Method and device for determining fetal heart rate, equipment and storage medium

Abstract

The embodiment of the invention discloses a fetal heart rate determining method and device, equipment and a storage medium. The method comprises the steps that an original fetal heart signal is obtained; the original fetal heart signals are input into a target residual network, target fetal heart signals output by the target residual network are obtained, and the target residual network is used for conducting noise reduction processing on the original fetal heart signals; and performing fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate. Through the mode, the pre-trained target residual network can firstly perform noise reduction processing on the original fetal heart signal to obtain the target fetal heart signal after noise reduction, then fetal heart rate determination processing is performed based on the target fetal heart signal to obtain the target fetal heart rate, the influence degree of noise on fetal heart rate calculation is reduced, and the accuracy of fetal heart rate calculation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fetal heart rate monitoring, and in particular to a method for determining a fetal heart rate, a determining device, equipment and a storage medium. Background Art

[0002] Fetal heart rate (FHR) is an important basic parameter for perinatal assessment of fetal health. The fetal heart cycle is calculated using the ultrasonic Doppler fetal heart signal with Doppler frequency shift effect. It is widely used in clinical monitoring of fetal heart rate because of its high efficiency, non-invasiveness and safety. In the field of time domain signal processing, the main technology for calculating instantaneous fetal heart rate using ultrasonic Doppler fetal heart signal is autocorrelation technology, which uses the property that the period of the autocorrelation function is consistent with the signal period. The instantaneous fetal heart rate can be obtained by calculating the period of the autocorrelation function.

[0003] However, the fetal heart rate signal obtained by ultrasound Doppler is a time series signal containing a lot of noise. Fetal movement, displacement of the fetal heart rate probe, and maternal movement will all generate corresponding noise. Therefore, when encountering noise, using the autocorrelation method to calculate the fetal heart rate may result in the calculation result being doubled or halved. When serious calculation errors occur, the abnormal instantaneous fetal heart rate will bring a certain degree of psychological burden to the user. However, the current traditional signal noise reduction method is difficult to eliminate the noise in the ultrasound Doppler fetal heart rate signal in multiple dimensions, and thus it is difficult to reduce the impact of noise on the calculation of the fetal heart rate. Summary of the invention

[0004] The main purpose of the present invention is to provide a method for determining a fetal heart rate, a determination device, a device and a storage medium, which can solve the problem in the prior art that it is difficult to reduce the influence of noise on the calculation of the fetal heart rate.

[0005] To achieve the above object, the present invention provides a method for determining a fetal heart rate in a first aspect, the method comprising:

[0006] Obtain original fetal heart rate signal;

[0007] Inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network, wherein the target residual network is used to perform noise reduction processing on the original fetal heart signal;

[0008] A fetal heart rate determination process is performed based on the target fetal heart rate signal to obtain a target fetal heart rate.

[0009] In a feasible implementation, inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network includes:

[0010] Dividing the original fetal heart signal into equal lengths to obtain a plurality of divided first fetal heart signals;

[0011] Inputting the first fetal heart signals into the target residual network for noise reduction processing respectively, to obtain respective second fetal heart signals output by the target residual network;

[0012] The second fetal heart signals are spliced ​​together to obtain the target fetal heart signal.

[0013] In a feasible implementation, the step of performing fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate includes:

[0014] Performing low-pass filtering on the target fetal heart signal to obtain an envelope signal corresponding to the target fetal heart signal;

[0015] Performing an autocorrelation operation according to the envelope signal to obtain an autocorrelation curve;

[0016] Using the autocorrelation curve to perform feature extraction, obtain the peak position of the autocorrelation curve, where the peak position is used to reflect the actual heartbeat moment;

[0017] The target fetal heart rate is obtained according to the peak position and a preset fetal heart rate algorithm.

[0018] In a feasible implementation, the method further includes:

[0019] Acquire a training sample set, wherein the training sample set includes a third fetal heart signal marked with a real cardiac cycle label;

[0020] Inputting the third fetal heart signal into a pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network;

[0021] Determining whether the initial residual network converges based on the fourth fetal heart signal and the real cardiac cycle label;

[0022] When the initial residual network converges, the converged initial residual network is used as the target residual network.

[0023] In a feasible implementation, the determining whether the initial residual network converges based on the fourth fetal heart signal and the real cardiac cycle label includes:

[0024] Determine a current loss value of the initial residual network using the fourth fetal heart signal and the true cardiac cycle label;

[0025] If the current loss value is at the minimum value, it is determined that the initial residual network converges, and the step of using the converged initial residual network as the target residual network is performed;

[0026] If the current loss value is not at the minimum value, it is determined that the initial residual network has not converged, and the model parameters of the initial residual network are adjusted, and the step of inputting the third fetal heart signal into the pre-constructed initial residual network to obtain the fourth fetal heart signal output by the initial residual network is returned to execute.

[0027] In a feasible implementation, the initial residual network includes at least one residual block, and the residual block includes at least two convolutional layers connected in series. Then, the third fetal heart signal is input into a pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network, including:

[0028] Inputting the third fetal heart signal into the residual block to perform a two-layer convolution operation on the third fetal heart signal to obtain a first signal feature after the convolution operation;

[0029] Adding the first signal feature to the third fetal heart signal to obtain a second signal feature;

[0030] The second signal feature is activated using a preset activation function to obtain the fourth fetal heart signal.

[0031] To achieve the above object, the second aspect of the present invention provides a device for determining a fetal heart rate, the device comprising:

[0032] Signal acquisition module: used to obtain the original fetal heart signal;

[0033] A noise reduction processing module is used to input the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network, wherein the target residual network is used to perform noise reduction processing on the original fetal heart signal;

[0034] Fetal heart rate determination module: used to perform fetal heart rate determination processing based on the target fetal heart rate signal to obtain the target fetal heart rate.

[0035] In a feasible implementation, the device further includes:

[0036] Sample acquisition module: used to acquire a training sample set, wherein the training sample set includes a third fetal heart signal marked with a real cardiac cycle label;

[0037] Network training module: used for inputting the third fetal heart signal into a pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network; determining whether the initial residual network converges based on the fourth fetal heart signal and the real cardiac cycle label; when the initial residual network converges, the converged initial residual network is used as the target residual network.

[0038] To achieve the above-mentioned purpose, the third 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 shown in the first aspect and any feasible implementation method.

[0039] To achieve the above-mentioned purpose, the fourth 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 shown in the first aspect and any feasible implementation method.

[0040] The embodiments of the present invention have the following beneficial effects:

[0041] The present invention provides a method for determining a fetal heart rate, the method comprising: obtaining an original fetal heart signal; inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network, the target residual network being used to perform noise reduction processing on the original fetal heart signal; performing fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate. Through the above method, the pre-trained target residual network can first perform noise reduction processing on the original fetal heart signal to obtain a noise-reduced target fetal heart signal, and then perform fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate, thereby reducing the influence of noise on the calculation of the fetal heart rate and improving the accuracy of the fetal heart rate calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] 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.

[0043] in:

[0044] Figure 1 is a flow chart of a method for determining a fetal heart rate in an embodiment of the present invention;

[0045] Figure 2 is another flow chart of a method for determining a fetal heart rate in an embodiment of the present invention;

[0046] Figure 3 Schematic diagram of a residual block of a residual network in an embodiment of the present invention;

[0047] Figure 4 is a flow chart of a device for determining fetal heart rate in an embodiment of the present invention;

[0048] Figure 5 4 is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] 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.

[0050] See also Figure 1 , Figure 1 The flowchart of a method for determining a fetal heart rate 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 method shown comprises the following steps:

[0051] 101. Obtaining original fetal heart rate signal;

[0052] 102. Inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network, wherein the target residual network is used to perform noise reduction processing on the original fetal heart signal;

[0053] 103. Perform fetal heart rate determination processing based on the target fetal heart rate signal to obtain a target fetal heart rate.

[0054] It should be noted that the present invention adopts the residual network in deep learning to reduce the noise of the ultrasonic Doppler fetal heart signal, and uses the self-learning method to eliminate the noise in the signal from multiple dimensions, thereby reducing the influence of noise on the calculation of the fetal heart rate by the autocorrelation technology, which can effectively reduce the incidence of doubling and halving the instantaneous fetal heart rate, and obtain a relatively stable fetal heart rate, wherein the above-mentioned original fetal heart signal is an ultrasonic Doppler fetal heart signal collected in real time; the target residual network is a pre-trained residual network, and the self-learning method is used during training to automatically learn the data features of the noise-free fetal heart signal, and effectively eliminate the noise signal in the fetal heart signal. Exemplarily, by designing the residual network and performing network training, the network weight value of the residual network has reached a stable output of a fetal heart signal with less noise (that is, the network reaches a stable state when the loss function is small), and the network training can be stopped to obtain the target residual network; the target fetal heart signal is the fetal heart signal after noise reduction by the target residual network, that is, the ultrasonic Doppler fetal heart signal after noise reduction.

[0055] Further, based on the target fetal heart rate signal, the fetal heart rate determination process is performed, and the main method for detecting the fetal heart rate of the target fetal heart rate is to use autocorrelation technology, and the period of the ultrasonic Doppler fetal heart signal is calculated by autocorrelation technology. However, before the use of the autocorrelation technology, only ordinary low-pass filtering is used to eliminate the influence of noise on the autocorrelation calculation, so the influence of noise must be eliminated from the signal source, thereby reducing the incidence of doubling and halving. The present invention uses the residual network in the deep learning method to perform noise reduction on the ultrasonic Doppler fetal heart signal, and uses the self-learning method to automatically learn the data characteristics of the noise-free fetal heart signal, effectively eliminate the noise signal in the fetal heart signal, and then use the autocorrelation technology to obtain the fetal heart rate. This method not only obtains a stable fetal heart rate, but also avoids the phenomenon that the noise signal causes the fetal heart rate calculation error, effectively reducing the incidence of doubling and halving. The fetal heart signal with noise is input into the residual network to obtain the fetal heart signal after noise reduction; the envelope signal is obtained by performing low-pass filtering and other operations on the fetal heart signal after noise reduction, and the fetal heart rate is calculated in the envelope signal using the autocorrelation technology.

[0056] The present invention provides a method for determining a fetal heart rate, the method comprising: obtaining an original fetal heart signal; inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network, the target residual network being used to perform noise reduction processing on the original fetal heart signal; performing fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate. Through the above method, the pre-trained target residual network can first perform noise reduction processing on the original fetal heart signal to obtain a noise-reduced target fetal heart signal, and then perform fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate, thereby reducing the influence of noise on the calculation of the fetal heart rate and improving the accuracy of the fetal heart rate calculation.

[0057] See also Figure 2 , Figure 2 FIG. 4 is another flow chart of a method for determining a fetal heart rate in an embodiment of the present invention. Figure 2 The method comprises the following steps:

[0058] 201. Obtaining an original fetal heart rate signal;

[0059] It should be noted that step 201 and Figure 1 The contents of step 101 are similar and will not be described here to avoid repetition. Figure 1 The content of step 101 is shown.

[0060] In order to reduce the pressure of data processing, the original fetal heart signal can be divided and then denoised in sequence, that is, the original fetal heart signal is input into the target residual network, and obtaining the target fetal heart signal output by the target residual network may include the following steps 202, 203 and 204.

[0061] 202. Divide the original fetal heart signal into equal lengths to obtain a plurality of divided first fetal heart signals;

[0062] 203. Input the first fetal heart signals into the target residual network for noise reduction processing to obtain respective second fetal heart signals output by the target residual network;

[0063] 204. Concatenate the second fetal heart signals to obtain the target fetal heart signal;

[0064] It should be noted that, due to the high sampling rate of the ultrasonic fetal heart signal, many data points will be collected in a very short time, which increases the amount of calculation. In order to reduce the amount of calculation, the collected fetal heart signal needs to be divided, which can be divided into equal lengths, such as dividing every 10 seconds, so as to obtain multiple first fetal heart signals, and denoise the first fetal heart signals respectively, and input them into the target residual network for denoising respectively, and splice the denoised second fetal heart signals together, splice them in order, and obtain the target fetal heart signal. That is, the denoised ultrasonic Doppler fetal heart signal is obtained by using the trained residual network, which can be divided into equal lengths, and after inputting the network to obtain equal length outputs, the output results are spliced, so that the denoised fetal heart signal can be obtained.

[0065] For example, a 100-second ultrasound fetal heart signal x is divided into 10 segments, labeled x1, x2, ... x10, and sent to the trained residual network to calculate the outputs y1, y2, ... y10 of equal length, which are then merged to obtain the complete denoised fetal heart signal y.

[0066] 205. Perform fetal heart rate determination processing based on the target fetal heart rate signal to obtain a target fetal heart rate.

[0067] It should be noted that step 205 and Figure 1 The contents of step 103 are similar, and are not described here to avoid repetition. For details, please refer to Figure 1 The content of step 103 is shown.

[0068] In a feasible implementation, the step of performing fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate includes:

[0069] A01. Low-pass filter the target fetal heart signal to obtain an envelope signal corresponding to the target fetal heart signal;

[0070] A02, performing an autocorrelation operation according to the envelope signal to obtain an autocorrelation curve;

[0071] A03. Extract features using the autocorrelation curve to obtain the peak position of the autocorrelation curve, where the peak position is used to reflect the actual heartbeat moment;

[0072] A04. Obtain a target fetal heart rate according to the peak position and a preset fetal heart rate algorithm.

[0073] Furthermore, the envelope signal of the ultrasonic Doppler fetal heart signal is obtained by using traditional signal processing methods such as low-pass filtering. For example, the envelope of the signal is calculated. In essence, the signal envelope is the baseband part of the signal. For an input signal x, its spectrum center frequency point is moved to zero frequency, and then a low-pass filter is used to filter out the high-frequency carrier to obtain the envelope baseband signal. In the time domain, the contour of the signal waveform is calculated. Then, according to the obtained envelope signal, the autocorrelation operation of the envelope signal is performed to obtain the peak value of the true heartbeat moment in the envelope signal. Among them, the autocorrelation operation is to calculate the autocorrelation of the signal. Since the signals of two consecutive beats in the fetal heartbeat signal have a strong similarity, they are amplified by the autocorrelation operation, and other irregular noise signals will become smaller and smaller in the autocorrelation operation. Finally, the autocorrelation function will show a periodicity consistent with the fetal heartbeat law. At this time, the time interval between the two peaks in the autocorrelation function is the time interval of the fetal heartbeat. The correlation curve calculated each time is obtained, and the peak value of the autocorrelation curve is detected; the fetal heart rate is calculated according to the peak position and the corresponding fetal heart rate offset. The calculation formula is as follows:

[0074]

[0075] Among them, Fs is the sampling frequency of the fetal heart signal, dev is the offset corresponding to the fetal heart rate, pos is the peak position of the autocorrelation curve, that is, the peak position obtained during autocorrelation, and FHR is the fetal heart rate.

[0076] The offset dev can be determined by the difference between the peak point calculated by the envelope and the heartbeat time marked by the doctor. Dev+pos can determine the actual fetal heartbeat peak position.

[0077] In a feasible implementation, the target residual network is pre-trained, and the above method may also include a training method for the target residual network, specifically including the following steps B01 to B04:

[0078] B01. Acquire a training sample set, wherein the training sample set includes a third fetal heart signal marked with a real cardiac cycle label;

[0079] In order to allow the residual network to fully learn the characteristics of the real fetal heart signal, a large amount of real fetal heart data is collected in advance, such as a real fetal heart data set collected from the hospital with cardiac cycle labels annotated by doctors. This data set includes a large number of fetal heart signals and their real cardiac cycle labels as training samples. 70% of the data set is used as a training set and 30% as a test set for model training and testing. Manually annotated fetal heart data is used as a comparison standard for evaluating the degree of algorithm optimization.

[0080] B02. Input the third fetal heart signal into a pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network.

[0081] After obtaining the above training samples, the third fetal heart signal as the training sample can be input into the pre-built initial residual network for network training to obtain the fourth fetal heart signal output by the initial residual network.

[0082] In a feasible implementation, the initial residual network may include at least one residual block. In the present application, 17 residual blocks are taken as an example. Each residual block includes at least two convolutional layers connected in series. In the present application, two convolutional layers are taken as an example. Each convolutional layer may perform a convolution operation using preset n convolution kernels. The size of the convolution kernel may be 1*3. Then, the third fetal heart signal is input into the pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network, including steps C01 to C03:

[0083] C01, inputting the third fetal heart signal into the residual block to perform a two-layer convolution operation on the third fetal heart signal to obtain a first signal feature after the convolution operation;

[0084] C02. Adding the first signal feature to the third fetal heart signal to obtain a second signal feature;

[0085] C03. Activate the second signal feature using a preset activation function to obtain the fourth fetal heart signal.

[0086] It can be understood that the third fetal heart signal x is input into the residual block to perform a two-layer convolution operation on the third fetal heart signal to obtain the first signal feature F(x) after the convolution operation; the first signal feature F(x) is added to the third fetal heart signal x to obtain the second signal feature F(x)+x; the second signal feature is activated using a preset activation function (ReLU function) to obtain the fourth fetal heart signal.

[0087] See also Figure 3 , Figure 3: is a schematic diagram of a residual block of a residual network in an embodiment of the present invention, wherein the residual network includes at least one residual block, the residual block includes two convolutional layers Weight layer connected in series, x is the input of the residual block, F(x) is the output of the convolution operation, F(x)+x is the output of the residual block, x identity is the upsampling of x, relu is the activation function, such as Figure 3 As shown, x is the input ultrasonic Doppler fetal heart signal segment, which is a one-dimensional time series signal with a length of L. The fetal heart signal is subjected to two layers of convolution operations with n 1x3 convolution kernels. F(x) is the output after the convolution operation. The input fetal heart signal value needs to be added at the end of the module to ensure that network degradation does not occur after the network is deepened. Among them, F(x) represents the residual between the features that have been learned and the features to be learned, and x represents the features that have been learned in the shallow layer before. As the number of network layers continues to deepen, each time the features learned after adding the network are compared with the residuals before adding them, the model is more sensitive to changes in the output, which facilitates better model training. Furthermore, x identity indicates that when the size of x is different from that of F(x), an upsampling operation is required, and the ReLU function is used for activation after convolution. The output of the network is a denoised fetal heart signal of the same length as the input fetal heart signal x. This application takes the overall network framework of the residual network as an example, connecting 17 residual blocks of the same operation in series.

[0088] B03, determining whether the initial residual network converges based on the fourth fetal heart signal and the real cardiac cycle label;

[0089] Then, the output of the initial residual network, i.e., the fourth fetal heart signal, is used to evaluate whether the network has converged and whether a fetal heart signal with little noise can be output. Specifically, whether the initial residual network has converged is determined based on the fourth fetal heart signal and the real cardiac cycle label, which specifically includes D01 to D03:

[0090] D01, using the fourth fetal heart signal and the real cardiac cycle label to determine the current loss value of the initial residual network;

[0091] D02. If the current loss value is at the minimum value, it is determined that the initial residual network converges, and the step of using the converged initial residual network as the target residual network is performed;

[0092] D03. If the current loss value is not at the minimum value, it is determined that the initial residual network has not converged, and the model parameters of the initial residual network are adjusted, and the step of inputting the third fetal heart signal into the pre-constructed initial residual network to obtain the fourth fetal heart signal output by the initial residual network is returned to execute.

[0093] B04. When the initial residual network converges, the converged initial residual network is used as the target residual network.

[0094] It can be understood that whether it is at the minimum value can be determined by whether the loss value changes or is compared with a preset threshold, which is not limited here. The current loss value is then evaluated to determine the network convergence. If the current loss value is at the minimum value, the initial residual network is determined to have converged, the training is completed, and B04 is executed to obtain the target residual network. On the contrary, if the current loss value is not at the minimum value, it is determined that the initial residual network has not converged and needs to continue training, and then the model parameters of the initial residual network are adjusted. The model parameters can be the weights of the convolutional layer, etc., and the execution of step B02 is returned.

[0095] The present invention provides a method for determining a fetal heart rate, the method comprising: obtaining an original fetal heart signal; dividing the original fetal heart signal into equal lengths to obtain a plurality of first fetal heart signals after division; inputting the first fetal heart signals into a target residual network for noise reduction processing respectively to obtain respective second fetal heart signals output by the target residual network; splicing the respective second fetal heart signals to obtain a target fetal heart signal; performing fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate. Through the above-mentioned method, the pre-trained target residual network can first perform noise reduction processing on the original fetal heart signal to obtain a target fetal heart signal after noise reduction, and then perform fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate, thereby reducing the influence of noise on the calculation of the fetal heart rate, improving the accuracy of the fetal heart rate calculation, and segmenting the signal during noise reduction to reduce the amount of calculation and improve the signal processing efficiency.

[0096] See also Figure 4 , Figure 4 FIG. 1 is a flow chart of a device for determining a fetal heart rate according to an embodiment of the present invention, Figure 4 The device shown includes:

[0097] Signal acquisition module 401: used to obtain the original fetal heart signal;

[0098] Noise reduction processing module 402: used to input the original fetal heart signal into the target residual network to obtain the target fetal heart signal output by the target residual network, and the target residual network is used to perform noise reduction processing on the original fetal heart signal;

[0099] The fetal heart rate determination module 403 is used to perform fetal heart rate determination processing based on the target fetal heart rate signal to obtain a target fetal heart rate.

[0100] It should be noted that Figure 4 The functions of each module in the device shown are Figure 1 The contents of each step in the method shown are similar, and are not described here to avoid repetition. For details, please refer to Figure 1The content of each step in the method shown.

[0101] In a feasible implementation, the device further includes:

[0102] Sample acquisition module: used to acquire a training sample set, wherein the training sample set includes a third fetal heart signal marked with a real cardiac cycle label;

[0103] Network training module: used for inputting the third fetal heart signal into a pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network; determining whether the initial residual network converges based on the fourth fetal heart signal and the real cardiac cycle label; when the initial residual network converges, the converged initial residual network is used as the target residual network.

[0104] It should be noted that the functions of the sample acquisition module and the network training module are similar to the contents of the above steps B01 to B04. To avoid repetition, they are not described here. For details, please refer to the contents of the above steps B01 to B04.

[0105] The present invention provides a device for determining a fetal heart rate, the device comprising: a signal acquisition module: used to obtain an original fetal heart signal; a noise reduction processing module: used to input the original fetal heart signal into a target residual network, and obtain a target fetal heart signal output by the target residual network, the target residual network is used to perform noise reduction processing on the original fetal heart signal; a fetal heart rate determination module: used to perform fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate. Through the above method, the pre-trained target residual network can first perform noise reduction processing on the original fetal heart signal to obtain a noise-reduced target fetal heart signal, and then perform fetal heart rate determination processing based on the target fetal heart signal to obtain a target fetal heart rate, thereby reducing the influence of noise on the calculation of the fetal heart rate and improving the accuracy of the fetal heart rate calculation.

[0106] Figure 5 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 5 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 5The 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.

[0107] 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 or Figure 2 Steps shown.

[0108] 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 or Figure 2 Steps shown.

[0109] 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).

[0110] 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.

[0111] 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 method for determining fetal heart rate, characterized in that: The method comprises: Obtain original fetal heart signal; Inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network, wherein the target residual network is used to perform noise reduction processing on the original fetal heart signal; A fetal heart rate determination process is performed based on the target fetal heart rate signal to obtain a target fetal heart rate.

2. The method according to claim 1, characterized in that Inputting the original fetal heart signal into a target residual network to obtain a target fetal heart signal output by the target residual network includes: Dividing the original fetal heart signal into equal lengths to obtain a plurality of divided first fetal heart signals; Inputting the first fetal heart signals into the target residual network for noise reduction processing respectively, to obtain respective second fetal heart signals output by the target residual network; The second fetal heart signals are spliced ​​together to obtain the target fetal heart signal.

3. The method according to claim 1, characterized in that The performing fetal heart rate determination processing based on the target fetal heart signal to obtain the target fetal heart rate includes: performing low-pass filtering on the target fetal heart signal to obtain an envelope signal corresponding to the target fetal heart signal; performing an autocorrelation operation on the envelope signal to obtain an autocorrelation curve; Performing feature extraction using the autocorrelation curve to obtain a peak position of the autocorrelation curve, wherein the peak position is used to reflect the actual heartbeat moment; The target fetal heart rate is obtained according to the peak position and a preset fetal heart rate algorithm.

4. The method according to claim 1, characterized in that The method further comprises: Acquire a training sample set, wherein the training sample set includes a third fetal heart signal marked with a real cardiac cycle label; Inputting the third fetal heart signal into a pre-built initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network; Determining whether the initial residual network converges based on the fourth fetal heart signal and the true cardiac cycle label; When the initial residual network converges, the converged initial residual network is used as the target residual network.

5. The method according to claim 4, characterized in that: The determining whether the initial residual network converges based on the fourth fetal heart signal and the true cardiac cycle label includes: Determining a current loss value of the initial residual network using the fourth fetal heart signal and the true cardiac cycle label; If the current loss value is at the minimum value, determining that the initial residual network has converged, and performing the step of using the converged initial residual network as the target residual network; If the current loss value is not at the minimum value, it is determined that the initial residual network has not converged, and the model parameters of the initial residual network are adjusted, and the step of inputting the third fetal heart signal into the pre-constructed initial residual network to obtain the fourth fetal heart signal output by the initial residual network is returned to execute.

6. The method according to claim 4, characterized in that: The initial residual network includes at least one residual block, and the residual block includes at least two convolutional layers connected in series. Then, inputting the third fetal heart signal into the pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network includes: Inputting the third fetal heart signal into the residual block to perform a two-layer convolution operation on the third fetal heart signal to obtain a first signal feature after the convolution operation; adding the first signal feature and the third fetal heart signal to obtain a second signal feature; The second signal feature is activated using a preset activation function to obtain the fourth fetal heart signal.

7. A device for determining fetal heart rate, characterized in that: The device comprises: Signal acquisition module: used to obtain original fetal heart signal; Noise reduction processing module: used to input the original fetal heart signal into the target residual network to obtain the target fetal heart signal output by the target residual network, and the target residual network is used to perform noise reduction processing on the original fetal heart signal; Fetal heart rate determination module: used to perform fetal heart rate determination processing based on the target fetal heart rate signal to obtain the target fetal heart rate.

8. The device according to claim 7, characterized in that The device further comprises: Sample acquisition module: used to acquire a training sample set, wherein the training sample set includes the third fetal heart signal marked with a real cardiac cycle label; The network training module is used to input the third fetal heart signal into a pre-constructed initial residual network for network training to obtain a fourth fetal heart signal output by the initial residual network; determine whether the initial residual network has converged based on the fourth fetal heart signal and the true cardiac cycle label; when the initial residual network converges, the converged initial residual network is used as the target residual network.

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 6.

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 is caused to perform the steps of the method according to any one of claims 1 to 6.

Images