Enterostomy bag excrement monitoring device and monitoring method

Abstract

The invention relates to the technical field of excrement detection, in particular to an enterostomy bag excrement monitoring device and method. The enterostomy bag excrement monitoring device comprises an ostomy bag provided with an inlet and used for being connected to an enterostomy of a patient; the monitor comprises at least two monitoring electrodes, the two monitoring electrodes are located on the two sides of the ostomy bag, and the two monitoring electrodes form a capacitance sensor in the ostomy bag; and the information collection device is in signal connection with the monitor and receives capacitance information of the monitor in real time. According to the technical scheme provided by the invention, compared with the prior art, the problem that the analog data representing the data of the excreta content received by the sensor is directly converted into the excreta capacity and is influenced by an abnormal value is solved; according to the scheme, the capacitance data sequence is regressed to the label representing the true value of the excrement capacity by using the prediction model, so that the accuracy of excrement capacity monitoring is improved.

Description

Technical Field

[0001] The present application relates to the technical field of excrement detection, and in particular to an enterostomy bag excrement monitoring device and a monitoring method. Background Art

[0002] For patients who have just completed ostomy surgery, an enterostomy bag is usually required to collect their excreta. However, patients with enterostomy bags need to pay attention to whether the excreta exceeds the threshold, which brings additional burdens to patients, such as when using them in public places and at night.

[0003] Patent CN215020155U discloses a product for patients undergoing ostomy surgery for rectal cancer. The weight of the ostomy bag body can be weighed by a lap plate. At the same time, the weight of the ostomy bag body can be well supported by the cooperation of a support rod, a pressure sensor and an alarm, thereby automatically weighing the weight of the stool in the ostomy bag body.

[0004] Patent CN208641394U discloses an ileostomy enema device. The device consists of a stoma plug and an infusion device connected by a quick connector, and an alarm is realized by a pressure sensor installed on the end cover. The above methods all detect the content of excrement by weighing the excrement.

[0005] No matter whether the above technology realizes excrement detection by detecting the weight change or expansion of the enterostomy bag itself, it is necessary to add a sleeve layer outside the enterostomy bag to form a relative force between the enterostomy bags. The sensor is not directly integrated with the enterostomy bag and it is an indirect measurement of excrement in the enterostomy bag. Since there may be poor contact between the sensor and the enterostomy bag, the sensor cannot accurately sense the weight change or expansion of the enterostomy bag.

[0006] Therefore, in practice, the excrement content represented by the collected relevant data is a changing value. There is no effective way in the current scheme to summarize or find out the actual content of the current excrement from the changing value of the excrement content, so the accuracy of the current monitoring of the excrement content in the ostomy bag is not high. Summary of the invention

[0007] The content of this application is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this application is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0008] As a first aspect of the present application, in order to solve the technical problems mentioned in the above background technology section, the present application provides an enterostomy bag feces monitoring device, comprising:

[0009] An ostomy bag having an access opening for connection to the patient's intestinal stoma;

[0010] A monitor, comprising at least two monitoring electrodes, the two monitoring electrodes being located on both sides of the stoma, the two monitoring electrodes forming a capacitive sensor in the stoma bag;

[0011] The information collection device is connected to the monitor signal, receives the capacitance information of the monitor in real time, and obtains a capacitance data sequence that changes with time;

[0012] An information processing device, signal-connected to the information collecting device, wherein the information processing device has a built-in prediction model;

[0013] The information processing device receives the capacitance data sequence, and based on the built-in prediction model, the information processing device selects a true value representing the current actual excrement capacity from the capacitance data sequence to obtain the real-time excrement content.

[0014] In the technical solution provided in the present application, compared with the existing technical solution, the analog data received by the sensor that represents the excrement content is directly converted into excrement capacity, which will be affected by abnormal values. In this solution, a prediction model is used to regress the capacitance data sequence to a label that represents the true value of the excrement capacity, thereby increasing the accuracy of excrement capacity monitoring.

[0015] Furthermore, the information collection device prepares all capacitance data generated within a preset time period into a capacitance data sequence to obtain multiple capacitance data sequences; each capacitance data sequence is taken as a sample.

[0016] In the technical solution provided by the present application, the capacitance data generated within a preset time period is taken as a sample. Therefore, for each sample, after inputting it into the prediction model, the corresponding excrement value can be obtained. Moreover, the adoption of this solution can avoid the influence of the actual value of the excrement capacity on the prediction accuracy during the process of excrement increase.

[0017] Furthermore, the last 80% of the capacitance data sequence of the adjacent previous sample overlaps with the first 80% of the capacitance data sequence of the adjacent next sample.

[0018] In this solution, by making the capacitance data sequences in two adjacent samples overlap with each other, it is possible to increase the real-time monitoring performance while ensuring that there are sufficient samples to support the generation of each prediction value.

[0019] Furthermore, the information processing device includes:

[0020] A preprocessing unit, connected to the information collection device signal, preprocesses the received capacitance data sequence;

[0021] The neural network unit has a built-in prediction model. The preprocessed capacitance data sequence is input into the neural network unit to obtain the true value of the current actual excrement capacity.

[0022] Furthermore, the prediction model includes:

[0023] An input layer, for receiving a capacitance data sequence;

[0024] Hidden layer: the hidden layer has at least 3 layers, and each hidden layer has 10 neurons;

[0025] The output layer is connected to the hidden layer signal and uses the output activation function to output the corresponding true value.

[0026] Furthermore, the loss function of the prediction model is M;

[0027] ;

[0028] Where n represents the number of samples, represents the predicted value of the i-th sample, y i Represents the actual value of the i-th sample.

[0029] In the technical solution provided in the present application, the mean square error is used as the loss function to avoid falling into the local minimum during the optimization process, and to find the global minimum. This feature makes it simple and efficient to solve the MSE through optimization algorithms such as gradient descent, thereby increasing the accuracy of the prediction.

[0030] Furthermore, the preprocessing module normalizes the data.

[0031] Furthermore, the training method of the prediction model includes the following steps:

[0032] S1: Obtain at least 1000 samples in advance, and mark the actual excrement volume in each sample;

[0033] S2: preprocessing each capacitance data sequence using a preprocessing unit;

[0034] S3: Divide the 1000 capacitance data sequences into training set, validation set and test set;

[0035] S4; Set initial parameters for the prediction model, which will be optimized during the training process.

[0036] S5: Input the capacitance data sequence in the training set into the prediction model, and continuously adjust the model parameters through the optimization algorithm to minimize the loss function.

[0037] When predicting the volume of excrement, a high degree of accuracy is required to avoid excessive excrement in the ostomy bag and infection. The prediction accuracy of excrement is limited by the model structure. The more complex the model structure, the stronger the model learning ability, and the easier it is to find patterns in the sample. However, if the model structure is too complex, it will take a long time and consume a lot of electricity during calculation, which is not convenient to carry around. To this end, this application provides the following technical solutions:

[0038] Furthermore, it also includes a training device, wherein the prediction model is input into the training device for training;

[0039] The training device comprises:

[0040] An input unit, used to input samples;

[0041] A first high-order unit, having a first high-order network model built therein;

[0042] A second high-order unit, having a second high-order network model built in;

[0043] The training unit contains the prediction model to be trained;

[0044] The input unit sends the samples to the first high-order unit, the second high-order unit and the training unit respectively;

[0045] Among them, the first high-order network model and the second high-order network model have the same structure, and the complexity of the network structure of the prediction model is less than that of the first high-order network model;

[0046] The first high-order network model is used to guide the implicit information of the reconstruction of the prediction model;

[0047] The second high-order model is used to guide the prediction model to capture detailed information.

[0048] In the technical solution provided in the present application, when training the prediction model, two high-order network models are used to guide the training process of the prediction model. One high-order network model guides the reconstruction of implicit information, and the other high-order network model guides the capture of detail information. Therefore, although the network structure of the prediction model finally used is simpler, it can utilize the learning ability of the high-order network model during the training process, and thus has sufficient accuracy. At the same time, in actual use, the prediction model has low energy consumption and is easy to be set in a low-energy ostomy bag so that the patient can carry it with him.

[0049] As a second aspect of the present application, the present application provides a method for monitoring excrement from an enterostomy bag, which uses the aforementioned enterostomy bag excrement monitoring device to monitor the excrement volume in the enterostomy bag.

[0050] The beneficial effects of this application are:

[0051] In the technical solution provided in the present application, compared with the existing technical solution, the analog data received by the sensor that represents the excrement content is directly converted into excrement capacity, which will be affected by abnormal values. In this solution, a prediction model is used to regress the capacitance data sequence to a label that represents the true value of the excrement capacity, thereby increasing the accuracy of excrement capacity monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The illustrative embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application.

[0053] In addition, throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the components and elements are not necessarily drawn to scale.

[0054] In the attached picture:

[0055] Figure 1 This is a schematic diagram of the structure of an enterostomy bag fecal matter monitoring device.

[0056] Figure 2 This is a graph showing the change trend of capacitance value after excrement enters the ostomy bag.

[0057] Figure 3 This is a scatter plot of relative error in content prediction.

[0058] Figure 4 It is a schematic diagram of the structure of the first high-order network model, the second high-order network model and the hidden layer of the prediction model.

[0059] Figure 5 This is the frequency distribution diagram of relative error in content prediction. DETAILED DESCRIPTION

[0060] Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms, and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0061] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0062] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0063] Example 1: Reference Figure 1 , a device for monitoring excrement from an intestinal ostomy bag, comprising an intestinal ostomy bag, a monitor, an information collection device, and an information processing device. Among them, the intestinal ostomy bag is provided with an entrance for connecting to the intestinal ostomy of a patient. The structure of the intestinal ostomy bag itself is a common intestinal ostomy bag, and the excrement in the intestinal ostomy of the patient flows directly into the intestinal ostomy bag. The monitor comprises two monitoring electrodes, and the monitoring electrodes comprise a positive electrode and a negative electrode, and the positive electrode and the negative electrode are respectively located inside the intestinal ostomy bag and on both sides of the intestinal ostomy bag. Therefore, the negative electrode and the positive electrode form a capacitive sensor. When excrement enters the intestinal ostomy bag, because the conductivity of the excrement is much greater than that of the air, the dielectric constant monitored by the capacitive sensor increases, and the capacitance value increases. As the excrement increases, the air discharged from the intestinal ostomy bag also increases, which in turn causes the dielectric constant between the front and back electrodes to become larger and larger, and the capacitance to become larger and larger. Therefore, the capacitance value can be used to determine the excrement capacity in the intestinal ostomy bag.

[0064] The above is the principle of using a monitor to measure the volume of excrement in an ostomy bag in this application. How to convert the capacitance value or dielectric constant into the actual volume of excrement is carried out in the subsequent method.

[0065] Because patients still need to live a normal life after wearing an ostomy bag, the excrement entering the ostomy bag will shake due to the patient's movement. This shaking process will cause the capacitance value to change, so in practice, it is difficult to distinguish whether the change in capacitance value is caused by the increase in excrement or the shaking of excrement. This makes it difficult to monitor the excrement volume.

[0066] To this end, the present application provides an information collection device and an information processing device. The information collection device is arranged on the ostomy bag, and is used to receive the capacitance information of the monitor in real time, and obtain a capacitance data sequence that changes with time. The information processing device is used to convert the capacitance data sequence that changes with time into the excrement volume in the ostomy bag. The information processing device can be arranged on the ostomy bag.

[0067] In this solution, a prediction model is used to process the received capacitance data sequence to obtain the actual excrement volume in the ostomy bag. To this end, the capacitance data sequence needs to be processed.

[0068] In some implementations: the information collection device generates all capacitance data generated within a preset time period as a capacitance data sequence to obtain multiple capacitance data sequences; each capacitance data sequence is used as a sample.

[0069] Because the volume of excreta in the ostomy bag increases gradually, only the capacitance data in the previous period of time is of reference value for the current excreta volume. The excreta volume represented by capacitance data that is too far away is not actually the current true value.

[0070] To this end, in this solution, all capacitance data generated within a preset time period are made into a capacitance data sequence, so each sample is a sequence of fixed length.

[0071] The length of the preset time period can be set in advance, or it can be set according to the rate of excrement production. For example, the total volume of the ostomy bag is 300ml, and it needs to be replaced when it reaches 250ml. For this reason, for safety reasons, it is necessary to complete a feces capacity prediction within the time it takes for the excrement to fill the 50ml. If the patient produces 50ml of excrement per hour, the excrement capacity prediction needs to be completed at least once within an hour. The above is to set the preset time period based on safety. In addition, the preset time period can also be set according to accuracy. For example, the accuracy required in this solution cannot exceed 10ml, so the maximum value of the preset time period cannot be greater than the error value divided by the excrement generation rate. The excrement generation rate of each patient is inconsistent, so the length of the preset time period can be personalized according to the individual's excrement generation rate, or it can be set uniformly.

[0072] In some other implementations, the last 80% of the capacitance data sequence of the adjacent previous sample overlaps with the first 80% of the capacitance data sequence of the adjacent next sample.

[0073] This embodiment is based on the above embodiment, allowing adjacent samples to have overlapping parts, so the distance between samples can be set smaller, and then during monitoring, the distance between the generated monitoring data is smaller, which can better monitor the excrement volume in real time.

[0074] Furthermore, when converting capacitance into excrement capacity, it actually also includes body posture information. However, body posture information is generally difficult to accurately indicate the state of excrement. For example, when the ostomy bag is placed vertically, the excrement is below the ostomy bag, and only the capacitive sensing unit below the ostomy bag can contact the excrement. When the ostomy bag is lying flat, the contact area between the excrement and the capacitive sensing unit will increase, so the current position of the excrement in the ostomy bag can be obtained by monitoring the placement direction of the ostomy bag, but the problem of excrement hanging on the wall will still exist when using this for accurate calculation. In other words, excrement is not a pure liquid. When the human body is in motion, excrement will remain on the inner wall of the ostomy bag, thereby affecting the judgment of the liquid level position and liquid level height of the excrement. In this way, it is necessary to use a network model to empirically analyze the capacity of excrement. If the body posture of the ostomy bag is directly used as an additional sample, the two pieces of information will be split from each other. For this reason, the present application provides the following technical solutions:

[0075] The stoma bag is provided with a gyroscope, which records the state R of the stoma bag.

[0076] A gyroscope is a sensor that senses the direction of gravity. The state of the ostomy bag can be obtained by using the gyroscope. The state in this scheme is mainly divided into r states based on the angle between the top of the ostomy bag and the vertical line. For example, r=3, which means that the first state is that the top of the ostomy bag is facing upward, the second state is that the top of the ostomy bag is horizontal, and the third state is that the ostomy bag is facing downward. The larger the corresponding r, the more states the ostomy bag has.

[0077] The capacitance data with the same state in the sample are extracted to generate r capacitance data columns; the state type is used as the ordinate, the time is used as the abscissa, and the capacitance value is used as the pixel value to generate a capacitance image.

[0078] Before adding state R, the sample is two-dimensional information, the first dimension is capacitance, and the second dimension is time. After adding the state, it changes from two-dimensional information to three-dimensional information. To this end, time and state are used as horizontal and vertical coordinates respectively, and capacitance is converted into a capacitance image as a pixel value, which can accurately show the distribution of capacitance data at different times and different states within a preset time period.

[0079] Therefore, in this solution, the final sample obtained is a capacitance image including state, time and capacitance value.

[0080] The following describes the process of processing samples by the information processing device.

[0081] The information processing device includes a preprocessing unit and a neural network unit. The preprocessing unit is connected to the information collection device signal and preprocesses the received capacitance data sequence. The preprocessing mainly includes normalization processing, outlier removal and other conventional operations. The preprocessing is mainly for normalization processing of capacitance values.

[0082] The neural network unit has a built-in prediction model. The preprocessed capacitance data sequence is input into the neural network unit to obtain the true value of the current actual excrement capacity.

[0083] Furthermore, the prediction model includes:

[0084] An input layer, for receiving a capacitance data sequence;

[0085] Hidden layer: the hidden layer has at least 3 layers, and each hidden layer has 10 neurons;

[0086] The output layer is connected to the hidden layer signal and uses the output activation function to output the corresponding real value.

[0087] Furthermore, the loss function of the prediction model is M;

[0088] ;

[0089] Where n represents the number of samples, represents the predicted value of the i-th sample, y i Represents the actual value of the i-th sample.

[0090] The above is the specific structure of the prediction model and the corresponding loss function. The following is the training process of the prediction model:

[0091] The training method of the prediction model includes the following steps:

[0092] S1: Obtain at least 1000 samples in advance, and mark the actual excrement volume in each sample;

[0093] S2: preprocessing each capacitance data sequence using a preprocessing unit;

[0094] S3: Divide the 1000 capacitance data sequences into training set, validation set and test set;

[0095] S4; Set initial parameters for the prediction model, which will be optimized during the training process.

[0096] S5: Input the capacitance data sequence in the training set into the prediction model, and continuously adjust the model parameters through the optimization algorithm to minimize the loss function.

[0097] In other embodiments, when training the prediction model, it is generally done on a high-performance computer. The training accuracy of the prediction model is affected by the complexity of the model. The more complex the model structure, the more it can discover the intrinsic connection in the data. Similarly, when it is used later, the operation time is longer and the energy consumption is higher. Therefore, large network structures are not conducive to use in miniaturized, portable devices. To this end, this solution provides the following training device for training and provides a new loss function.

[0098] like Figure 4 As shown, a training device is also included, and the prediction model is input into the training device for training;

[0099] The training device includes: an input unit, a first high-order unit, a second high-order unit, and a training unit. The input unit is used to input samples; the first high-order unit has a built-in first high-order network model; the second high-order unit has a built-in second high-order network model; the training unit has a built-in prediction model to be trained; the input unit sends samples to the first high-order unit, the second high-order unit, and the training unit respectively; the first high-order network model and the second high-order network model have the same structure, and the complexity of the network structure of the prediction model is less than that of the first high-order network model; the first high-order network model is used to guide the implicit information of the prediction model reconstruction; the second high-order network model is used to guide the prediction model to capture detailed information.

[0100] In the training device, the first high-order network model, the second high-order network model and the prediction model will all input the same samples, and during the iteration process of the prediction model, the first high-order network model and the second high-order network model will send their learned intermediate state information to the prediction model to guide the training during the iteration process of the prediction model.

[0101] Specifically, the first high-order network model, the second high-order network model, and the hidden layer of the prediction model are all MobileNet structures. Among them, the model structure of the first high-order network model and the second high-order network model is the same, while the model structure of the prediction model is simpler, mainly with the number of channels and the number of convolution kernel filters being different. In this solution, the number of channels of the first high-order network model is 4 times the number of channels of the prediction model, and the number of convolution kernels is also 4 times the number of convolution kernels of the prediction model.

[0102] Taking a sample as an example, after the sample is input into the training device, the training process is as follows:

[0103] S1: The first high-order unit, the second high-order unit and the training unit receive the same sample respectively; the first high-order network model obtains a first feature map, the second high-order network model obtains a second feature map, and the prediction model obtains a third feature map;

[0104] The first high-order network model, the second high-order network model and the prediction model all include an input layer, a hidden layer and an output layer, where the first feature map, the second feature map and the third feature map are all the final output information of the hidden layer.

[0105] S2: Upgrading the dimension of the third feature map so that the dimension of the third feature map is the same as that of the first feature map and the second feature map.

[0106] Because the network structure of the prediction model is simple, the dimension of the third feature map is lower than that of the first and second feature maps. Therefore, dimension upgrading is required.

[0107] S3: Perform average pooling on the first feature map, the second feature map, and the third feature map to obtain a first feature vector, a second feature vector, and a third feature vector.

[0108] S4: Apply one-dimensional convolutional layers to the first eigenvector, the second eigenvector, and the third eigenvector respectively to realize cross-channel information interaction; apply the Sigmoid function to normalize the output of the one-dimensional convolutional layer to obtain the attention weight of each channel; in this scheme, the implicit information of the first eigenvector and the second eigenvector can be added as attention weights.

[0109] S5: multiply the third feature map by the attention weight of each channel to obtain a weighted output feature map;

[0110] S6: Calculate the difference between the predicted model and the attention weight of the first high-order model and the second high-order model on each channel to calculate the loss function L P , by minimizing the loss function L P Adjust model parameters.

[0111] The loss function is;

[0112] ;

[0113] Wherein, λ1, λ2, and λ3 are respectively the first weight coefficient, the second weight coefficient, and the third weight coefficient which are preset in advance;

[0114] ;

[0115] L1 is the reconstruction loss function, which is the mean absolute error between the output feature map and the first feature map, used to describe the detail difference information, I3 is the output feature map, and I1 is the first feature map;

[0116] ;

[0117] L2 is the perceptual loss function, which is the mean absolute error between VGG (I3) and VGG (I2), and is used for the perceptual loss between the output feature map and the second feature map;

[0118] VGG (I3) is the first perceptual information obtained by inputting the output feature map into the output layer of the VGG model, and VGG (I2) is the second perceptual information obtained by inputting the output feature map into the output layer of the VGG model.

[0119] The VGG model is a pre-trained deep convolutional prediction model developed by the Visual Geometry Group at the University of Oxford.

[0120] ;

[0121] L3 is the embedding consistency distance loss function, which is used to describe the consistency difference between the first eigenvector, the second eigenvector, and the third eigenvector in the attention weight;

[0122] in , represents the attention weight of the i-th feature mapping to the j-th channel of the first high-order network model,

[0123] , represents the attention weight of the i-th feature mapping to the j-th channel of the second high-order network model, represents the attention weight of the i-th feature map to the j-th channel of the prediction model, i represents the feature index, j represents the channel index, t represents the number of feature maps, and c represents the number of channels.

[0124] The above is the training process and the corresponding loss function. In this way, the prediction model can be trained when there are enough samples.

[0125] This application selects a portion of the data to train the trained prediction model. Figure 3 The scatter plot of relative error of the prediction model content after training is displayed. Figure 3 The horizontal axis in the figure is the sampling point, and the vertical axis is the relative error. When a large number of points are taken, the relative error is small. It can be seen from the figure that the relative error of the content prediction is small. Figure 5 The frequency distribution diagram of the relative error of the prediction model after training is shown. It can be seen from the figure that the error approximately follows a normal distribution. The 95% confidence interval of the relative error is [0.084, 0.14], indicating that the prediction model has a high prediction accuracy.

[0126] Example 2: The present application provides a method for monitoring excrement from an enterostomy bag, using the aforementioned enterostomy bag excrement monitoring device to monitor the excrement volume in the enterostomy bag.

[0127] The above description is only some preferred embodiments of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present application is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present application to form a technical solution.

Claims

1. A device for monitoring excrement from an intestinal ostomy bag, characterized in that: include: A stoma bag having an access opening for connecting to the patient's intestinal stoma; A monitor, comprising at least two monitoring electrodes, the two monitoring electrodes being located on both sides of the stoma, the two monitoring electrodes forming a capacitive sensor in the stoma bag; The information collection device is connected to the monitor signal, receives the capacitance information of the monitor in real time, and obtains a capacitance data sequence that changes with time; An information processing device, signal-connected to the information collecting device, wherein the information processing device has a built-in prediction model; The information processing device receives the capacitance data sequence, and based on the built-in prediction model, the information processing device selects a true value representing the current actual excrement capacity from the capacitance data sequence to obtain the real-time excrement content.

2. The enterostomy bag excrement monitoring device according to claim 1, characterized in that: The information collection device forms all capacitance data generated within a preset time period into a capacitance data sequence to obtain multiple capacitance data sequences; each capacitance data sequence is used as a sample.

3. The enterostomy bag excrement monitoring device according to claim 1, characterized in that: The last 80% of the capacitance data sequence of the adjacent previous sample overlaps with the first 80% of the capacitance data sequence of the adjacent next sample.

4. The enterostomy bag excrement monitoring device according to claim 1, characterized in that: The information processing device includes: A preprocessing unit, connected to the information collection device signal, preprocesses the received capacitance data sequence; The neural network unit has a built-in prediction model. The preprocessed capacitance data sequence is input into the neural network unit to obtain the true value of the current actual excrement capacity.

5. The enterostomy bag excrement monitoring device according to claim 4, characterized in that: Predictive models include: An input layer, for receiving a capacitance data sequence; Hidden layer: the hidden layer has at least 3 layers, and each hidden layer has 10 neurons; The output layer is connected to the hidden layer signal and uses the output activation function to output the corresponding true value.

6. The enterostomy bag excrement monitoring device according to claim 5, characterized in that: The loss function of the prediction model is M; ； Where n represents the number of samples, represents the predicted value of the i-th sample, y i Represents the actual value of the i-th sample.

7. The enterostomy bag excrement monitoring device according to claim 5, characterized in that: The training method of the prediction model includes the following steps: S1: Obtain at least 1000 samples in advance, and mark the actual excrement volume in each sample; S2: preprocessing each capacitance data sequence using a preprocessing unit; S3: Divide the 1000 capacitance data sequences into training set, validation set and test set; S4; setting initial parameters for the prediction model, inputting the capacitance data sequence in the training set into the prediction model, and continuously adjusting the model parameters through the optimization algorithm to minimize the loss function.

8. The enterostomy bag excrement monitoring device according to claim 5, characterized in that: Also included is a training device, wherein the prediction model is input into the training device for training; The training device comprises: An input unit, used to input samples; A first high-order unit, having a first high-order network model built therein; A second high-order unit, having a second high-order network model built in; The training unit contains the prediction model to be trained; The input unit sends the samples to the first high-order unit, the second high-order unit and the training unit respectively; Among them, the first high-order network model and the second high-order network model have the same structure, and the complexity of the network structure of the prediction model is less than that of the first high-order network model; The first high-order network model is used to guide the implicit information of the reconstruction of the prediction model; The second high-order model is used to guide the prediction model to capture detailed information.

9. A method for monitoring excrement from an intestinal ostomy bag, characterized in that: The enterostomy bag excrement monitoring device according to any one of claims 1 to 8 is used to monitor the excrement volume in the ostomy bag.

Images