the evaluation of the stimulation intensity 2.5mA". Then, the first evaluation index, the first characterization attribute, the second evaluation index and the second characterization attribute are loaded into the controller.

[0265] Alternatively, the server generates the first characterization attribute and the second characterization attribute described above, obtains the target loss value between the first evaluation index and the second evaluation index, assuming "0.15". Subsequently, the first evaluation index, the first characterization attribute and the target loss value are loaded into the controller, or the second evaluation index, the second characterization attribute and the target loss value are loaded into the controller.

[0266] Through the above different ways, the server loads the first optimization parameter and the second optimization parameter in the target optimization parameter group into the controller, so that the controller can more accurately determine the target control optimization result of the implanted sensor control signal data based on these parameters and related information, so as to better realize the adjustment of the brain waves of Parkinson's patients and the relief of symptoms.

[0267] Figure 2 The hardware structure of the deep learning system 100 for implementing the machine learning-based implanted sensor control data analysis method provided by the embodiments of the present application is shown, as shown in Figure 2 As shown, the deep learning system 100 can include a processor 110, a machine readable storage medium 120, a bus 130 and a communication unit 140.

[0268] In an alternative embodiment, the deep learning system 100 can be a single server or a group of servers. The group of servers can be centralized or distributed (e.g., the deep learning system 100 can be a distributed system). In an alternative embodiment, the deep learning system 100 can be local or remote. For example, the deep learning system 100 can access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the deep learning system 100 can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In an alternative embodiment, the deep learning system 100 can be implemented on a cloud platform. By way of example only, the cloud platform can include a private cloud, a public cloud, a community cloud, a distributed cloud, an inter-cloud, an intra-cloud, a multi-cloud, or any integration thereof.

[0269] The machine-readable storage medium 120 can store data and / or instructions. In an alternative embodiment, the machine-readable storage medium 120 can store data acquired from an external terminal. In an alternative embodiment, the machine-readable storage medium 120 can store data and / or instructions used by the deep learning system 100 to perform or use to complete the exemplary methods described in the present application. In an alternative embodiment, the machine-readable storage medium 120 can include a mass storage, a removable storage, a volatile read-write memory, a read-only memory, or any integration thereof. Exemplary mass storage can include a magnetic disk, an optical disk, a solid-state disk, or the like. Exemplary removable storage can include a flash drive, a floppy disk, an optical disk, a memory card, a compact disk, a magnetic tape, or the like.

[0270] In a specific implementation process, the plurality of processors 110 executes computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can perform the machine learning-based implanted sensor regulation data analysis method of the method embodiments as described above. The processor 110, the machine-readable storage medium 120, and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to describe the transceiving action of the communication unit 140.

[0271] The specific implementation process of the processor 110 can refer to the implementation principles and technical effects of the various method embodiments executed by the deep learning system 100 as described above, which are similar, and will not be described here in detail.

[0272] In addition, the embodiment of the present application also provides a readable storage medium, in which computer executable instructions are pre-installed. When the processor executes the computer executable instructions, the machine learning-based implanted sensor regulation data analysis method as described above is realized.

[0273] For simplicity of exposition, and to aid in the understanding of one or more embodiments of the application, the foregoing description of embodiments of the application is sometimes divided into sections, each of which can contain multiple features. Similarly, it should be noted that, for simplicity of exposition, and to aid in the understanding of one or more embodiments of the application, the foregoing description of embodiments of the application is sometimes divided into sections, each of which can contain multiple features.

Claims

1. A method for analyzing implant sensor control data based on machine learning, characterized in that: The method comprises: Acquiring implanted sensor control signal data to be optimized and obtaining X reference optimization parameter groups; wherein at least one of the X reference optimization parameter groups has an unbalanced correlation between reference optimization parameter pairs; and X is a positive integer greater than 1; determining, under a target implantable neural control state label, control effect losses required for optimizing the control signal of the target implantable sensor based on the X reference optimization parameter groups, respectively, to generate X control effect losses; the target implantable sensor control signal belonging to the implantable sensor control signal data; Based on the X control effect losses, extracting a target optimization parameter group for the implanted sensor control signal data under the target implantable neural control state label from the X reference optimization parameter groups; Based on the target optimization parameter group under the target implantable neural regulation state label, the implanted sensor control signal in the implanted sensor control signal data is regulated and optimized to generate a target regulation optimization result of the implanted sensor control signal data, and the target regulation optimization result of the implanted sensor control signal data is used as a training sample for machine learning.

2. The implant sensor control data analysis method based on machine learning according to claim 1, characterized in that: The X reference optimization parameter groups include a first reference optimization parameter group and a second reference optimization parameter group, and reference optimization parameter pairs in the first reference optimization parameter group exhibit an unbalanced correlation; wherein the reference optimization parameter pairs in the first reference optimization parameter group are determined based on different reference optimization parameters within the same reference optimization parameter range; or, the reference optimization parameter pairs in the first reference optimization parameter group are determined based on reference optimization parameters within different reference optimization parameter ranges, and the optimization parameters within the different reference optimization parameter ranges are different; The reference optimization parameter pairs in the second reference optimization parameter group are balancedly correlated with each other, wherein the reference optimization parameter pairs in the second reference optimization parameter group are determined based on the same reference optimization parameter within the same reference optimization parameter range.

3. The implant sensor control data analysis method based on machine learning according to claim 1 or 2, characterized in that: The obtaining of X reference optimization parameter groups includes: determining, under the target implantable neural control state label, a control effect loss required for control optimization of the target implantable sensor control signal according to the balanced optimization parameter group Ya; the balanced optimization parameter group Ya belongs to Y balanced optimization parameter groups, and the reference optimization parameter pairs covered by the balanced optimization parameter groups in the Y balanced optimization parameter groups exhibit balanced correlations, where Y is a positive integer greater than 1, and a is a positive integer not less than Y; If Y control effect losses corresponding to the Y balanced optimization parameter groups are obtained, extracting a basic optimization parameter group of the target implanted sensor control signal under the target implanted neural control state label from the Y balanced optimization parameter groups based on the Y control effect losses; Based on the variation range in the variation range sequence, the basic optimization parameter group is updated to generate the X reference optimization parameter groups.

4. The method for analyzing implanted sensor control data based on machine learning according to claim 3, characterized in that: The X reference optimization parameter groups include a third reference optimization parameter group; the reference optimization parameter pairs in the third reference optimization parameter group exhibit an unbalanced correlation; A reference optimization parameter in the third reference optimization parameter group is obtained by updating a first basic optimization parameter in the basic optimization parameter group based on a first variation range in the variation range sequence; Another reference optimization parameter in the third reference optimization parameter group is obtained by updating the second basic optimization parameter in the basic optimization parameter group based on the second variation range in the variation range sequence.

5. The method for analyzing implanted sensor control data based on machine learning according to claim 4, characterized in that: The X reference optimization parameter groups also include a fourth reference optimization parameter group; the reference optimization parameter pairs in the fourth reference optimization parameter group present a balanced correlation; A reference optimization parameter in the fourth reference optimization parameter group is obtained by updating the first basic optimization parameter based on the third variation range in the variation range sequence; Another reference optimization parameter in the fourth reference optimization parameter group is obtained by updating the second basic optimization parameter based on the inverse of the third variation range.

6. The method for analyzing implanted sensor control data based on machine learning according to claim 1 or 2, characterized in that: The determining, under the target implantable neural control state label, control effect losses required for performing control optimization on the target implantable sensor control signal based on the X reference optimization parameter groups, to generate X control effect losses, includes: Acquire a target control signal node in the target implanted sensor control signal, and determine, from the target implanted sensor control signal, a candidate signal node of the target control signal node under the target implanted neural control state label; Determining a control deviation between a control parameter of the target control signal node and a control parameter of the candidate signal node; Comparing the control deviation with the X reference optimization parameter groups respectively to generate comparison information corresponding to the X reference optimization parameter groups respectively; Based on the control information corresponding to the reference optimization parameter group Yb, determining the control optimization knowledge point corresponding to the target control signal node under the reference optimization parameter group Yb; the reference optimization parameter group Yb belongs to the X reference optimization parameter groups, and b is a positive integer not greater than X, until the control optimization knowledge points corresponding to the target control signal node under the X reference optimization parameter groups are obtained; Based on the obtained X control optimization knowledge points, control optimization processing is performed on the control parameters of the target control signal node respectively to generate X control optimization results; Based on the X control optimization results, determining the control effect losses required for performing control optimization on the target implanted sensor control signal according to the X reference optimization parameter groups under the target implanted neural control state label, and generating X control effect losses.

7. The method for analyzing implanted sensor control data based on machine learning according to claim 6, characterized in that: The candidate signal nodes include a first candidate signal node and a second candidate signal node, and the control deviation includes a first control deviation and a second control deviation, the first control deviation being the control deviation between the control parameter of the first candidate signal node and the control parameter of the target control signal node, and the second control deviation being the control deviation between the control parameter of the second candidate signal node and the control parameter of the target control signal node; Comparing the control deviation with the X reference optimization parameter groups respectively to generate comparison information corresponding to the X reference optimization parameter groups respectively includes: Obtaining the number of control optimization knowledge points under the target implantable neural control state label, and obtaining first comparison information between the first control deviation and the reference optimization parameter group Yb based on the number of control optimization knowledge points under the target implantable neural control state label; Based on the number of control optimization knowledge points under the target implantable neural control state label, obtaining second comparison information between the second control deviation and the reference optimization parameter group Yb; Using the first comparison information and the second comparison information as the comparison information corresponding to the reference optimization parameter group Yb, until the comparison information corresponding to the X reference optimization parameter groups is obtained; Furthermore, the step of determining the control optimization knowledge point corresponding to the target control signal node under the reference optimization parameter group Yb based on the control information corresponding to the reference optimization parameter group Yb includes: Determining a first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb based on the first comparison information; Determine, based on the second comparison information, a second knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; Based on the first knowledge point mapping value and the second knowledge point mapping value, the control optimization knowledge point of the target control signal node under the reference optimization parameter group Yb is determined.

8. The method for analyzing implanted sensor control data based on machine learning according to claim 7, characterized in that: The reference optimization parameter group Yb includes a first reference optimization parameter and a second reference optimization parameter, the first reference optimization parameter is smaller than the second reference optimization parameter, and the control optimization knowledge points include four knowledge points; The determining, based on the first comparison information, a first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb includes: If the first control information indicates that the first control deviation is a negative deviation, and the first control deviation is less than a first reference optimization parameter, outputting the first mapping value as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; If the first control information indicates that the first control deviation is a negative deviation, and the first control deviation is not less than the first reference optimization parameter, outputting the second mapping value as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; If the first control information indicates that the first control deviation is not a negative deviation, and the first control deviation is less than the second reference optimization parameter, outputting the third mapping value as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; If the first control information represents that the first control deviation is an increasing deviation, and the first control deviation is not less than the second reference optimization parameter, then the fourth mapping value is output as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; the first mapping value, the second mapping value, the third mapping value and the fourth mapping value have mapping relationships with the four control optimization knowledge points respectively.

9. The method for analyzing implanted sensor control data based on machine learning according to claim 7, characterized in that: The reference optimization parameter group Yb includes a first reference optimization parameter and a second reference optimization parameter, the first reference optimization parameter is smaller than the second reference optimization parameter, and the control optimization knowledge points include three knowledge points; The determining, based on the first comparison information, a first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb includes: If the first control information indicates that the first control deviation is less than the first reference optimization parameter, outputting the fifth mapping value as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; If the first control information indicates that the first control deviation is greater than the second reference optimization parameter, outputting the sixth mapping value as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; If the first control information indicates that the first control deviation is equal to the first reference optimization parameter, or the first control deviation is equal to the second reference optimization parameter, outputting the seventh mapping value as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; If the first control information represents that the first control deviation is greater than the first reference optimization parameter, and the first control deviation is less than the second reference optimization parameter, then the seventh mapping value is output as the first knowledge point mapping value of the target control signal node under the reference optimization parameter group Yb; the fifth mapping value, the sixth mapping value, and the seventh mapping value have mapping relationships with the three control optimization knowledge points respectively.

10. A deep learning system, characterized in that: The deep learning system includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores machine-executable instructions, and the machine-executable instructions are loaded and executed by the processor to implement the machine learning-based implant sensor control data analysis method described in any one of claims 1-9.

Images