Noninvasive internal environment parameter data determination method, system, equipment, medium and product

Abstract

The embodiment of the invention discloses a non-invasive internal environment parameter data determination method, system and device, a medium and a product, and belongs to the technical field of data processing. The method comprises the following steps: acquiring a predetermined physiological signal of a target object, wherein the predetermined physiological signal comprises a target electrocardiosignal and a target photoelectric volume signal under a current time window; the preset physiological signals are input into a pre-training prediction model, the preset electrolyte concentration and / or the preset organism metabolite concentration of the target object are / is obtained, and the prediction model becomes the pre-training prediction model after dictionary-encoder joint training, mask pre-training and downstream task training; a first-stage loss function used by mask pre-training is a weighted sum of a first loss function and a second loss function, the first loss function is a mask prediction loss function, and the second loss function is a physical constraint loss function determined based on a Fizhz-south cloud kinetic equation. According to the embodiment of the invention, the accuracy of determining the non-invasive internal environment parameter data can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to a method, system, device, medium and product for determining non-invasive internal environment parameter data. Background Technology

[0002] Existing methods for monitoring the internal environment include invasive and non-invasive blood gas analysis. The former has significant limitations, such as time-limited monitoring parameters and the increased risk of infection from invasive collection; the latter suffers from lower accuracy. To improve patients' blood gas monitoring experience, there is an urgent need for a non-invasive method for determining internal environment parameters. Summary of the Invention

[0003] This invention provides a method, system, device, medium, and product for determining non-invasive internal environment parameter data, in order to solve the problem that existing methods for determining non-invasive internal environment parameter data have low accuracy in predicting parameter data.

[0004] According to one aspect of the present invention, a method for determining non-invasive internal environment parameter data is provided, the method comprising:

[0005] Acquire predetermined physiological signals of the target object, wherein the predetermined physiological signals include the target electrocardiogram signal and the target photoplethysmogram signal under the current time window;

[0006] The predetermined physiological signal is input into the pre-trained prediction model to obtain the predetermined electrolyte concentration and / or predetermined metabolic product concentration of the target object. After training by dictionary-encoder joint training, mask pre-training and downstream task training, the prediction model becomes the pre-trained prediction model. The first-stage loss function used in the mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is the physical constraint loss function determined based on the Fittshur-Narcoma dynamics equations.

[0007] According to another aspect of the present invention, a non-invasive internal environment parameter monitoring system is provided, comprising:

[0008] The electrocardiogram (ECG) acquisition unit is used to acquire the ECG signals of the target object in real time.

[0009] A photoplethysmography (PPG) acquisition unit is used to acquire the PPG signal of the target object in real time.

[0010] A processor for determining at least one target physiological indicator data of the target object according to the method described in any embodiment.

[0011] According to another aspect of the present invention, a non-invasive internal environment parameter data determination device is provided, the device comprising:

[0012] The acquisition module is used to acquire predetermined physiological signals of the target object, wherein the predetermined physiological signals include the target electrocardiogram signal and the target photoplethysmography signal under the current time window;

[0013] The prediction module is used to input the predetermined physiological signal into the pre-trained prediction model to obtain the predetermined electrolyte concentration and / or predetermined metabolic product concentration of the target object. The prediction model becomes the pre-trained prediction model after joint training of dictionary-encoder, mask pre-training and downstream task training. The first-stage loss function used in the mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is the physical constraint loss function determined based on the Fittshur-Naghui dynamic equations.

[0014] According to another aspect of the present invention, a server device is provided, the server device comprising:

[0015] One or more processors;

[0016] Storage device for storing one or more programs.

[0017] When one or more programs are executed by one or more processors, the one or more processors implement the non-invasive internal environment parameter data determination method as described in any embodiment of the present invention.

[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions for causing a processor to execute and implement the non-invasive internal environment parameter data determination method according to any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer program product is provided, which, when executed by a processor, implements the non-invasive internal environment parameter data determination method as described in any embodiment of the present invention.

[0020] In the technical solution of this invention, the prediction model undergoes sequential training through dictionary-encoder joint training, mask pre-training, and downstream task training to become a pre-trained prediction model. The loss function used in the first stage of mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is a physical constraint loss function determined based on the Fittshur-Namco dynamics equations. Since the Fittshur-Namco dynamics equations are an approximate model of neuronal and myocardial electrophysiology, the combined use of the first loss function and the second loss function determined by the Fittshur-Namco dynamics equations can improve the training effect of mask pre-training, thereby improving the accuracy of the pre-trained prediction model and thus improving the accuracy of its predicted predetermined electrolyte concentration and / or predetermined body metabolite concentration.

[0021] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating the method for determining non-invasive internal environment parameter data provided in an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of the data flow during dictionary-encoder joint training provided in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the structure of the pulse transform coding layer provided in an embodiment of the present invention;

[0026] Figure 4 A schematic diagram of mask pre-training for embedding physical knowledge provided in an embodiment of the present invention;

[0027] Figure 5 This is a schematic diagram of the structure of the non-invasive internal environment parameter monitoring system provided in an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of the non-invasive internal environment parameter data determination device provided in an embodiment of the present invention;

[0029] Figure 7 This is a schematic diagram of the structure of a server device provided in an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] It should be noted that the collection, gathering, updating, analysis, processing, use, transmission, and storage of user personal information involved in the technical solutions provided in the embodiments of the present invention all comply with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken to prevent unauthorized access to user personal information data and to safeguard user personal information security, network security, and national security.

[0033] Figure 1 This is a flowchart illustrating a method for determining non-invasive internal environment parameter data according to an embodiment of the present invention. This embodiment is applicable to situations where electrolyte concentration and / or predetermined concentrations of metabolic products are automatically determined based on electrocardiogram (ECG) signals and photoplethysmography (PPG) signals. The method is executed by a non-invasive internal environment parameter data determining device, which can be implemented in hardware and / or software and can be configured in server equipment such as a computer or server. Figure 1 As shown, the method in this embodiment includes:

[0034] S110. Obtain the predetermined physiological signal of the target object, wherein the predetermined physiological signal includes the target electrocardiogram signal and the target photoplethysmogram signal under the current time window.

[0035] The target object wears a signal acquisition device, which includes an electrocardiogram (ECG) sensor and a photoplethysmography (PPG) sensor. The ECG and PPG signals acquired by the ECG and PPG sensors are divided into segments using a time window, and the ECG and PPG signals under the current time window are used as the target ECG and target PPG signals.

[0036] The time window width must cover at least the duration of a complete cardiac cycle, such as 5 to 30 seconds, to ensure that the target electrocardiogram signal and the target photoplethysmogram signal can reflect stable heart rate and hemodynamic characteristics.

[0037] Specifically, the sampling rate of the ECG signal and the photoplethysmography (PPG) signal can be selected as 500Hz. After acquiring the ECG signal and PPG signal, a bandpass filter of 0.5-40Hz is used to filter the ECG signal and the PPG signal respectively to obtain the filtered ECG signal and the filtered PPG signal; then, the filtered ECG signal and the filtered PPG signal are normalized to obtain the normalized ECG signal and the normalized PPG signal; then, a time window is used to segment the normalized ECG signal and the normalized PPG signal to obtain the target ECG signal and the target PPG signal.

[0038] S120. Input the predetermined physiological signal into the pre-trained prediction model to obtain the predetermined electrolyte concentration and / or predetermined metabolic product concentration of the target object. After the prediction model is trained by dictionary-encoder joint training, mask pre-training and downstream task training, it becomes a pre-trained prediction model. The first-stage loss function used in the mask pre-training is the weighted sum of the first loss function and the second loss function. The first loss function is the mask prediction loss function, and the second loss function is the physical constraint loss function determined based on the Fittshur-Nagun dynamic equation.

[0039] Once the predetermined physiological signals are determined, they are input into a pre-trained prediction model to obtain the predetermined electrolyte concentration and / or predetermined metabolic product concentration of the target subject. The predetermined electrolyte concentration includes some or all of the calcium ion concentration, potassium ion concentration, and chloride ion concentration, while the predetermined metabolic product concentration includes some or all of the lactate concentration and blood glucose concentration, thus improving the flexibility of determining the target non-invasive internal environment parameter data.

[0040] The prediction model includes a pulse feature dictionary and an encoder. After undergoing a dictionary-encoder joint training phase, a pre-training phase, and a downstream task training phase, it becomes a pre-trained prediction model.

[0041] Both dictionary-encoder joint training and pre-training are unsupervised training, while downstream training is considered supervised training.

[0042] The encoder of the prediction model consists of multiple pulse transform coding layers, which are connected in series and stacked. The spiking neurons in the pulse transform coding layers are integral firing neurons.

[0043] Figure 2 This is a schematic diagram of the structure of the spiking transform coding layer provided in an embodiment of the present invention. Specifically, the spiking transform coding layer adopts a spiking transform structure based on a spiking neural network (SNN). Its basic unit consists of a multi-head self-attention layer, residual connections, a fully connected layer (dense layer), and spiking neurons, as shown below. Figure 2 As shown, the sum of the output features of the multi-head self-attention layer and the input features of the multi-head self-attention layer is input to the first fully connected layer. The output features of the first fully connected layer are input to the spiking neuron, and the output features of the spiking neuron are input to the second fully connected layer. The sum of the output features of the second fully connected layer and the input features of the first fully connected layer serves as the output features of the corresponding pulse transform coding layer.

[0044] Among them, the spiking neuron is an integral-and-fire (IF) neuron, which can be represented as:

[0045] ;

[0046] in, The membrane potential of neuron k at time t. It is a film capacitor. It is the emission threshold. It is a reset potential. It is a transmission identifier. When the current... When input is given to neuron k, its membrane potential gradually increases until it reaches a threshold. Simultaneously, the neuron fires a spike, then resets to its normal value. Waiting for the next launch.

[0047] Integral-fire neurons are used to process temporal signals. In the neuron model, they have the functions of membrane potential accumulation and impulse firing for dynamic information integration. This special process includes the following three steps: First, one-dimensional convolution is used to capture local dependencies, and then the signal is nonlinearly normalized to a scalar value between 0 and 1 using the sigmoid function. Second, the current signal will stimulate a neuron to generate an impulse after a period of membrane potential accumulation. For example, at time t in the learning process, the current value in the current sequence will be calculated first. A function This function is based on the current values ​​in the current sequence. And some other variables, such as the membrane potential at the previous time t−1. Then add it to the previous membrane potential. To obtain a new accumulated membrane potential. This process can be represented by the following formula:

[0048] .

[0049] if Less than a given threshold It will still be used as the next step. The previous membrane potential; otherwise, it will be set to This indicates membrane potential leakage due to pulse delivery.

[0050] Define pulse feature dictionary For the feature vector output by the pulse transform encoder Find the closest term index in the pulse feature dictionary to obtain the quantized feature map. . The input signal is fed into the decoder, and the decoder outputs the corresponding reconstructed signal. See Figure 3 The formula for this process is as follows:

[0051] ;

[0052] in, For the encoder input, It is the output of the encoder. It is an atomic vector in the pulse feature dictionary.

[0053] Regarding the dictionary-encoder joint training phase: First training set is obtained, consisting of multiple first training samples, including electrocardiograms and photoplethysmograms within the target time window; the atomic feature vectors in the pulse feature dictionary are initialized; and joint training of the dictionary-encoder is completed based on the second-stage loss function and the first training sample set.

[0054] Specifically, a vector quantization (VQ) mechanism is used to optimize the pulse feature dictionary. During training, each first training sample in the first training set is input into the encoder, and the encoder outputs the feature vector corresponding to the received first sample. Find the closest index in the pulse feature dictionary to obtain the quantized feature map. . The input signal is fed into the decoder, and the decoder outputs the corresponding reconstructed signal. .

[0055] The model minimizes the L2 distance between the impulse feature dictionary vector and the encoder output vector using the L2 norm. Additionally, a regularization term is added to constrain the encoder's parameter updates. The overall loss is represented by the following second-stage loss function:

[0056] ;

[0057] in, For L2 norm, which is the reconstruction loss, used to constrain the error between the encoder output vector and the corresponding atomic feature vector in the pulse feature dictionary; is the first regularization term, and is the pulse feature dictionary loss, used to constrain the distance between the encoder output vector and the corresponding atomic feature vector in the pulse feature dictionary; The second regularization term, representing the commitment loss, is used to constrain the encoder output vector to anchor the corresponding atomic feature vector in the pulse feature dictionary.

[0058] Among them, the first regular direction Continuous features of encoder output The gradient-stopping operation prevents gradients from flowing to the encoder during backpropagation, and is specifically used to constrain the quantized features. Towards Approximation is used to optimize the parameters of the pulse feature dictionary, making the dictionary atoms fit the feature distribution of the encoder output more closely.

[0059] Second regularization Quantized discrete features The gradient stopping operation prevents gradients from flowing to the encoder during backpropagation, and is specifically used to constrain the encoder's output features. Vectorized features Approximation optimizes the encoder's parameters, enabling the encoder to learn and generate features that are more easily quantized by a dictionary.

[0060] β and γ are weighting coefficients used to balance the contributions of the two parts of the loss, and their sizes can be adjusted according to task requirements (such as reconstruction accuracy and quantization efficiency).

[0061] To improve the ability to model the physical consistency between ECG signals and photoplethysmography (PPG) signals, this invention introduces a Physics-Informed Neural Network (PINN) constraint mechanism based on the FitzHugh–Nagumo dynamics equations during mask pre-training, and combines it with a Spiking Transformer encoder to achieve joint optimization of temporal feature reconstruction and physical constraints. Firstly, as... Figure 4As shown, the acquired target sample set is randomly masked. The unmasked portion is input into a pulse transform encoder. The output features of the pulse transform encoder are input into a multilayer perceptron regression head. The multilayer perceptron regression head outputs the predicted token index of the masked portion. This token index is input into a decoder, which outputs the corresponding decoding result. A physical information neural network constraint mechanism is used to determine the reconstructed signal for the masked signal. The training sample set includes multiple training samples, which are predetermined physiological signals. The predicted output is logits, and the true token index is... The mask prediction loss is expressed in the form of cross-entropy:

[0062] ;

[0063] in, The probability distribution is obtained by softmax transformation of the encoder output logits. This loss is used to supervise the model in learning the contextual dependency structure between token indices. Then, the predicted token vectors are mapped into continuous signal representations by the decoder. This is used for physical constraint calculations. Considering the nonlinear excitation-recovery characteristics of cardiomyocyte action potentials, this invention uses the FitzHugh-Nagumo kinetic equation as the physiological constraint basis for the model. This equation is a classic approximate model of neuronal and myocardial electrophysiology, in the following form:

[0064] ;

[0065] in, This represents the continuous signal reconstructed by the decoder; As a recovery variable, it reflects the channel recovery characteristics; The drive signal output by the Spiking Transformer encoder; These are system parameters that can be jointly learned during training.

[0066] Based on this formula, the residual corresponding to the physical information neural network constraint mechanism is defined as follows:

[0067] ;

[0068] Simultaneously, a physical constraint loss based on this physical information neural network constraint mechanism is constructed:

[0069] ;

[0070] The MSE (mean squared error) form is used to minimize the squared mean of the equation residuals, achieving a least-squares fit to the dynamic equations. Finally, the first-stage loss function used in mask pre-training is defined as:

[0071] ;

[0072] in, For loss weight coefficients, fixed values ​​or adaptive weighting strategies can be adopted, such as Relative Loss Balancing with Random Lookback (ReLoBRaLo), which dynamically adjusts the weights of each loss term for multi-objective optimization tasks to improve model training efficiency and accuracy.

[0073] In mask pre-training, the pulse transform encoder can learn temporal features that conform to the electrophysiological laws of myocardium under unsupervised conditions, thereby improving the physiological consistency, generalization and signal reconstruction quality of the model.

[0074] Regarding downstream task training: In one embodiment, a second training set is obtained, which includes multiple second training samples, including electrocardiogram samples, target photoplethysmography (PPG) signal samples, and labels corresponding to the electrocardiogram samples and PPG signal samples; based on the second training set, downstream task training of the intermediate prediction model is completed to obtain a pre-trained prediction model, which is a prediction model determined through the pre-training stage.

[0075] Specifically, after completing the pre-training based on the physical information neural network constraint mechanism, i.e., after mask pre-training, the pulse feature dictionary parameters and encoder parameters determined by the mask pre-training are loaded for downstream supervised training. In one embodiment, invasive blood gas analysis results are used as label samples to construct a second training set of indicators for blood glucose, lactate, potassium ions, calcium ions, sodium ions, and chloride ions. The intermediate prediction model takes the segmented ECG signal and photoplethysmography signal as input and outputs the predicted value. The loss is calculated using the task loss function, as follows:

[0076] ;

[0077] in, This represents the predicted value output by the regression head of the multilayer perceptron. For label values, Task loss measures the deviation between the model's predicted values ​​and the actual values.

[0078] The multilayer perceptron regression head maps the pulse features output by the encoder into continuous predicted values, adapting to regression tasks. The task loss is optimized by backpropagation to minimize the average error between the predicted and actual values, and to improve the encoder and regression head parameters.

[0079] Finally, the prediction model trained on the downstream task is used as the pre-trained prediction model. The real-time collected electrocardiogram signal and photoplethysmogram signal are input into the pre-trained prediction model, and the encoder and multilayer perceptron regression head output the predicted values ​​of some or all of the indicators such as blood glucose, potassium ion, calcium ion, lactate, sodium ion, and chloride ion, realizing continuous, non-invasive, and intelligent monitoring of internal environmental indicators, which can be integrated with clinical monitoring systems.

[0080] In the technical solution of this invention, the prediction model undergoes sequential training through dictionary-encoder joint training, mask pre-training, and downstream task training to become a pre-trained prediction model. The loss function used in the first stage of mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is a physical constraint loss function determined based on the Fittshur-Namco dynamics equations. Since the Fittshur-Namco dynamics equations are an approximate model of neuronal and myocardial electrophysiology, the combined use of the first loss function and the second loss function determined by the Fittshur-Namco dynamics equations can improve the training effect of mask pre-training, thereby improving the accuracy of the pre-trained prediction model and thus improving the accuracy of its predicted predetermined electrolyte concentration and / or predetermined body metabolite concentration.

[0081] Figure 5 This is a schematic diagram of the non-invasive internal environment parameter data determination device provided in an embodiment of the present invention. The non-invasive internal environment parameter data determination device can be disposed in a processor. Figure 5 As shown, the non-invasive internal environment parameter data determination device includes:

[0082] The acquisition module 110 is used to acquire the predetermined physiological signals of the target object, wherein the predetermined physiological signals include the target electrocardiogram signal and the target photoplethysmography signal under the current time window;

[0083] The prediction module 120 is used to input the predetermined physiological signal into the pre-trained prediction model to obtain the predetermined electrolyte concentration and / or predetermined body metabolite concentration of the target object. The prediction model becomes the pre-trained prediction model after being trained by dictionary-encoder joint training, mask pre-training and downstream task training. The first-stage loss function used in the mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is the physical constraint loss function determined based on the Fittshur-Naghui dynamic equation.

[0084] In the technical solution of this invention, the prediction model undergoes sequential training through dictionary-encoder joint training, mask pre-training, and downstream task training to become a pre-trained prediction model. The loss function used in the first stage of mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is a physical constraint loss function determined based on the Fittshur-Namco dynamics equations. Since the Fittshur-Namco dynamics equations are an approximate model of neuronal and myocardial electrophysiology, the combined use of the first loss function and the second loss function determined by the Fittshur-Namco dynamics equations can improve the training effect of mask pre-training, thereby improving the accuracy of the pre-trained prediction model and thus improving the accuracy of its predicted predetermined electrolyte concentration and / or predetermined body metabolite concentration.

[0085] In one embodiment, the second loss function includes a first mean squared error loss function and a second mean squared error loss function, wherein the first mean squared error loss function is determined based on the membrane potential evolution equation in the Fittshur-Naimun dynamics equation, and the second mean squared error loss function is determined based on the recovery variable evolution equation in the Fittshur-Naimun dynamics equation.

[0086] In one embodiment, the second-stage loss function used in the dictionary-encoder joint training includes an L2 norm term, a first regularization term, and a second regularization term;

[0087] The L2 norm term is the reconstruction loss, used to constrain the error between the encoder output vector and the corresponding atomic feature vector in the pulse feature dictionary;

[0088] The first regularization term is the pulse feature dictionary loss, which is used to constrain the distance between the encoder output vector and the corresponding atomic feature vector in the pulse feature dictionary;

[0089] The second regularization term is the commitment loss, used to constrain the encoder output vector to anchor the corresponding atomic feature vector in the pulse feature dictionary.

[0090] In one embodiment, the joint training of the dictionary-encoder includes:

[0091] Obtain a first training set, which includes multiple first training samples, including electrocardiograms and photoplethysmograms under the target time window.

[0092] Initialize each of the atomic feature vectors in the pulse feature dictionary;

[0093] The dictionary-encoder is jointly trained based on the second-stage loss function and the first training sample set.

[0094] In one embodiment, the encoder includes multiple pulse transform coding layers, which are connected in series and stacked.

[0095] The spiking neurons in the pulse transform coding layer are integral firing neurons.

[0096] In one embodiment, the downstream task training includes:

[0097] Obtain a second training set, which includes multiple second training samples. The second training samples include electrocardiogram samples, target photoplethysmography (PPG) signal samples, and labels corresponding to the electrocardiogram samples and the target PPG signal samples.

[0098] The intermediate prediction model is trained on downstream tasks based on the second training set to obtain a pre-trained prediction model, which is a prediction model determined through the pre-training stage.

[0099] The non-invasive internal environment parameter data determination device provided in the embodiments of the present invention can execute the non-invasive internal environment parameter data determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0100] It is worth noting that the various units and modules included in the above-mentioned non-invasive internal environment parameter data determination device are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the protection scope of the embodiments of the present invention.

[0101] Figure 6 This is a schematic diagram of the internal environment detection system provided in an embodiment of the present invention. Figure 6 As shown, the internal environment monitoring system includes:

[0102] The non-invasive ECG acquisition unit 210 is used to acquire the ECG signal of the target object in real time.

[0103] The non-invasive photoplethysmography acquisition unit 220 is used to acquire the photoplethysmography signal of the target object in real time.

[0104] Processor 230 is configured to determine at least one target physiological indicator data of the target object according to any of the methods described.

[0105] The non-invasive ECG acquisition unit 210 acquires the ECG signal of the target object using a non-invasive acquisition method, and the non-invasive photoplethysmography acquisition unit 220 also acquires the photoplethysmography signal of the target object using a non-invasive acquisition method.

[0106] It should be noted that the non-invasive ECG acquisition unit and the non-invasive photoplethysmography acquisition unit can be based on existing technologies, and will not be described in detail in this embodiment.

[0107] The technical solution provided by the embodiments of the present invention has a pre-trained prediction model with high prediction accuracy. Therefore, the internal environment monitoring system based on the pre-trained prediction model, non-invasive electrocardiogram acquisition unit and non-invasive photoplethysmography acquisition unit can accurately determine the predetermined electrolyte concentration and / or predetermined body metabolite concentration of the target object in real time.

[0108] Figure 7 This is a schematic diagram of a server device provided in an embodiment of the present invention. The server device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The server device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0109] like Figure 7 As shown, server device 10 includes at least one processor 11 and memory, such as read-only memory (ROM) 12 and random access memory (RAM) 13, communicatively connected to at least one processor 11. The memory stores computer programs executable by at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer programs stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of server device 10. The processor 11, ROM 12, and RAM 13 are interconnected via bus 14. Input / output (I / O) interface 15 is also connected to bus 14.

[0110] Multiple components in server device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows server device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0111] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as non-invasive internal environment parameter data determination methods.

[0112] In some embodiments, the noninvasive internal environment parameter data determination method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on server device 10 via read-only memory (ROM) 12 and / or communication unit 19. When the computer program is loaded into random access memory (RAM) 13 and executed by processor 11, one or more steps of the noninvasive internal environment parameter data determination method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the noninvasive internal environment parameter data determination method by any other suitable means (e.g., by means of firmware).

[0113] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0114] Computer programs for implementing the noninvasive internal environment parameter data determination method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer programs can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0115] This invention provides a computer-readable storage medium storing computer instructions for causing a processor to execute a non-invasive internal environment parameter data determination method.

[0116] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0117] To provide interaction with the user, the systems and techniques described herein can be implemented on a server device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the server device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0118] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0119] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0120] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.

[0121] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements a method for determining non-invasive internal environment parameter data according to any embodiment of the invention.

[0122] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0123] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and no limitation is imposed herein.

[0124] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for determining non-invasive internal environment parameter data, characterized in that, The method includes: Acquire predetermined physiological signals of the target object, wherein the predetermined physiological signals include the target electrocardiogram signal and the target photoplethysmogram signal under the current time window; The predetermined physiological signal is input into the pre-trained prediction model to obtain the predetermined electrolyte concentration and / or predetermined metabolic product concentration of the target object. After training by dictionary-encoder joint training, mask pre-training and downstream task training, the prediction model becomes the pre-trained prediction model. The first-stage loss function used in the mask pre-training is a weighted sum of a first loss function and a second loss function. The first loss function is the mask prediction loss function, and the second loss function is the physical constraint loss function determined based on the Fittshur-Narcoma dynamics equations.

2. The method according to claim 1, characterized in that, The second loss function includes a first mean squared error loss function and a second mean squared error loss function. The first mean squared error loss function is determined based on the membrane potential evolution equation in the Fittshur-Naimun dynamic equation, and the second mean squared error loss function is determined based on the recovery variable evolution equation in the Fittshur-Naimun dynamic equation.

3. The method according to claim 1, characterized in that, The second-stage loss function used in the dictionary-encoder joint training includes an L2 norm term, a first regularization term, and a second regularization term; The L2 norm term is the reconstruction loss, used to constrain the error between the encoder output vector and the corresponding atomic feature vector in the pulse feature dictionary; The first regularization term is the pulse feature dictionary loss, which is used to constrain the distance between the encoder output vector and the corresponding atomic feature vector in the pulse feature dictionary; The second regularization term is the commitment loss, used to constrain the encoder output vector to anchor the corresponding atomic feature vector in the pulse feature dictionary.

4. The method according to claim 3, characterized in that, The dictionary-encoder joint training includes: Obtain a first training set, which includes multiple first training samples, including electrocardiograms and photoplethysmograms under the target time window. Initialize the atomic feature vectors in the pulse feature dictionary; The dictionary-encoder is jointly trained based on the second-stage loss function and the first training sample set.

5. The method according to claim 4, characterized in that, The encoder includes multiple pulse transform coding layers, which are connected in series and stacked. The spiking neurons in the pulse transform coding layer are integral firing neurons.

6. The method according to claim 1, characterized in that, The downstream task training includes: Obtain a second training set, which includes multiple second training samples. The second training samples include electrocardiogram samples, target photoplethysmography (PPG) signal samples, and labels corresponding to the electrocardiogram samples and the target PPG signal samples. The downstream task training of the intermediate prediction model is completed based on the second training set to obtain the pre-trained prediction model. The intermediate prediction model is the prediction model determined by the mask pre-training.

7. A non-invasive internal environment parameter monitoring system, characterized in that, include: The electrocardiogram (ECG) acquisition unit is used to acquire the ECG signals of the target object in real time. A photoplethysmography (PPG) acquisition unit is used to acquire the PPG signal of the target object in real time. A processor for determining at least one target physiological indicator data of the target object using the method according to any one of claims 1-6.

8. A server device, characterized in that, The server device includes: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the non-invasive internal environment parameter data determination method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the non-invasive internal environment parameter data determination method according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method for determining non-invasive internal environment parameter data according to any one of claims 1-6.

Images