CPR quality identification model training method, CPR quality identification model using method and related device

Abstract

The embodiment of the invention discloses a CPR quality recognition model training method, a CPR quality recognition model using method and a related device. The training method comprises the steps that a plurality of sample blood oxygen signals collected in advance and CPR quality labels corresponding to the sample blood oxygen signals are obtained; model training is conducted through the sample blood oxygen signals and a preset neural network model, a first recognition result used for reflecting CPR quality corresponding to the sample blood oxygen signals is obtained, and the neural network model is a two-way long-short-term memory network combined with a self-attention mechanism; determining whether the neural network model converges according to the CPR label and the first identification result; and if the neural network model converges, taking the convergent neural network model as a target recognition model of CPR quality. By means of the mode, model training can be conducted through the blood oxygen signals and the neural network model of the two-way long-short-term memory network combined with the self-attention mechanism, the target recognition model of the CPR quality is obtained, the CPR quality is recognized based on the target recognition model, and the accuracy of CPR quality recognition is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cardiopulmonary resuscitation quality recognition, and in particular to a training method, a use method and a related device of a CPR quality recognition model. Background Art

[0002] Cardiovascular disease has become the leading cause of human morbidity and mortality. Every year, a large number of patients die from it worldwide, most of whom suffer from sudden cardiac death. Sudden cardiac death has become a major killer that threatens human life. The most effective and direct medical treatment for this situation is cardiopulmonary resuscitation (CPR). The key to successful CPR is to perform high-quality CPR as early as possible. In clinical practice, the frequency and amplitude of compressions are often insufficient due to various reasons. Therefore, it is particularly important to monitor the quality of CPR during the CPR process. At present, there is no mature method to monitor the quality of CPR, and there is a lack of accurate means to monitor the quality of CPR. Summary of the invention

[0003] The main purpose of the present invention is to provide a training method, a use method and a related device for a CPR quality recognition model, which can solve the problem in the prior art of lacking a means to accurately monitor the quality of cardiopulmonary resuscitation.

[0004] To achieve the above object, the present invention provides a first aspect of a training method for a CPR quality recognition model, the training method comprising:

[0005] Acquire a pre-collected training sample set, wherein the training sample set includes at least a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals;

[0006] Performing model training using the sample blood oxygen signal and a preset neural network model to obtain a first recognition result, wherein the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal;

[0007] Determining whether the neural network model converges according to the CPR label and the first recognition result;

[0008] If the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality.

[0009] In a feasible implementation, the neural network model includes a bidirectional long short-term memory network, a self-attention layer, and a fully connected layer connected in sequence, and the sample blood oxygen signal and the preset neural network model are used to perform model training to obtain a first recognition result, including:

[0010] Inputting the sample blood oxygen signal into the bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features;

[0011] Inputting the plurality of blood oxygen signal features into the self-attention layer for feature fusion to obtain fused blood oxygen signal features;

[0012] The fused blood oxygen signal features are input into the fully connected layer to perform CPR quality prediction to obtain a first recognition result.

[0013] In a feasible implementation, the training method further includes:

[0014] Performing signal filtering processing on the sample blood oxygen signal to obtain a filtered sample blood oxygen signal;

[0015] The filtered sample blood oxygen signal and a preset neural network model are used to perform model training to obtain a first recognition result.

[0016] In a feasible implementation, the bidirectional long short-term memory network at least includes an input layer, a forget gate, a memory gate, a temporary cell state layer, a cell state layer, an output gate, and a hidden state layer connected in sequence; the sample blood oxygen signal is input into the bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features, including:

[0017] The sample blood oxygen signal is input into the input layer, the forget gate, the memory gate, the temporary cell state layer, the cell state layer, the output gate and the hidden state layer for feature extraction to obtain a hidden state sequence, and the blood oxygen signal feature includes the hidden state sequence.

[0018] In a feasible implementation, the inputting the fused blood oxygen signal feature into the fully connected layer to perform CPR quality prediction to obtain a first recognition result includes:

[0019] Performing global average pooling processing on the fused blood oxygen signal features to obtain pooled blood oxygen signal features;

[0020] The first recognition result is obtained by using the blood oxygen signal features after the pooling process and the preset Softmax function.

[0021] To achieve the above object, the second aspect of the present invention provides a method for using a CPR quality recognition model, the method comprising:

[0022] During the CPR process of the target object, obtaining a blood oxygen signal to be identified of the target object;

[0023] The blood oxygen signal to be identified is input into a target recognition model for identification to obtain a second recognition result, wherein the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be identified. The target recognition model is obtained by using the training method described in the first aspect and any feasible implementation method.

[0024] To achieve the above object, the present invention provides a training device for a CPR quality recognition model in a third aspect, the training device comprising:

[0025] Sample acquisition module: used to acquire a pre-collected training sample set, wherein the training sample set at least includes a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals;

[0026] Model training module: used to perform model training using the sample blood oxygen signal and a preset neural network model to obtain a first recognition result, wherein the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal;

[0027] Convergence judgment module: used to determine whether the neural network model converges according to the CPR label and the first recognition result;

[0028] Model determination module: used to use the converged neural network model as the target recognition model of CPR quality if the neural network model converges.

[0029] To achieve the above-mentioned object, the fourth aspect of the present invention provides a device for using a CPR quality recognition model, the device comprising:

[0030] Signal acquisition module: used to obtain the blood oxygen signal to be identified of the target object during the CPR process of the target object;

[0031] Quality identification module: used to input the blood oxygen signal to be identified into the target recognition model for identification to obtain a second recognition result, wherein the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be identified, and the target recognition model is obtained by adopting the training method described in the first aspect and any feasible implementation method.

[0032] To achieve the above-mentioned purpose, the fifth aspect of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the training method described in the first aspect and any feasible implementation method or the using method described in the second aspect.

[0033] To achieve the above-mentioned purpose, the sixth aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the training method described in the first aspect and any feasible implementation method or the using method described in the second aspect.

[0034] The embodiments of the present invention have the following beneficial effects:

[0035] The present invention provides a training method for a recognition model of CPR quality, the training method comprising: obtaining a pre-collected training sample set, the training sample set at least comprising a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals; using the sample blood oxygen signals and a preset neural network model to perform model training to obtain a first recognition result, the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal; according to the CPR label and the first recognition result, determining whether the neural network model converges; if the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality. In the above manner, the blood oxygen signal and the neural network model of the bidirectional long short-term memory network combined with the self-attention mechanism can be used for model training to obtain a target recognition model for CPR quality, and the CPR quality can be recognized based on the target recognition model, thereby improving the accuracy of CPR quality recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] in:

[0038] Figure 1 It is a flowchart of a training method of a CPR quality recognition model in an embodiment of the present invention;

[0039] Figure 2 Another flow chart of a method for training a CPR quality recognition model in an embodiment of the present invention;

[0040] Figure 3 It is a flow chart of a method for using a CPR quality recognition model in an embodiment of the present invention;

[0041] Figure 4 It is a structural block diagram of a training device for a CPR quality recognition model in an embodiment of the present invention;

[0042] Figure 5 It is a structural block diagram of a device for using a CPR quality recognition model in an embodiment of the present invention;

[0043] Figure 6 4 is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] It should be noted that BiLSTM is the abbreviation of Bi-directional Long Short-Term Memory, which is composed of forward LSTM and backward LSTM. Due to its design characteristics, LSTM has a good effect in extracting time series data features and is very suitable for modeling time series data. Self-Attention is a variant of the attention mechanism, which is better at capturing the internal correlation of data or features.

[0046] In addition, studies have shown that the blood oxygen pulse signal (referred to as blood oxygen signal) can reflect the compression quality of cardiopulmonary resuscitation (referred to as CPR), so the present application predicts the compression quality of cardiopulmonary resuscitation based on the blood oxygen signal.

[0047] The present invention provides a method for identifying CPR quality through blood oxygen signals, and uses a BiLSTM combined with a self-attention neural network model to realize the identification and monitoring of CPR quality. It is committed to the development of a portable, non-invasive, economical CPR monitoring feedback system that can track and reflect ROSC in real time and can be widely promoted and applied.

[0048] See also Figure 1 , Figure 1 The flowchart of a method for training a CPR quality recognition model in an embodiment of the present invention is shown in FIG. 1 . The method can be applied to a terminal or a server. The terminal can be a desktop terminal or a mobile terminal. The mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The server can be implemented by an independent server or a server cluster composed of multiple servers. This embodiment is described by applying to a terminal as an example. Figure 1 The training method shown includes the following steps:

[0049] 101. Acquire a pre-collected training sample set, where the training sample set at least includes a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals;

[0050] It can be understood that the present invention uses blood oxygen signals to predict CPR quality. In order to enable the neural network model to learn the relationship between blood oxygen signals and CPR quality, it is necessary to use a sample set consisting of a large number of blood oxygen signals marked with CPR quality labels to train the original neural network model, so as to obtain a recognition model that can identify CPR quality based on blood oxygen signals.

[0051] During the training process, it is first necessary to obtain a pre-collected training sample set, which includes a number of sample blood oxygen signals and CPR quality labels for each sample blood oxygen signal. The sample blood oxygen signal can be a blood oxygen signal during historical cardiopulmonary resuscitation. The CPR quality label is used to reflect the compression quality of historical cardiopulmonary resuscitation, such as high-quality compression or low-quality compression. The dimensions of the specific classification labels can be adaptively set according to actual conditions. This is only an example and is not specifically limited. Among them, the blood oxygen signal can be of different characteristics, and the CPR quality label can be manually labeled. The training sample set is obtained by counting the different characteristic blood oxygen signals with CPR quality labels that have been manually labeled.

[0052] In a feasible implementation, in order to further improve the model training effect and recognition accuracy, the sample blood oxygen signal can also be subjected to signal filtering processing to obtain a filtered sample blood oxygen signal, remove noise in the signal, and then perform the following step 102 based on the filtered sample blood oxygen signal to perform model training to obtain a first recognition result.

[0053] 102. Perform model training using the sample blood oxygen signal and a preset neural network model to obtain a first recognition result, where the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal;

[0054] 103. Determine whether the neural network model converges according to the CPR label and the first recognition result;

[0055] 104. If the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality.

[0056] After obtaining the training sample set, the sample blood oxygen signal and the preset neural network model can be used for model training to obtain the first recognition result, wherein the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism (i.e., BiLSTM combined with a Self-Attention model), and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal. The first recognition result can be high-quality compression or low-quality compression, wherein ROSC represents the restoration of spontaneous circulation. For patients with cardiac arrest, cardiopulmonary resuscitation is performed. If the patient restores spontaneous circulation, it means that the CPR compression quality is high. If the patient does not have ROSC, it means that the CPR compression quality is low. The label can also be set to whether the spontaneous circulation is restored. Then, through the CPR label and the first recognition result, it is determined whether the neural network model converges, such as inputting the CPR label and the first recognition result into a preset loss function, and judging whether it is converged enough by the loss function value. When it is determined that the neural network model converges, it means that the model at this time can accurately identify the CPR quality corresponding to the blood oxygen signal, and the converged neural network model can be used as the target recognition model of CPR quality. In a feasible implementation, the target recognition model can be further tested using a test sample set, and the model parameters can be further optimized based on the test recognition results to obtain a more stable recognition model. The collected sample blood oxygen signals can be divided into a training set and a test set, and the neural network model is trained using the training set and tested using the test set until the model converges and can recognize the CPR quality model.

[0057] The present invention provides a training method for a recognition model of CPR quality, the training method comprising: obtaining a pre-collected training sample set, the training sample set at least comprising a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals; using the sample blood oxygen signals and a preset neural network model to perform model training to obtain a first recognition result, the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal; according to the CPR label and the first recognition result, determining whether the neural network model converges; if the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality. In the above manner, the blood oxygen signal and the neural network model of the bidirectional long short-term memory network combined with the self-attention mechanism can be used for model training to obtain a target recognition model for CPR quality, and the CPR quality can be recognized based on the target recognition model, thereby improving the accuracy of CPR quality recognition.

[0058] See also Figure 2 , Figure 2 FIG. 4 is another flow chart of a method for training a CPR quality recognition model in an embodiment of the present invention, Figure 2The training method shown includes the following steps:

[0059] 201. Acquire a pre-collected training sample set, where the training sample set at least includes a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals;

[0060] It should be noted that step 201 and Figure 1 The contents of step 101 are similar, and are not described here to avoid repetition. For details, please refer to Figure 1 The content of step 101 is shown.

[0061] The neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, which includes a bidirectional long short-term memory network, a self-attention layer, and a fully connected layer connected in sequence. The processed signal is extracted from the blood oxygen signal features through BiLSTM combined with the Self-Attention model, and a fully connected layer is added for feature fusion.

[0062] 202. Input the sample blood oxygen signal into the bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features;

[0063] In order to better extract the data features of the blood oxygen signal, the sample blood oxygen signal is input into the bidirectional long short-term memory network for feature extraction to obtain several blood oxygen signal features.

[0064] In a feasible implementation, the bidirectional long short-term memory network includes at least an input layer, a forget gate, a memory gate, a temporary cell state layer, a cell state layer, an output gate, and a hidden state layer connected in sequence; then the sample blood oxygen signal is input into the bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features, including: inputting the sample blood oxygen signal into the input layer, the forget gate, the memory gate, the temporary cell state layer, the cell state layer, the output gate, and the hidden state layer for feature extraction to obtain a hidden state sequence, and the blood oxygen signal feature includes the hidden state sequence.

[0065] Specifically, by pre-building the BiLSTM network layer, the LSTM model corresponding to the BiLSTM network layer is composed of the input layer X at time t t , cell state C t , temporary cell state Hidden state h t 、Forget gate t 、Memory Gate t , output gate o t The calculation formula is as follows:

[0066] f t =σ(W f *[h t-1,x t ]+b f );

[0067] i t =σ(W i *[h t-1 ,x t ]+b i );

[0068]

[0069] o t =σ(W o *[h t-1 ,x t ]+b o );

[0070] h t =o t *tanhC t ;

[0071] Finally, we can get a hidden state sequence {h0,h1,h2,…,h n-1}; In the above formula, σ is the sigmoid activation function; W f , W i , W C , W o are the weight parameters of the forget gate, memory gate, cell state, and output gate respectively; b f , b i , b C , b o are the bias coefficients of the forget gate, memory gate, cell state, and output gate respectively; x t It is the vector representation of the tth time step of the input time series (original blood oxygen signal).

[0072] Among them, the hidden state sequence {h0,h1,h2,…,h n-1}The input S of the corresponding self-attention layer K =[S K 1,S K 2,S K 3,……S K N ], K represents the number of features.

[0073] 203. Input the plurality of blood oxygen signal features into the self-attention layer for feature fusion to obtain fused blood oxygen signal features;

[0074] In order to better capture the internal correlation of data or features, a self-attention layer is used to perform feature fusion of blood oxygen signal features. The several blood oxygen signal features are input into the self-attention layer for feature fusion to obtain fused blood oxygen signal features.

[0075] In one feasible implementation, by pre-building a Self-Attention layer, since the self-attention mechanism can take into account the intrinsic connections between different features and learn the internal dependencies between different features, the self-attention mechanism can be used to calculate the weight of the CPR compression data feature in the blood oxygen signal sample through a one-step matrix calculation, so as to better solve the long-term dependency problem of time series data.

[0076] The input of the self-attention layer is S K =[S K 1,S K 2,S K 3,……S K N ], where S K For the feature matrix output by the above BiLSTM, the calculation formula of the self-attention mechanism is as follows:

[0077] R=S K *(softmax(W T tanh(S K ))) T ;

[0078] Among them, W is the parameter vector available for training, T is the transposition, and the data characteristics of the final blood oxygen signal are expressed as follows:

[0079] S R =tanh(R);

[0080] 204. Input the fused blood oxygen signal feature into the fully connected layer to perform CPR quality prediction to obtain a first recognition result, where the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal;

[0081] Furthermore, the fully connected layer is used to predict the results, and the softmax function is used after global average pooling to obtain the ROSC quality recognition result. The fused blood oxygen signal feature S R The fully connected layer is input to perform CPR quality prediction to obtain a first recognition result.

[0082] In a feasible implementation, the inputting the fused blood oxygen signal feature into the fully connected layer to perform CPR quality prediction to obtain a first recognition result includes:

[0083] Performing global average pooling processing on the fused blood oxygen signal features to obtain pooled blood oxygen signal features;

[0084] The first recognition result is obtained by using the blood oxygen signal features after the pooling process and the preset Softmax function.

[0085] Among them, the above blood oxygen signal data features can be globally averaged and pooled to greatly reduce network parameters, thereby reducing the overfitting phenomenon. Then, the pooled feature vector S is calculated using the Softmax function to obtain the ROSC quality result, as shown in the following formula:

[0086] R=S K *(softmax(W S S i +b i ));

[0087] Among them, W s is the weight matrix of global average pooling, b i is the bias coefficient, S i is one of the vectors in the pooled feature vector S, S K is the feature matrix output by the bidirectional long short-term memory neural network layer.

[0088] 205. Determine whether the neural network model converges according to the CPR label and the first recognition result;

[0089] 206. If the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality;

[0090] It should be noted that steps 205 and 206 are Figure 1 The contents of steps 103 and 104 are similar, and are not described here to avoid repetition. For details, please refer to Figure 1 The contents of steps 103 and 104 are shown.

[0091] The present invention provides a training method for a recognition model of CPR quality, the training method comprising: obtaining a pre-collected training sample set, the training sample set at least comprising a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals; inputting the sample blood oxygen signals into a bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features; inputting the several blood oxygen signal features into a self-attention layer for feature fusion to obtain fused blood oxygen signal features; inputting the fused blood oxygen signal features into a fully connected layer for CPR quality prediction to obtain a first recognition result, the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal; according to the CPR label and the first recognition result, determining whether the neural network model converges; if the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality. The bidirectional long short-term memory network used in the present invention can extract more important compression features in the blood oxygen signal under ventricular fibrillation conditions, obtain a more accurate feature expression, extract the internal connection between feature matrices through the self-attention mechanism, establish a BiLSTM combined with a Self-Attention model for CRP quality recognition, and have a good effect on the CRP quality recognition result of the blood oxygen signal.

[0092] See also Figure 3 , Figure 3 The flowchart of the method for using the CPR quality recognition model in the embodiment of the present invention is as follows. The method can be applied to both a terminal and a server. The terminal can be a desktop terminal or a mobile terminal. The mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The server can be implemented by an independent server or a server cluster composed of multiple servers. This embodiment is described by applying to a terminal as an example. Figure 3 The method of use shown includes the following steps:

[0093] 301. During a CPR process of a target object, obtaining a blood oxygen signal to be identified of the target object;

[0094] 302. Input the blood oxygen signal to be identified into a target recognition model for identification to obtain a second recognition result, where the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be identified.

[0095] The target recognition model is as follows: Figure 1 or Figure 2 The target object is a subject to be resuscitated by cardiopulmonary resuscitation, and the training method is adopted as follows Figure 1 or Figure 2 The target recognition model obtained by the training method and the blood oxygen signal to be recognized collected in real time can recognize the real-time compression quality of the target object during CPR.

[0096] The present invention provides a method for using a recognition model of CPR quality, the method comprising: obtaining a blood oxygen signal to be recognized by the target object during the CPR process of the target object; inputting the blood oxygen signal to be recognized into the target recognition model for recognition, and obtaining a second recognition result, wherein the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be recognized, wherein the target recognition model is adopted as follows Figure 1 or Figure 2 The method can be used to train a target recognition model of CPR quality using a blood oxygen signal and a neural network model of a bidirectional long short-term memory network combined with a self-attention mechanism to identify the CPR quality corresponding to the blood oxygen signal to be identified, thereby improving the accuracy of CPR quality recognition.

[0097] See also Figure 4 , Figure 4 FIG. 4 is a structural block diagram of a training device for a CPR quality recognition model in an embodiment of the present invention, such as Figure 4 The training device shown comprises:

[0098] Sample acquisition module 401: used to acquire a pre-collected training sample set, wherein the training sample set at least includes a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals;

[0099] Model training module 402: used to perform model training using the sample blood oxygen signal and a preset neural network model to obtain a first recognition result, wherein the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal;

[0100] Convergence judgment module 403: used to determine whether the neural network model converges according to the CPR label and the first recognition result;

[0101] Model determination module 404: used to use the converged neural network model as a target recognition model for CPR quality if the neural network model converges.

[0102] It should be noted that Figure 4 The content of each module in the training device shown is similar to Figure 1 The contents of each step in the training method are similar, and will not be described here to avoid repetition. For details, please refer to Figure 1 The content of each step in the training method.

[0103] The present invention provides a training device for a recognition model of CPR quality, the training device comprising: a sample acquisition module: used to acquire a pre-collected training sample set, the training sample set at least comprising a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals; a model training module: used to perform model training using the sample blood oxygen signals and a preset neural network model to obtain a first recognition result, the neural network model being a bidirectional long short-term memory network combined with a self-attention mechanism, the first recognition result being used to reflect the CPR quality corresponding to the sample blood oxygen signal; a convergence judgment module: used to determine whether the neural network model converges according to the CPR label and the first recognition result; a model determination module: used to use the converged neural network model as a target recognition model of CPR quality if the neural network model converges. The above device can be used to perform model training using blood oxygen signals and a neural network model of a bidirectional long short-term memory network combined with a self-attention mechanism to obtain a target recognition model of CPR quality, recognize CPR quality based on the target recognition model, and improve the accuracy of CPR quality recognition.

[0104] See also Figure 5 , Figure 5 FIG. 1 is a structural block diagram of a device for using a CPR quality recognition model in an embodiment of the present invention, such as Figure 5 The device shown for use includes:

[0105] Signal acquisition module 501: used to acquire the blood oxygen signal to be identified of the target object during the CPR process of the target object;

[0106] The quality identification module 502 is used to input the blood oxygen signal to be identified into the target identification model for identification to obtain a second identification result, where the second identification result is used to reflect the CPR quality corresponding to the blood oxygen signal to be identified.

[0107] The target recognition model is as follows: Figure 1 or Figure 2 The training method is obtained.

[0108] It should be noted that Figure 5 The contents of each module in the device shown are Figure 3 The contents of each step in the method of use are similar, and are not described here to avoid repetition. For details, please refer to Figure 3 The content of each step in the method of use.

[0109] The present invention provides a device for using a recognition model of CPR quality, the device comprising: a signal acquisition module: used to obtain a blood oxygen signal to be recognized by a target object during the CPR process of the target object; a quality recognition module: used to input the blood oxygen signal to be recognized into a target recognition model for recognition, and obtain a second recognition result, the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be recognized, wherein the target recognition model is adopted as follows Figure 1 or Figure 2 The device can use the blood oxygen signal and the neural network model of the bidirectional long short-term memory network combined with the self-attention mechanism to train the CPR quality target recognition model to identify the CPR quality corresponding to the blood oxygen signal to be identified, thereby improving the accuracy of CPR quality recognition.

[0110] Figure 6 FIG. 1 shows an internal structure diagram of a computer device in an embodiment. The computer device may be a terminal or a server. Figure 6 As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the above method. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement the above method. Those skilled in the art can understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0111] In one embodiment, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps: Figure 1 , Figure 2 or Figure 3 Steps shown.

[0112] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor performs the following steps: Figure 1 , Figure 2 or Figure 3 Steps shown.

[0113] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0114] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A training method for a CPR quality recognition model, characterized in that: The training method comprises: Acquire a pre-collected training sample set, wherein the training sample set includes at least a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals; Performing model training using the sample blood oxygen signal and a preset neural network model to obtain a first recognition result, wherein the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal; Determining whether the neural network model converges according to the CPR label and the first recognition result; If the neural network model converges, the converged neural network model is used as a target recognition model for CPR quality.

2. The training method according to claim 1, characterized in that: The neural network model includes a bidirectional long short-term memory network, a self-attention layer, and a fully connected layer connected in sequence, and the sample blood oxygen signal and the preset neural network model are used to perform model training to obtain a first recognition result, including: Inputting the sample blood oxygen signal into the bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features; Inputting the plurality of blood oxygen signal features into the self-attention layer for feature fusion to obtain fused blood oxygen signal features; The fused blood oxygen signal features are input into the fully connected layer to perform CPR quality prediction to obtain a first recognition result.

3. The training method according to claim 1, characterized in that: The training method further comprises: Performing signal filtering processing on the sample blood oxygen signal to obtain a filtered sample blood oxygen signal; The filtered sample blood oxygen signal and a preset neural network model are used to perform model training to obtain a first recognition result.

4. The training method according to claim 2, characterized in that: The bidirectional long short-term memory network at least includes an input layer, a forget gate, a memory gate, a temporary cell state layer, a cell state layer, an output gate and a hidden state layer connected in sequence; the sample blood oxygen signal is input into the bidirectional long short-term memory network for feature extraction to obtain a number of blood oxygen signal features, including: The sample blood oxygen signal is input into the input layer, the forget gate, the memory gate, the temporary cell state layer, the cell state layer, the output gate and the hidden state layer for feature extraction to obtain a hidden state sequence, and the blood oxygen signal feature includes the hidden state sequence.

5. The training method according to claim 2, characterized in that: The step of inputting the fused blood oxygen signal features into the fully connected layer to perform CPR quality prediction to obtain a first recognition result includes: Performing global average pooling processing on the fused blood oxygen signal features to obtain pooled blood oxygen signal features; The first recognition result is obtained by using the blood oxygen signal features after the pooling process and the preset Softmax function.

6. A method for using a CPR quality recognition model, characterized in that: The method of use includes: During the CPR process of the target object, obtaining a blood oxygen signal to be identified of the target object; The blood oxygen signal to be identified is input into a target recognition model for identification to obtain a second recognition result, wherein the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be identified, and the target recognition model is obtained using the training method described in any one of claims 1 to 5.

7. A training device for a CPR quality recognition model, characterized in that: The training device comprises: Sample acquisition module: used to acquire a pre-collected training sample set, wherein the training sample set at least includes a number of sample blood oxygen signals and CPR quality labels corresponding to the sample blood oxygen signals; Model training module: used to perform model training using the sample blood oxygen signal and a preset neural network model to obtain a first recognition result, wherein the neural network model is a bidirectional long short-term memory network combined with a self-attention mechanism, and the first recognition result is used to reflect the CPR quality corresponding to the sample blood oxygen signal; Convergence judgment module: used to determine whether the neural network model converges according to the CPR label and the first recognition result; Model determination module: used to use the converged neural network model as the target recognition model of CPR quality if the neural network model converges.

8. A device for using a CPR quality recognition model, characterized in that: The use device comprises: Signal acquisition module: used to obtain the blood oxygen signal to be identified of the target object during the CPR process of the target object; Quality identification module: used to input the blood oxygen signal to be identified into the target recognition model for identification to obtain a second recognition result, wherein the second recognition result is used to reflect the CPR quality corresponding to the blood oxygen signal to be identified, and the target recognition model is obtained by using the training method as described in any one of claims 1 to 5.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 5 or claim 7.

10. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method according to any one of claims 1 to 5 or claim 7.

Images