Bladder pressure detection method, device and storage medium

Abstract

The present application provides a bladder pressure detection method, device, and storage medium, relating to the field of medical detection technology. The method obtains a bladder image obtained after an endoscope is inserted into the bladder, extracts bladder feature data related to bladder pressure from the bladder image, and the bladder feature data includes at least one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features. Based on the bladder feature data, model input features are determined, and the model input features are input into a pre-trained pressure recognition model to determine the current bladder pressure. The bladder feature data changes with changes in bladder pressure, that is, the bladder feature data is related to bladder pressure. Therefore, bladder pressure can be predicted based on the bladder feature data, which can reduce the number of inserted devices while improving the accuracy of bladder pressure detection.

Description

Technical Field

[0001] The present application relates to the field of medical detection technology, and in particular to a bladder pressure detection method, device, and storage medium. Background Art

[0002] During bladder surgery, medical staff can insert an endoscope and laser fiber into the surgical site. They observe the surgical site through the endoscope and control the angle of the laser fiber to break up the stones. Irrigation fluid flows through the endoscope channel, flushing away stone powder, blood clots, or tissue fragments. Turbid fluid is then aspirated through the endoscope's outlet channel. Because saline needs to be continuously infused and urine accumulates in the patient's body, excessive pressure can cause tissue ischemia or even rupture. Therefore, bladder pressure must be continuously monitored during surgery to stabilize it within an appropriate range through relevant procedures.

[0003] Currently, bladder pressure is typically detected using a pressure sensor, typically located in the middle of the irrigation fluid inlet tube. This sensor measures the pressure of the irrigation fluid and estimates bladder pressure. However, this method of estimating bladder pressure suffers from poor real-time performance and low accuracy, posing a risk of tissue rupture and compromising surgical safety. While inserting additional pressure sensors can accurately measure bladder pressure, it increases the number of instruments inserted, increasing patient trauma and the complexity of the procedure. Summary of the Invention

[0004] The present application provides a bladder pressure detection method, device, and storage medium, which can reduce the number of inserted instruments while improving the accuracy of bladder pressure detection.

[0005] The present application provides a bladder pressure detection method, comprising:

[0006] Acquire a bladder image; the bladder image is a real-time image acquired after the endoscope is inserted into the bladder;

[0007] extracting bladder feature data related to bladder pressure based on the bladder image; the bladder feature data including at least one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features;

[0008] An input feature of a pressure recognition model is determined based on the bladder feature data, and the input feature of the pressure recognition model is input into a pre-trained pressure recognition model to determine a current bladder pressure, where the current bladder pressure represents the pressure on the inner wall of the bladder.

[0009] Optionally, the bladder feature data includes bladder wall texture features, and the step of extracting bladder feature data related to bladder pressure based on the bladder image includes:

[0010] performing a first preprocessing on the bladder image, including image denoising, contrast enhancement, and edge sharpening, to obtain a first preprocessed image;

[0011] obtaining a first bladder wall mask based on the first preprocessed image and a pre-trained first segmentation model; the first bladder wall mask is used to describe the position and shape of the bladder wall region in the first preprocessed image; the first segmentation model is used to output the first bladder wall mask based on the input first preprocessed image;

[0012] Bladder wall texture features in a bladder wall region are extracted based on the first bladder wall mask and the first preprocessed image.

[0013] Optionally, the extracting bladder wall texture features in the bladder wall region based on the first bladder wall mask and the first preprocessed image includes:

[0014] Multiplying the first preprocessed image and the first bladder wall mask pixel by pixel to obtain a first feature map, where the first feature map is an image of the bladder wall region in the first preprocessed image;

[0015] Determine a gray level co-occurrence matrix based on the first feature map, a preset angle, and a preset pixel distance, and determine a contrast, an angular second moment, and an inverse disparity based on the gray level co-occurrence matrix;

[0016] Determining local binary pattern feature values ​​of the first feature map using a local binary pattern based on a preset domain radius and preset sampling points;

[0017] Obtaining extracted features of the bladder wall region based on the first feature map and a pre-trained feature extraction model; the feature extraction model is configured to output extracted features based on the input first feature map;

[0018] The bladder wall texture features were obtained by combining contrast, angular second moment, inverse disparity, local binary pattern eigenvalues ​​and extracted features.

[0019] Optionally, the bladder characteristic data includes bladder wall motion characteristics, and the step of extracting bladder characteristic data related to bladder pressure based on the bladder image includes:

[0020] Performing video preprocessing including image denoising, feature enhancement and video stabilization on continuous frame bladder images to obtain a preprocessed image sequence;

[0021] determining a second bladder wall mask for each frame of the preprocessed image sequence based on the preprocessed image sequence and a pre-trained second segmentation model; the second bladder wall mask is used to describe the position and shape of the bladder wall region in each frame of the preprocessed image sequence; and the second segmentation model is used to output a second bladder wall mask corresponding to each frame of the preprocessed image sequence based on the input preprocessed image sequence;

[0022] Multiplying each frame image in the preprocessed image sequence by the corresponding second bladder wall mask pixel by pixel to obtain a plurality of second feature maps, where the second feature maps are bladder wall area images of each frame image in the preprocessed image sequence;

[0023] Determine the bladder wall displacement field sequence in each second feature map based on the optical flow method; the bladder wall displacement field sequence represents the motion trajectory of each pixel point on the bladder wall;

[0024] The bladder wall motion features are extracted based on the bladder wall displacement field sequence of the bladder wall area in each frame image.

[0025] Optionally, the bladder wall motion feature includes an average bladder wall motion amplitude, and the average bladder wall motion amplitude includes the motion amplitudes of the bladder wall in the horizontal direction and the vertical direction in the bladder image.

[0026] Optionally, the bladder characteristic data includes bladder wall reflection characteristics, and the step of extracting bladder characteristic data related to bladder pressure based on the bladder image includes:

[0027] performing a second preprocessing on the bladder image including image denoising and illumination correction to obtain a second preprocessed image;

[0028] Based on the second preprocessed image and a pre-trained reflection feature extraction model, bladder wall reflection features are obtained, and the bladder wall reflection features include specular reflection features and diffuse reflection features; the reflection feature extraction model is used to output specular reflection features and diffuse reflection features based on the input second preprocessed image.

[0029] Optionally, determining the input features of the pressure recognition model based on the bladder feature data includes:

[0030] determining the bladder feature data as a model input feature; or,

[0031] Patient characteristic information is acquired, and the model input feature is generated based on the bladder characteristic data and the patient characteristic information.

[0032] Optionally, the patient characteristic information includes at least one of gender, height, age and weight.

[0033] To achieve the above objectives and other related objectives, the present application provides a bladder pressure detection device, comprising:

[0034] A data acquisition module is used to acquire a bladder image; the bladder image is a real-time image acquired after the endoscope is inserted into the bladder;

[0035] a feature extraction module, configured to extract bladder feature data related to bladder pressure based on the bladder image; the bladder feature data comprising at least one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features;

[0036] The pressure prediction module is used to determine the input features of the pressure recognition model based on the bladder characteristic data, and input the input features of the pressure recognition model into a pre-trained pressure recognition model to determine the current bladder pressure, wherein the current bladder pressure represents the pressure borne by the inner wall of the bladder.

[0037] To achieve the above objectives and other related objectives, the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a computer processor, the computer executes one or more of the aforementioned bladder pressure detection methods.

[0038] As described above, the bladder pressure detection method, device, and storage medium provided by this application have the following beneficial effects:

[0039] The present application discloses a method for detecting bladder pressure. The method obtains a bladder image obtained after inserting an endoscope into the bladder, extracts bladder feature data related to bladder pressure from the bladder image, and the bladder feature data includes at least one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features. A model input feature is determined based on the bladder feature data, and the model input feature is input into a pre-trained pressure recognition model to determine the current bladder pressure. The bladder feature data changes with changes in bladder pressure, that is, the bladder feature data is related to bladder pressure. Therefore, bladder pressure can be predicted based on the bladder feature data, which can improve the accuracy of bladder pressure detection while reducing the number of inserted devices.

[0040] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are incorporated into and constitute a part of the specification, illustrating embodiments consistent with the present application and, together with the specification, serving to explain the principles of the present application. It is obvious that the drawings described below are merely some embodiments of the present application, and a person of ordinary skill in the art can derive other drawings based on these drawings without inventive effort. In the drawings:

[0042] Figure 1 is a flow chart of a bladder pressure detection method shown in an exemplary embodiment of the present application;

[0043] Figure 2It is a structural block diagram of a bladder pressure detection device shown in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will describe the embodiments of the present application with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand the other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be understood that the preferred embodiments are only for the purpose of illustrating the present application and are not intended to limit the scope of protection of the present application.

[0045] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. Therefore, the illustrations only show components related to the present application and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component can be changed at will, and the component layout type may also be more complicated.

[0046] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present application difficult to understand.

[0047] See also Figure 1 , Figure 1 FIG is a flow chart of a bladder pressure detection method according to an exemplary embodiment of the present application. Figure 1 It can be seen that the bladder pressure detection method may include:

[0048] Step S110: Acquire a bladder image.

[0049] The bladder image is a real-time image obtained after the endoscope is inserted into the bladder.

[0050] In one embodiment of the present application, when medical personnel use an endoscope to perform bladder stone surgery, the endoscope is inserted into the bladder, breaks up the stones using a laser, and continuously instills an infusion solution to lower the temperature within the bladder and remove the broken stone powder. The endoscope host can obtain real-time bladder images through the endoscopic probe. The bladder images include information such as the stones and bladder wall.

[0051] Step S120 : extracting bladder characteristic data related to bladder pressure based on the bladder image.

[0052] The bladder feature data includes at least one of bladder wall texture features, bladder wall movement features, and bladder wall reflection features.

[0053] In one embodiment of the present application, bladder feature data related to bladder pressure can be extracted based on the bladder image. The bladder wall texture feature characterizes the wrinkle characteristics of the bladder inner wall surface. When the internal pressure of the bladder is high, there are fewer wrinkles, and when the internal pressure of the bladder is low, there are more wrinkles. The bladder wall movement feature characterizes the movement amplitude of the bladder over a period of time. When the internal pressure of the bladder is high, the movement amplitude of the bladder wall is small with the movement of breathing and other movements. When the internal pressure of the bladder is low, the movement amplitude of the bladder wall with the movement of breathing and other movements is large. The bladder wall reflection feature characterizes the reflection intensity of the bladder under the illumination of a light source. When the internal pressure of the bladder is high, the smooth muscle of the bladder wall is stretched, the mucosal layer becomes thinner and the surface is smoother. At this time, the mirror reflection is strong, which appears as a strong reflective area of ​​mirror highlight in the bladder image. When the internal pressure of the bladder is low, the bladder inner wall is relatively relaxed, the surface of the bladder inner wall has more wrinkles and high roughness. At this time, the light is scattered in multiple directions, the reflection is mainly diffuse reflection, and the imaging is more uniform.

[0054] It should be noted that, in addition to bladder wall texture features, bladder wall motion features, and bladder wall reflection features, bladder characteristic data related to bladder pressure may also include features such as wall thickness, tension, and curvature. When the endoscope is an ultrasonic endoscope or a three-dimensional endoscope, three-dimensional modeling can be performed based on the acquired bladder image to obtain wall thickness, tension, and curvature.

[0055] Exemplarily, bladder wall texture features can be extracted based on the bladder image; or, bladder wall motion features can be extracted based on the bladder image; or, bladder wall reflection features can be extracted based on the bladder image; or, bladder wall texture features and bladder wall motion features can be extracted based on the bladder image; or, bladder wall texture features and bladder wall reflection features can be extracted based on the bladder image; or, bladder wall texture features, bladder wall motion features, and bladder wall reflection features can be extracted based on the bladder image.

[0056] In one embodiment, when the bladder feature data includes bladder wall texture features, extracting bladder feature data related to bladder pressure based on the bladder image may include: performing a first preprocessing on the bladder image including image denoising, contrast enhancement, and edge sharpening to obtain a first preprocessed image; obtaining a first bladder wall mask based on the first preprocessed image and a pre-trained first segmentation model; the first bladder wall mask is used to describe the position and shape of the bladder wall region in the first preprocessed image; the first segmentation model is used to output the first bladder wall mask based on the input first preprocessed image; and extracting bladder wall texture features in the bladder wall region based on the first bladder wall mask and the first preprocessed image.

[0057] Optionally, when extracting bladder wall texture features, the process of performing a first preprocessing on the bladder image may include using a non-local means denoising (NLM) algorithm to perform image denoising on the bladder image to obtain a denoised grayscale image; then, contrast-limited adaptive histogram equalization (CLAHE) may be used to perform contrast enhancement on the denoised grayscale image to obtain an intermediate image, and contrast is enhanced by local area histogram equalization to solve the problem of uneven illumination and highlight the details of mucosal folds; and edge sharpening is performed on the intermediate image using a Laplacian operator to obtain a first preprocessed image.

[0058] It should be noted that, when extracting bladder wall texture features, the first segmentation model can be a U-Net network, the input of the first segmentation model is the first preprocessed image, and the output is the first bladder wall mask. The process of training the first segmentation model can include: marking the bladder walls of multiple sample images as training samples, using Dice Loss and / or Adam optimizer and a preset overlap (Intersection over Union, IoU) to train the first segmentation model to obtain a trained first segmentation model.

[0059] Optionally, when extracting bladder wall texture features, the process of extracting bladder wall texture features in the bladder wall area based on the first bladder wall mask and the first preprocessed image may include: multiplying the first preprocessed image and the first bladder wall mask pixel by pixel to obtain a first feature map, where the first feature map is an image of the bladder wall area in the first preprocessed image; determining a grayscale co-occurrence matrix based on the first feature map, a preset angle, and a preset pixel distance, and determining contrast, angular second-order moment, and inverse disparity based on the grayscale co-occurrence matrix; determining local binary pattern eigenvalues ​​of the first feature map using a local binary pattern and based on a preset domain radius and preset sampling points; obtaining extracted features of the bladder wall area based on the first feature map and a pre-trained feature extraction model; and splicing contrast, angular second-order moment, inverse disparity, local binary pattern eigenvalues, and extracted features to obtain bladder wall texture features.

[0060] It should be noted that the feature extraction model can be a ResNet-18 model. The ResNet-18 model includes an input layer, a convolutional layer, a maximum pooling layer, four residual blocks, an average pooling layer, and a fully connected layer. The feature extraction model can be trained based on training samples. The training samples may include a sample bladder wall area (that is, a region of interest) extracted from a sample bladder image and the corresponding bladder pressure. The feature extraction model is trained with the goal of minimizing the loss function through a preset training set, a validation set, a test set, and a loss function to obtain a first feature extraction model. The fully connected layer of the first feature extraction model is removed to obtain a trained feature extraction model. The trained feature extraction model can obtain the extracted features of the bladder wall area based on the input bladder wall area.

[0061] For example, contrast, angular second moment and inverse disparity are all one-dimensional data, the local binary pattern eigenvalue is a 59-dimensional vector, and the extracted feature is a 512-dimensional vector, which can be spliced ​​to obtain a 574-dimensional bladder wall texture feature.

[0062] Optionally, the bladder wall texture features can be reduced in dimension based on principal component analysis (PCA) and recursive feature elimination (RFE) and then input into a pre-trained pressure recognition model, or the concatenated 574-dimensional bladder wall texture features can be directly used to input into the pre-trained pressure recognition model to obtain the current bladder pressure.

[0063] In one embodiment, when the bladder feature data includes bladder wall motion features, bladder feature data related to bladder pressure is extracted based on bladder images, including: performing video preprocessing including image denoising, feature enhancement, and video stabilization on consecutive frame bladder images to obtain a preprocessed image sequence; determining a second bladder wall mask for each frame in the preprocessed image sequence based on the preprocessed image sequence and a pre-trained second segmentation model; the second bladder wall mask is used to describe the position and shape of the bladder wall area in each frame of the preprocessed image sequence; the second segmentation model is used to output a second bladder wall mask corresponding to each frame based on the input preprocessed image sequence; pixel-by-pixel multiplication of each frame in the preprocessed image sequence and the corresponding second bladder wall mask to obtain multiple second feature maps, the second feature maps being bladder wall area images of each frame in the preprocessed image sequence; determining a bladder wall displacement field sequence in each second feature map based on an optical flow method; the bladder wall displacement field sequence characterizes the motion trajectory of each pixel point of the bladder wall; and extracting bladder wall motion features based on the bladder wall area and displacement vector of each frame.

[0064] It should be noted that the number of consecutive frames or the processing cycle can be preset. After the processing cycle is preset, the bladder wall motion features can be extracted from the consecutive frame bladder images in each cycle. For example, the processing cycle can be 1 second.

[0065] Optionally, performing video preprocessing including image denoising, feature enhancement and video stabilization on the continuous frame bladder images to obtain the preprocessed image sequence may include: applying non-local mean denoising to the continuous frame bladder images frame by frame to perform image denoising to obtain a denoised frame sequence; applying adaptive histogram equalization to feature enhancement on the denoised frame sequence frame by frame to obtain an enhanced frame sequence; and performing video stabilization on the enhanced frame sequence to obtain the preprocessed image sequence.

[0066] Image denoising of continuous frame bladder images can be performed by applying non-mean denoising to each frame of the bladder image to obtain denoised continuous frame images, which can enhance the signal-to-noise ratio of the motion signal. Adaptive histogram equalization can then be applied to each frame in the denoised continuous frame images to achieve the goal of feature enhancement and obtain enhanced continuous frame images. ORB feature point detection is performed on each frame in the enhanced continuous frame images to determine multiple first feature points in each frame. RANSAC affine estimation is used to calculate the optimal transformation matrix of the first feature points of adjacent frames. Based on the optimal transformation matrix, the current frame is mapped to the reference coordinate system to obtain a preprocessed image sequence. The coordinates of each frame image are unified for video stabilization, facilitating subsequent texture feature extraction.

[0067] Optionally, the preprocessed image sequence can be input into a pretrained second segmentation model to obtain a second bladder wall mask for each frame. The second segmentation model can be a U-Net network. The input of the second segmentation model is the preprocessed image sequence, and the output is a second bladder wall mask sequence. The second bladder wall mask sequence includes a second bladder wall mask corresponding to each frame. Each frame in the preprocessed image sequence corresponds to a second bladder wall mask. For each frame in the preprocessed image sequence, each frame is multiplied pixel by pixel with the corresponding second bladder wall mask to obtain a second feature map. The second feature map is an image of the bladder wall region for each frame in the preprocessed image sequence.

[0068] Optionally, the process of determining the bladder wall displacement field sequence in each second feature map based on the optical flow method may include:

[0069] For each pixel point in the bladder wall area in the second feature map, a gradient matrix is ​​determined. The gradient matrix includes spatial gradient and temporal gradient. The spatial gradient determination formula can be expressed as:

[0070] ;

[0071] in, Characterize any second feature map, and Respectively represent the two-dimensional spatial coordinates of the current pixel in the second feature map, Corresponding to the horizontal direction, Corresponding to the vertical direction, t represents the time dimension, which is the frame number in the preprocessed image sequence. Represents the pixel to the right of the current pixel in the same frame image, Represents the pixel below the current pixel in the same frame image, The coordinates of the t-th frame image are represented as The pixels in The spatial gradient in the axial direction, The coordinates of the t-th frame image are represented as The pixels in The spatial gradient along the axis.

[0072] The time gradient determination formula may include:

[0073] ;

[0074] Centered on each pixel, a local window is constructed (for example, the local window can be ), for each pixel in the window, establish the equation , for the local window Pixels, construct The raw displacement field of the bladder wall region of each second feature map is determined using the least squares method using equations including:

[0075] ;

[0076]

[0077] ;

[0078] in, Characterize the pixel point The instantaneous displacement in the direction, Characterize the pixel point The instantaneous displacement in the direction, is the displacement matrix, is the time matrix, For the Pixels in The spatial gradient in the axial direction, For the Pixels in The spatial gradient in the axial direction, For the The temporal gradient of pixels, is the transposed matrix of the displacement matrix;

[0079] The bladder wall displacement field sequence in each second characteristic graph can be represented as , including the original displacement field of each pixel in the second feature map, the original displacement field is represented as .

[0080] Optionally, the bladder wall motion feature may include an average bladder wall motion amplitude. Based on the bladder wall displacement field sequence of the bladder wall region in each second feature map, the process of extracting the bladder wall motion feature may include:

[0081] For each bladder wall region of the second feature map, determine the instantaneous displacement amplitude of each pixel:

[0082] ;

[0083] in, Characterizes the instantaneous displacement amplitude;

[0084] Determine the average bladder wall motion amplitude of the bladder wall region of each second feature map:

[0085] ;

[0086] in, Characterization The average motion amplitude of the bladder wall in the second feature map of the frame, Represents the total number of pixels in the second feature map.

[0087] Then, the set of the average bladder wall motion amplitudes of all second feature maps may be determined as the bladder wall motion feature.

[0088] In one embodiment, when the bladder feature data includes bladder wall reflection features, extracting bladder feature data related to bladder pressure based on the bladder image includes: performing a second preprocessing on the bladder image including image denoising and illumination correction to obtain a second preprocessed image; determining specular reflection features and diffuse reflection features based on the second preprocessed image and a pre-trained reflection feature extraction model; the bladder wall reflection features include specular reflection features and diffuse reflection features; and the reflection feature extraction model is used to output specular reflection features and diffuse reflection features based on the input second preprocessed image.

[0089] Optionally, Gaussian filtering or median filtering may be used to perform image denoising on the bladder image, and histogram equalization may be used to perform illumination correction on the bladder image to obtain a second preprocessed image.

[0090] Exemplary training of a reflectance feature extraction model may include obtaining a dataset of annotated endoscopic images. First, a series of internal bladder images acquired through a cystoscopic endoscope are obtained. For each original image, annotated images are annotated with specular and diffuse reflectance components. These annotations can be obtained synchronously with the endoscope using specialized optical measurement equipment, or by medical experts manually annotating some data based on optical knowledge and image features. Data augmentation techniques can then be used to obtain more annotated samples. Retinex or other image segmentation algorithms can also be applied to separate specular and diffuse reflectance components from the endoscopic images. A convolutional neural network can be selected as the initial reflectance feature extraction model, with parameters such as model weights and biases randomly initialized. The original images from the training dataset are input into the model, which performs calculations based on the current parameters to obtain predicted specular and diffuse reflectance components. The mean squared error (MSE) is used to calculate the error between the predicted values ​​and the true annotated values, including the error for both the specular and diffuse reflectance components. Based on this error, the model parameters are adjusted using an optimization algorithm (such as stochastic gradient descent).

[0091] Step S130 : determining input features of a pressure recognition model based on the bladder feature data, and inputting the input features of the pressure recognition model into a pre-trained pressure recognition model to determine the current bladder pressure.

[0092] The current bladder pressure represents the pressure on the inner wall of the bladder.

[0093] In one embodiment of the present application, the input features of the pressure recognition model can be determined based on the bladder feature data, and the input features of the pressure recognition model can be input into a pre-trained pressure recognition model to determine the current bladder pressure.

[0094] In one possible implementation, bladder feature data can be determined as input features for a pressure recognition model. When the bladder feature data includes one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features, the bladder feature can be determined as the input feature for the pressure recognition model. When the bladder feature includes at least two of the bladder wall texture features, bladder wall motion features, and bladder wall reflection features, the bladder feature data can be concatenated to obtain the input features for the pressure recognition model.

[0095] In another possible implementation, patient characteristic information can be obtained and model input features can be generated based on bladder characteristic data and the patient characteristic information. The patient characteristic information includes at least one of gender, height, age, and weight. The patient characteristic information and bladder characteristic information can be concatenated to generate input features for the pressure recognition model.

[0096] It should be noted that gender and height directly affect the anatomical characteristics of the bladder. For example, male urethra length is significantly longer than female urethra length, and height may be correlated with bladder capacity. Incorporating these parameters into the model can more accurately reflect the patient's actual anatomy and correct for pressure prediction errors caused by individual differences. Furthermore, aging leads to decreased bladder muscle contractility and mucosal elasticity, while weight may reflect the impact of fat distribution on bladder space. Using patient characteristic information, the model can dynamically adjust the weighting of bladder wall motion features to avoid misjudgments due to physiological degeneration or body size differences. Weight and age may be associated with a patient's metabolic syndrome or chronic inflammatory state, factors that may indirectly affect pressure distribution through bladder wall vascular texture (such as congestion and follicle formation). The model can capture such implicit associations through feature fusion.

[0097] The patient characteristic information and the bladder characteristic data may be spliced ​​together. After the patient characteristic information and the bladder characteristic data are spliced ​​together, the spliced ​​data may be normalized to obtain input features of the pressure recognition model.

[0098] In one embodiment of the present application, a pressure recognition model can be trained using training samples. The training samples may include model input features and bladder pressure labels. A mean square error loss function is selected, and the pressure recognition model is trained with the goal of minimizing the loss function. This results in a trained pressure recognition model, also known as a pre-trained pressure recognition model. The model input features during the model application phase and the model input features during the training phase have the same feature data type. For example, if the model input features during the training phase are bladder feature data, then the model input features during the model application phase will also be bladder feature data.

[0099] It should be noted that the pressure recognition model can be a DNN (Deep-Learning Neural Network).

[0100] During the model application phase, the model input features are fed into a pre-trained pressure recognition model to generate a pressure prediction value, which is then used as the current bladder pressure. During surgery, the current bladder pressure can be used to adjust the infusion rate, the opening and closing of the fluid injection, and the rate of fluid removal.

[0101] Figure 2 FIG is a block diagram of a bladder pressure detection device shown in an exemplary embodiment of the present application. Figure 2 As shown, the exemplary bladder pressure detection device 200 includes:

[0102] The data acquisition module 210 is used to acquire a bladder image; the bladder image is a real-time image acquired after the endoscope is inserted into the bladder.

[0103] The feature extraction module 220 is configured to extract bladder feature data related to bladder pressure based on the bladder image; the bladder feature data includes at least one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features.

[0104] The pressure prediction module 230 is used to determine the input features of the pressure recognition model based on the bladder characteristic data, and input the input features of the pressure recognition model into a pre-trained pressure recognition model to determine the current bladder pressure, where the current bladder pressure represents the pressure on the inner wall of the bladder.

[0105] It should be noted that the bladder pressure detection device provided in the above-described embodiment and the bladder pressure detection method provided in the above-described embodiment share the same concept. The specific manner in which the various modules and units perform their operations has been described in detail in the method embodiments and will not be repeated here. In actual applications, the bladder pressure detection device provided in the above-described embodiment can, as needed, allocate the aforementioned functions to different functional modules, i.e., divide the system's internal structure into different functional modules to perform all or part of the aforementioned functions, and this is not intended to be limiting herein.

[0106] An embodiment of the present application further provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the electronic device implements the bladder pressure detection method provided in the above-mentioned embodiments.

[0107] Another aspect of the present application provides a computer-readable storage medium storing a computer program. When executed by a computer processor, the computer program causes the computer to perform the bladder pressure detection methods provided in the aforementioned embodiments. The computer-readable storage medium may be included in the electronic device described in the aforementioned embodiments, or may exist independently and not be incorporated into the electronic device.

[0108] Another aspect of the present application provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the bladder pressure detection method provided in each of the above embodiments.

[0109] In the embodiments of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance. Throughout the specification and claims, the terms "including" and "comprising" are open-ended terms and should be interpreted as "including but not limited to."

[0110] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, any equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concepts disclosed in this application shall be covered by the claims of this application.

Claims

1. A bladder pressure detection device, characterized in that: include: A data acquisition module, used for acquiring bladder images; The bladder image is a real-time image obtained after the endoscope is inserted into the bladder; a feature extraction module, configured to extract bladder feature data related to bladder pressure based on the bladder image; the bladder feature data comprising at least one of bladder wall texture features, bladder wall motion features, and bladder wall reflection features; a pressure prediction module, configured to determine input features of a pressure recognition model based on the bladder characteristic data, and input the input features of the pressure recognition model into a pre-trained pressure recognition model to determine a current bladder pressure, wherein the current bladder pressure represents the pressure borne by the inner wall of the bladder; Among them, the bladder wall motion characteristics represent the movement amplitude of the bladder over a period of time.

2. The bladder pressure detection device according to claim 1, characterized in that: The bladder feature data includes bladder wall texture features, and the bladder feature data related to bladder pressure extracted based on the bladder image includes: performing a first preprocessing on the bladder image, including image denoising, contrast enhancement, and edge sharpening, to obtain a first preprocessed image; obtaining a first bladder wall mask based on the first preprocessed image and a pre-trained first segmentation model; the first bladder wall mask is used to describe the position and shape of the bladder wall region in the first preprocessed image; the first segmentation model is used to output the first bladder wall mask based on the input first preprocessed image; Bladder wall texture features in a bladder wall region are extracted based on the first bladder wall mask and the first preprocessed image.

3. The bladder pressure detection device according to claim 2, characterized in that: The step of extracting bladder wall texture features in the bladder wall region based on the first bladder wall mask and the first preprocessed image includes: Multiplying the first preprocessed image and the first bladder wall mask pixel by pixel to obtain a first feature map, where the first feature map is an image of the bladder wall region in the first preprocessed image; Determine a gray level co-occurrence matrix based on the first feature map, a preset angle, and a preset pixel distance, and determine a contrast, an angular second moment, and an inverse disparity based on the gray level co-occurrence matrix; Determining local binary pattern feature values ​​of the first feature map using a local binary pattern based on a preset domain radius and preset sampling points; Obtaining extracted features of the bladder wall region based on the first feature map and a pre-trained feature extraction model; the feature extraction model is configured to output extracted features based on the input first feature map; The bladder wall texture features were obtained by combining contrast, angular second moment, inverse disparity, local binary pattern eigenvalues ​​and extracted features.

4. The bladder pressure detection device according to claim 1, characterized in that: The bladder characteristic data includes bladder wall motion characteristics, and the bladder characteristic data related to bladder pressure extracted based on the bladder image includes: Performing video preprocessing including image denoising, feature enhancement and video stabilization on continuous frame bladder images to obtain a preprocessed image sequence; determining a second bladder wall mask for each frame of the preprocessed image sequence based on the preprocessed image sequence and a pre-trained second segmentation model; the second bladder wall mask is used to describe the position and shape of the bladder wall region in each frame of the preprocessed image sequence; and the second segmentation model is used to output a second bladder wall mask corresponding to each frame of the preprocessed image sequence based on the input preprocessed image sequence; Multiplying each frame image in the preprocessed image sequence by the corresponding second bladder wall mask pixel by pixel to obtain a plurality of second feature maps, where the second feature maps are bladder wall area images of each frame image in the preprocessed image sequence; Determine the bladder wall displacement field sequence in each second feature map based on the optical flow method; the bladder wall displacement field sequence represents the motion trajectory of each pixel point on the bladder wall; The bladder wall motion features are extracted based on the bladder wall displacement field sequence of the bladder wall area in each frame image.

5. The bladder pressure detection device according to claim 4, characterized in that: The bladder wall motion feature includes an average motion amplitude of the bladder wall. The average motion amplitude of the bladder wall includes the motion amplitudes of the bladder wall in the horizontal direction and the vertical direction in the bladder image.

6. The bladder pressure detection device according to claim 1, characterized in that: The bladder characteristic data includes bladder wall reflection characteristics, and the step of extracting bladder characteristic data related to bladder pressure based on the bladder image includes: performing a second preprocessing on the bladder image including image denoising and illumination correction to obtain a second preprocessed image; Based on the second preprocessed image and a pre-trained reflection feature extraction model, bladder wall reflection features are obtained, and the bladder wall reflection features include specular reflection features and diffuse reflection features; the reflection feature extraction model is used to output specular reflection features and diffuse reflection features based on the input second preprocessed image.

7. The bladder pressure detection device according to claim 1, characterized in that: The step of determining the input features of the pressure recognition model based on the bladder feature data includes: determining the bladder feature data as a model input feature; or, Patient characteristic information is acquired, and the model input feature is generated based on the bladder characteristic data and the patient characteristic information.

8. The bladder pressure detection device according to claim 7, characterized in that: The patient characteristic information includes at least one of gender, height, age and weight.

9. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the steps of the bladder pressure detection device according to any one of claims 1 to 8.

Images