Bladder urine control neural circuit urine storage function dynamic sensing method, readable storage medium and device

Abstract

The invention discloses a dynamic sensing method for a urine storage function of a bladder urine control neural circuit, a readable storage medium and a device, and the dynamic sensing method comprises the steps: collecting an electrophysiological signal of at least one channel of a user; according to the electrophysiological signals, a preset neural network model is adopted for calculation, and the bladder urine volume value of the user is obtained. According to the technical scheme, firstly, disposable empty bladder-gt is utilized; the method comprises the following steps of: discretizing continuous electroneurographic signal change into a series of ordered intrinsic state points through unsupervised clustering according to the process data of'full bladder ', automatically generating continuous and personalized pseudo labels for the whole urine storage process, and taking the pseudo labels and corresponding original multi-channel electrophysiological signals as training data, so as to achieve the purpose of training the whole urine storage process. A deep neural network model is subjected to fitting training, a common deep mode exceeding manual features can be captured by learning end-to-end mapping from a multi-channel signal to a urine storage state, and finally, robust and continuous bladder urine storage volume dynamic monitoring of a new input signal is realized.

Description

Technical Field

[0001] This invention relates to the field of medical technology, specifically to a method, readable storage medium, and device for dynamically sensing the urine storage function of a bladder urinary control neural circuit. Background Technology

[0002] In neurological disorders such as benign prostatic hyperplasia, hemiplegia, and spinal cord injury, patients may lose the ability to sense bladder fullness, leading to urinary retention. 59% of severely ill patients are unable to urinate spontaneously and require long-term catheterization to reduce bladder pressure, resulting in the bladder's inability to maintain a full state. The bladder muscles remain in a relaxed state for extended periods, leading to bladder function degeneration, bladder overdistension, and permanent detrusor muscle damage. This also prolongs hospital stays and increases medical costs.

[0003] Currently, commonly used clinical diagnostic and assessment tools for urinary retention mainly focus on two aspects: sensory function and muscle control function. These include imaging cystoscopy, urodynamic testing (UDS), electrosensory threshold testing (EPT) in neurophysiology, and ultrasound measurement of residual urine. While cystoscopy can detect lesions within the bladder, it is an invasive procedure, accompanied by discomfort, infection risks, and high costs. Clinical medical ultrasound, although non-invasive, is expensive, bulky, and its results are highly dependent on the operator, making dynamic functional assessment difficult. UDS is performed under artificial irrigation conditions, often failing to accurately reflect bladder behavior under physiological conditions, and patients frequently experience significant discomfort, leading to poor compliance with follow-up examinations. EPT mainly involves stimulating the perineal or urethral skin surface with surface electrodes to assess the sensitivity of nerves in the lower urinary tract or pelvis, and the results rely on subjective evaluation. Furthermore, local skin electrical stimulation responses cannot characterize the health and integrity of the neural circuits controlling the urinary tract, providing only indirect reference value.

[0004] In situations where continuous calibration using ultrasound is not feasible, how can we utilize neurophysiological signals to achieve non-invasive, continuous, real-time, and accurate dynamic estimation of an individual's bladder capacity, thereby enabling dynamic perception of bladder urge to urinate and solving the problems of invasiveness, bulkiness, and discontinuity in traditional monitoring methods? Summary of the Invention

[0005] Based on the above situation, the main objective of this invention is to provide a non-invasive, continuous, real-time and accurate dynamic monitoring method for bladder urgency using neuroelectrophysiological signals.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for dynamically sensing the urine storage function of a bladder urinary control neural circuit includes the following steps:

[0008] S100, acquire electrophysiological signals from at least one channel of the user, the electrophysiological signals including peripheral nerve electrophysiological signals and central nerve electrophysiological signals of the user's back spinal region;

[0009] S200, based on the electrophysiological signal, a preset neural network model is used to calculate and obtain the user's bladder urine volume value.

[0010] Preferably, the training process of the preset neural network model includes a training dataset acquisition process and a neural network training process. The training dataset acquisition process involves acquiring electrophysiological signals from at least one channel during the user's urine storage process from an empty bladder to a full bladder. An empty bladder is defined as when the residual urine volume in the user's bladder is less than a first capacity threshold, and a full bladder is defined as when the residual urine volume is greater than or equal to a second capacity threshold. The electrophysiological signals are peripheral nerve electrophysiological signals and central nerve electrophysiological signals acquired from the user's back spinal region. The physiological signal is segmented, and a high-level feature set is extracted from each segment to obtain a feature vector sequence of the electrophysiological signal. Unsupervised temporal clustering is performed on the feature vector sequence to obtain N cluster centroids. Based on the N cluster centroids and their timestamps, a pseudo-label for the urine storage volume corresponding to each segment of the electrophysiological signal is calculated. Each segment of the electrophysiological signal is matched one-to-one with the pseudo-label for the urine storage volume to generate a training dataset. The neural network training process involves training the neural network model using cross-validation with mean squared error or average error as the loss function to obtain the preset neural network model.

[0011] Preferably, the electrophysiological signals include signals from the sympathetic nerve segment and sensory nerve ascending pathway segment of the spine in the back of the human body, or signals from the pelvic nerve and pudendal nerve segment.

[0012] Preferably, the process of segmenting the electrophysiological signal and extracting the high-level feature set of each segment in the training dataset acquisition specifically involves: segmenting the electrophysiological signal in a sliding window manner, and extracting the time-frequency domain statistics, variational mode decomposition component energy, and wavelet coefficients of each segment of the electrophysiological signal to form the high-level feature set.

[0013] Preferably, the unsupervised temporal clustering of the feature vector sequence in the training dataset acquisition specifically involves performing K-means unsupervised temporal clustering on the feature vector sequence.

[0014] Preferably, the process of calculating the pseudo-label of urine storage volume corresponding to each segment of the electrophysiological signal based on the N cluster centroids and their timestamps during the acquisition of the training dataset specifically involves: determining the temporal order of the N cluster centroids based on their timestamps; and calculating a curve using a centroid or smooth regression algorithm based on the N cluster centroids and their timestamps. This curve is used to describe the pseudo-label of urine storage volume corresponding to each segment of the electrophysiological signal.

[0015] Preferably, the preset neural network model includes an input layer, a feature extraction layer, and an output layer. The input layer is used to receive the user's electrophysiological signals; the feature extraction layer outputs features from the user's electrophysiological signals; and the output layer is used to perform fitting learning between the features and the urine storage volume pseudo-label through a fully connected layer and a drop-out mechanism, thereby outputting the user's bladder urine volume value.

[0016] Preferably, the feature extraction layer includes a one-dimensional convolutional neural network, a recurrent neural network, or a hybrid network of convolutional neural networks and recurrent neural networks. The one-dimensional convolutional neural network is used to extract local features with translation invariance from the user's electrophysiological signals; the recurrent neural network is used to capture long-term temporal dependencies; the convolutional neural network in the hybrid network is used to extract robust features of each segment of the electrophysiological signal; and the recurrent neural network in the hybrid network is used to capture the variation pattern of the robust features in the time dimension.

[0017] Preferably, when the user's bladder volume is greater than or equal to a preset capacity threshold, it is determined that the user has the urge to urinate.

[0018] The present invention also discloses a computer-readable storage medium storing a data processing program, which, when executed by a processor, implements the steps of the method described in any one of the present invention.

[0019] This invention also discloses a dynamic sensing device for the bladder urinary control neural circuit's urine storage function, comprising an electrophysiological signal acquisition module and a urine volume calculation module. The electrophysiological signal acquisition module is used to acquire at least one channel of electrophysiological signals from the user, including peripheral nerve electrophysiological signals and central nerve electrophysiological signals from the user's back spinal region. The urine volume calculation module is used to calculate the user's bladder urine volume value based on the electrophysiological signals using a preset neural network model. The training process of the preset neural network model includes a training dataset acquisition process and a neural network training process. The training dataset acquisition process involves acquiring at least one channel of electrophysiological signals from the user's urine storage process from an empty bladder state to a full bladder state. An empty bladder state is defined as when the residual urine volume in the user's bladder is less than a first capacity threshold, as detected by ultrasound. When the residual urine volume is greater than or equal to the second capacity threshold, it is considered a full bladder state. The electrophysiological signals are peripheral nerve electrophysiological signals and central nerve electrophysiological signals collected from the spinal region of the user's back. The electrophysiological signals are segmented, and the high-level feature set of each segment is extracted to obtain the feature vector sequence of the electrophysiological signals. Unsupervised temporal clustering is performed on the feature vector sequence to obtain N cluster centroids, each centroid representing a neural electrical activity state during the urine storage process. The urine storage volume pseudo-label corresponding to each segment of the electrophysiological signal is calculated based on the N cluster centroids and the timestamps of the N cluster centroids. Each segment of the electrophysiological signal is matched one-to-one with the urine storage volume pseudo-label to generate a training dataset. Neural network training process: The neural network model is trained using cross-validation with mean squared error or average error as the loss function to obtain the preset neural network model.

[0020] The technical solution of this invention first utilizes the data from a single "empty bladder -> full bladder" process. Through unsupervised clustering, the continuous changes in neural electrical signals are discretized into a series of ordered intrinsic state points, automatically generating continuous and personalized pseudo-labels for the entire urine storage process. By using these pseudo-labels and the corresponding original multi-channel electrophysiological signals as training data, a deep neural network model is fitted and trained. By learning the end-to-end mapping from multi-channel signals to urine storage state, it can capture common deep patterns that surpass manual features, ultimately achieving robust and continuous dynamic monitoring of bladder urine storage volume for new input signals.

[0021] Other beneficial effects of the present invention will be explained in detail through the introduction of specific technical features and technical solutions in specific embodiments. Those skilled in the art should be able to understand the beneficial technical effects brought about by these technical features and technical solutions through the introduction of these technical features and technical solutions. Attached Figure Description

[0022] The preferred embodiment of the dynamic sensing method and device for bladder urinary control neural circuit storage function according to the present invention will be described below with reference to the accompanying drawings. In the drawings:

[0023] Figure 1 This is a flowchart of a method for dynamically sensing the urine storage function of a bladder urinary control neural circuit according to a preferred embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the dynamic sensing method for bladder control neural circuit storage function according to a preferred embodiment of the present invention, used to monitor bladder urine volume in healthy individuals.

[0025] Figure 3 This is a block diagram of a dynamic sensing device for the urine storage function of a bladder urinary control neural circuit according to a preferred embodiment of the present invention. Detailed Implementation

[0026] To provide a more detailed description of the technical solutions of this application and to facilitate a better understanding of this application, specific embodiments of this application are described below in conjunction with the accompanying drawings. However, it should be understood that all illustrative embodiments and their descriptions are used to explain this application and do not constitute the sole limitation of this application.

[0027] Figure 1 The flowchart of a dynamic sensing method for bladder urinary control neural circuit storage function according to a preferred embodiment of the present invention includes: step S100, acquiring electrophysiological signals of at least one channel of the user, the electrophysiological signals including peripheral nerve electrophysiological signals and central nerve electrophysiological signals of the user's back spinal region; step S200, calculating the user's bladder urine volume value based on the electrophysiological signals using a preset neural network model.

[0028] In a preferred embodiment, the training process of the preset neural network model includes a training dataset acquisition process and a neural network training process. The training dataset acquisition process involves acquiring electrophysiological signals from at least one channel during the user's urine storage process from an empty bladder to a full bladder. An empty bladder state is defined as when the residual urine volume in the user's bladder is less than a first capacity threshold, and a full bladder state is defined as when the residual urine volume is greater than or equal to a second capacity threshold. The electrophysiological signals are peripheral nerve electrophysiological signals and central nerve electrophysiological signals acquired from the user's back spinal region. The electrophysiological signals are segmented and... The high-level feature set of each segment is taken to obtain the feature vector sequence of the electrophysiological signal; unsupervised temporal clustering is performed on the feature vector sequence to obtain N cluster centroids, where each centroid typically represents a neural electrical activity state during the urine storage process; based on the N cluster centroids and their timestamps, a pseudo-label of the urine storage volume corresponding to each segment of the electrophysiological signal is calculated; each segment of the electrophysiological signal is mapped one-to-one with the pseudo-label of the urine storage volume to generate a training dataset; the neural network training process involves training the neural network model using cross-validation with mean squared error or average error as the loss function to obtain the preset neural network model.

[0029] The technical solution of this invention first utilizes the data from a single "empty bladder -> full bladder" process. Through unsupervised clustering, the continuous changes in neural electrical signals are discretized into a series of ordered intrinsic state points, automatically generating continuous and personalized pseudo-labels for the entire urine storage process. By using these pseudo-labels and the corresponding original multi-channel electrophysiological signals as training data, a deep neural network model is fitted and trained. By learning the end-to-end mapping from multi-channel signals to urine storage state, it can capture common deep patterns that surpass manual features, ultimately achieving robust and continuous dynamic monitoring of bladder urine storage volume for new input signals.

[0030] In a preferred embodiment, the electrophysiological signals include signals from the sympathetic nerve segments and sensory nerve ascending pathways of the spine in the back, or signals from the pelvic nerve and pudendal nerve segments. Specifically, the detection location of the electrophysiological signals can be determined based on the location of the damaged nerves in neuronal urinary tract diseases.

[0031] In a specific implementation, the first volume threshold can be 50 mL, and the second volume threshold can be 450 mL. That is, in an "empty bladder" state: after the user empties their bladder, the residual urine volume verified by ultrasound is <50 ml. In a "full bladder" state: the user's urine volume calibrated by ultrasound is ≥450 ml. In a specific implementation, when the user empties their bladder and experiences a strong urge to urinate, the urine volume can be detected using ultrasound imaging to determine if it meets the above standards.

[0032] In a preferred embodiment, the entire continuously acquired electrophysiological signal can be segmented using a sliding window (e.g., a 5-second window), and a high-level feature set can be extracted from each window. This high-level feature set may include time-frequency domain statistics of the multi-channel electrophysiological signal, component energies of variational mode decomposition, and wavelet coefficients, etc. In this way, the continuous multi-channel electrophysiological signal is transformed into a feature vector sequence.

[0033] In a preferred embodiment, the unsupervised temporal clustering of the feature vector sequence during training set acquisition can specifically be: performing K-means unsupervised temporal clustering on the feature vector sequence. K-means unsupervised temporal clustering of the feature vector sequence yields N cluster centroids (C1, C2, ..., CN), each centroid representing a specific neural electrical activity state during urine storage. Specifically, the number K of K-means unsupervised temporal clusters can be determined according to the Bayesian Information Criterion (BIC).

[0034] In a preferred embodiment, the calculation of the pseudo-label for urine storage volume corresponding to each segment of the electrophysiological signal based on the N cluster centroids and their timestamps during training dataset acquisition can specifically be as follows: Based on the timestamps of the N cluster centroids, determine the temporal order of their appearance (i.e., trajectory: C1->C2->...->CN); based on the N cluster centroids and their timestamps, calculate a curve using a centroid or smoothing regression algorithm (such as autoregression). This curve describes the pseudo-label for urine storage volume corresponding to each segment of the electrophysiological signal. In other words, it describes the process of urinary urgency changing from "no urge" to "strong urge." Thus, a corresponding, continuous pseudo-label for "urinary storage volume" is generated for each time window in the original electrophysiological signal. Associating the original electrophysiological signal segments (X1-XN) with the generated pseudo-labels forms a training dataset {X1-XN, Label}.

[0035] In a preferred embodiment, the preset neural network model may include an input layer, a feature extraction layer, and an output layer. The input layer is used to receive the user's electrophysiological signals, the feature extraction layer outputs the output of the user's electrophysiological signals to extract features, and the output layer is used to perform fitting learning between the features and the pseudo-label of urine storage volume through a fully connected layer and a drop-out mechanism, thereby outputting the user's bladder urine volume value.

[0036] In a specific implementation, the input layer may receive electrophysiological signals within a time window, or it may receive multi-channel electrophysiological signals after preliminary preprocessing, such as electrophysiological signals after amplification and filtering.

[0037] In a preferred embodiment, the feature extraction layer may include a one-dimensional convolutional neural network (1D-CNN), a recurrent neural network (RNN / LSTM), or a hybrid network of convolutional neural networks (CNN) and recurrent neural networks (RNN / LSTM). When using a one-dimensional convolutional neural network (1D-CNN), it is used to extract translation-invariant local features from the user's electrophysiological signals. When using a recurrent neural network (RNN / LSTM), it is used to capture long-term temporal dependencies. When using a hybrid network, the convolutional neural network in the hybrid network is used to extract robust features for each segment of the electrophysiological signal, and the recurrent neural network in the hybrid network is used to capture the temporal variation patterns of the robust features. In other embodiments, other neural networks may be selected as needed.

[0038] In other implementations, multi-dimensional features can be extracted from the user's electrophysiological signals, such as time-domain features, frequency-domain features, nonlinear decomposition features, and linear decomposition features. Time-domain features may include the mean, variance, skewness, kurtosis, and zero-crossing rate of the electrophysiological signal. Frequency-domain features may include the energy, centroid frequency, and spectral entropy of the electrophysiological signal's power spectral density within a preset frequency band. Nonlinear and linear decomposition features may include the energy and entropy of the intrinsic mode functions (EMFs) of the electrophysiological signal. The preset frequency band may be, for example, Delta, Theta, Alpha, or Beta bands. The energy and entropy of the EMFs of the electrophysiological signal can be obtained through variational mode decomposition (VMD). Alternatively, the linear features of the electrophysiological signal can be obtained by calculating the energy and statistical characteristics of the wavelet transform coefficients at various scales.

[0039] In specific implementations, the feature extraction layer may further include a channel adjustment module, which can employ the Squeeze-and-Excitation Network (SE) mechanism. The channel adjustment module is a component in deep learning models used to dynamically and adaptively recalibrate feature responses along the channel dimension (i.e., the depth of the feature map). The SE module can be flexibly embedded into neural network architectures and incurs only a small computational overhead, yet can significantly improve model performance.

[0040] In a specific implementation, the preset neural network model trained using the training dataset will learn the following function:

[0041] F(X)≈Label

[0042] Wherein, X represents the collected user electrophysiological signal, and F(X) represents the calculated user bladder volume value. That is, by inputting any segment of electrophysiological signal, the preset neural network model can directly output the estimated bladder volume value.

[0043] like Figure 2 As shown, the method of the present invention is used to monitor urine volume in a healthy person from an empty bladder state to a full bladder state. In this paper, label is a pseudo-label for "urine storage volume", predict is the estimated bladder urine volume value of the healthy person according to the method of the present invention, and windows is the cluster centroid window. The exemplary value in the figure is 20. The mean error (MAE) is less than 9%, and the accuracy can reach 91%.

[0044] In a preferred embodiment, when the user's bladder volume is greater than or equal to a preset capacity threshold, it is determined that the user has the urge to urinate. This allows for a urination reminder to the user.

[0045] The present invention also discloses a computer-readable storage medium storing a data processing program, which, when executed by a processor, implements the steps of the method described in any one of the present invention.

[0046] This invention also discloses a device for dynamic monitoring of bladder urine volume, such as... Figure 3 As shown, the system includes an electrophysiological signal acquisition module 100 and a urine volume calculation module 200. The electrophysiological signal acquisition module 100 is used to acquire at least one channel of electrophysiological signals from the user, including peripheral nerve electrophysiological signals and central nerve electrophysiological signals from the user's spinal region. The urine volume calculation module 200 is used to calculate the user's bladder urine volume value based on the electrophysiological signals using a preset neural network model. The training process of the preset neural network model includes a training dataset acquisition process and a neural network training process.

[0047] Training dataset acquisition process: Electrophysiological signals from at least one channel are acquired during the user's urine storage process, from an empty bladder to a full bladder. An empty bladder is defined as when the residual urine volume in the user's bladder is less than a first capacity threshold, and a full bladder is defined as when the residual urine volume is greater than or equal to a second capacity threshold. The electrophysiological signals are peripheral nerve electrophysiological signals and central nerve electrophysiological signals acquired from the user's back spinal region. The electrophysiological signals are segmented, and a high-level feature set is extracted from each segment to obtain a feature vector sequence. Unsupervised temporal clustering is performed on the feature vector sequence to obtain N cluster centroids, each centroid representing a neural electrical activity state during the urine storage process. A pseudo-label for the urine storage volume corresponding to each segment of the electrophysiological signal is calculated based on the N cluster centroids and their timestamps. Each segment of the electrophysiological signal is mapped one-to-one with the pseudo-label for the urine storage volume to generate a training dataset.

[0048] Neural network training process: The neural network model is trained using cross-validation with mean squared error or average error as the loss function to obtain the preset neural network model.

[0049] It should be noted that the aforementioned computer-readable media may include, but is not limited to: volatile memory, such as random access memory (RAM); non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); and combinations of the above types of memory.

[0050] In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0051] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0052] It should be noted that the use of step numbers (letters or numbers) to refer to certain specific method steps in this invention is merely for the purpose of convenience and brevity in description, and is by no means intended to restrict the order of these method steps. Those skilled in the art will understand that the order of the relevant method steps should be determined by the technology itself and should not be unduly restricted by the existence of step numbers.

[0053] Those skilled in the art will understand that, without conflict, the above-mentioned preferred solutions can be freely combined and superimposed.

[0054] It should be understood that the above embodiments are merely exemplary and not restrictive. Various obvious or equivalent modifications or substitutions that can be made by those skilled in the art regarding the above details without departing from the basic principles of the present invention will be included within the scope of the claims of the present invention.

Claims

1. A method for dynamically sensing the urine storage function of a bladder urinary control neural circuit, characterized in that, Including the following steps: S100, acquire electrophysiological signals from at least one channel of the user, the electrophysiological signals including peripheral nerve electrophysiological signals and central nerve electrophysiological signals of the user's back spinal region; S200, based on the electrophysiological signal, a preset neural network model is used to calculate and obtain the user's bladder urine volume value.

2. The method for dynamic sensing of urine storage function of the bladder urinary control neural circuit according to claim 1, characterized in that, The training process of the preset neural network model includes a training dataset acquisition process and a neural network training process; The training dataset acquisition process involves: acquiring electrophysiological signals from at least one channel during the user's urine storage process from an empty bladder to a full bladder; an empty bladder state is defined as when the residual urine volume in the user's bladder is less than a first capacity threshold, and a full bladder state is defined as when the residual urine volume is greater than or equal to a second capacity threshold; the electrophysiological signals are peripheral nerve electrophysiological signals and central nerve electrophysiological signals acquired from the user's back spinal region; the electrophysiological signals are segmented, and a high-level feature set is extracted from each segment to obtain a feature vector sequence; unsupervised temporal clustering is performed on the feature vector sequence to obtain N cluster centroids; and a pseudo-label for the urine storage volume corresponding to each segment of the electrophysiological signal is calculated based on the N cluster centroids and their timestamps. Each electrophysiological signal segment is matched one-to-one with the urine storage volume pseudo-label to generate a training dataset. The neural network training process involves training the neural network model using cross-validation with mean squared error or average error as the loss function to obtain the preset neural network model.

3. The method for dynamic sensing of urine storage function of the bladder urinary control neural circuit according to claim 1, characterized in that, The electrophysiological signals include signals from the sympathetic nerve segments and sensory nerve ascending pathway segments of the spine in the back of the human body, or signals from the pelvic nerve and pudendal nerve segments.

4. The method for dynamic sensing of urine storage function of the bladder urinary control neural circuit according to claim 1, characterized in that, The electrophysiological signal was segmented during the acquisition of the training dataset, and the high-level feature set of each segment was extracted specifically as follows: The electrophysiological signal is segmented using a sliding window approach, and the time-frequency domain statistics, variational mode decomposition component energy, and wavelet coefficients of each segment are extracted to form the high-level feature set.

5. The method for dynamic sensing of urine storage function of the bladder urinary control neural circuit according to claim 1, characterized in that, The unsupervised temporal clustering of the feature vector sequence in the training dataset collection specifically involves performing K-means unsupervised temporal clustering on the feature vector sequence.

6. The method for dynamic sensing of urine storage function of the bladder urinary control neural circuit according to claim 1, characterized in that, The pseudo-labels for urine storage volume corresponding to each segment of the electrophysiological signal, calculated based on the N cluster centroids and their timestamps during the training dataset acquisition, are specifically as follows: Based on the timestamps of the N cluster centroids, determine the temporal order in which the N cluster centroids appear. Based on the N cluster centroids and the timestamps of the N cluster centroids, a curve is calculated using a centroid or smooth regression algorithm. The curve is used to describe the pseudo-label of urine storage volume corresponding to each segment of the electrophysiological signal.

7. The method for dynamic sensing of urine storage function of the bladder urinary control neural circuit according to claim 1, characterized in that, The preset neural network model includes an input layer, a feature extraction layer, and an output layer. The input layer is used to receive the user's electrophysiological signals; The output of the feature extraction layer is used to extract features from the user's electrophysiological signals; The output layer is used to learn the fit between features and the urine storage pseudo-label through a fully connected layer and a drop-out mechanism, thereby outputting the user's bladder urine volume value; The feature extraction layer includes a one-dimensional convolutional neural network, a recurrent neural network, or a hybrid network of convolutional neural networks and recurrent neural networks. The one-dimensional convolutional neural network is used to extract local features with translation invariance from the user's electrophysiological signals; The recurrent neural network is used to capture long-term time dependencies; The convolutional neural network in the hybrid network is used to extract robust features of each segment of the electrophysiological signal, and the recurrent neural network in the hybrid network is used to capture the variation pattern of the robust features in the time dimension.

8. The method for dynamic sensing of bladder urinary control neural circuit storage function according to any one of claims 1-7, characterized in that, When the user's bladder volume is greater than or equal to a preset capacity threshold, it is determined that the user has the urge to urinate.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a data processing program, which, when executed by a processor, implements the steps of the method as described in any one of claims 1-8.

10. A dynamic sensing device for the urine storage function of a bladder urinary control neural circuit, characterized in that, Includes an electrophysiological signal acquisition module and a urine volume calculation module. The electrophysiological signal acquisition module is used to acquire at least one channel of electrophysiological signals from the user, including peripheral nerve electrophysiological signals and central nerve electrophysiological signals from the user's back spinal region. The urine volume calculation module is used to calculate the user's bladder urine volume value based on the electrophysiological signal using a preset neural network model; wherein, the training process of the preset neural network model includes a training dataset acquisition process and a neural network training process. Training dataset acquisition process: Electrophysiological signals from at least one channel are acquired during the user's urine storage process, from an empty bladder to a full bladder. An empty bladder is defined as when the residual urine volume in the user's bladder is less than a first capacity threshold, and a full bladder is defined as when the residual urine volume is greater than or equal to a second capacity threshold. The electrophysiological signals are peripheral nerve electrophysiological signals and central nerve electrophysiological signals acquired from the user's back spinal region. The electrophysiological signals are segmented, and a high-level feature set is extracted from each segment to obtain a feature vector sequence. Unsupervised temporal clustering is performed on the feature vector sequence to obtain N cluster centroids, each centroid representing a neural electrical activity state during the urine storage process. A pseudo-label for the urine storage volume corresponding to each segment of the electrophysiological signal is calculated based on the N cluster centroids and their timestamps. Each segment of the electrophysiological signal is mapped one-to-one with the pseudo-label for the urine storage volume to generate a training dataset. Neural network training process: The neural network model is trained using cross-validation with mean squared error or average error as the loss function to obtain the preset neural network model.

Images