Transmembrane state evaluation system and method for bladder three-dimensional form

Abstract

The invention provides a transmembrane state evaluation system and method for a bladder three-dimensional form, and is applied to the technical field of data processing. The method comprises the following steps: preprocessing bladder ultrasonic image information of a target patient to generate target ultrasonic image sequence information; processing the target ultrasonic image sequence information to generate a target enhanced ultrasonic image; processing the physiological parameter information of the urinary system of the patient based on the target bladder morphology evaluation model to generate a target physiological status feature; processing the target physiological status features, the target enhanced ultrasound image and the functional image data based on the target bladder morphology evaluation model to generate a predicted CT image of the target patient; and processing the predicted CT image of the target patient and the simulated positioning CT image information of the target patient based on the target bladder form evaluation model in combination with the bladder dynamic image information to generate a bladder three-dimensional form evaluation result.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a system and method for evaluating the transmembrane state of a bladder's three-dimensional morphology. Background Art

[0002] During external beam radiotherapy for pelvic tumors, variations in bladder filling can significantly impact the effectiveness of radiotherapy. Inconsistent bladder filling can lead to inadequate coverage of the target volume, ineffective tumor cell destruction, and poor treatment efficacy. Alternatively, the irradiation range can be excessive, causing unnecessary damage to surrounding normal tissues and increasing radiotherapy side effects.

[0003] In the prior art, bladder capacity measurement methods mainly include two-dimensional and three-dimensional ultrasound imaging measurements, CT image measurements, and measurement methods based on bladder bioelectrical impedance. However, these methods are limited to the measurement of bladder capacity, lack in-depth application in the evaluation of bladder three-dimensional morphology, and cannot meet the clinical needs for accurate evaluation of bladder morphology. In addition, although there are related patents focusing on the anterior-posterior registration of ultrasound bladders, they only stay at the registration between ultrasound images, and lack cross-modal evaluation between different modal images such as ultrasound and CT. In actual radiotherapy, positioning CT is the gold standard for plan design, and ultrasound images need to be subsequently registered with it to detect the three-dimensional morphology of the bladder in real time to determine whether treatment has started. However, existing technical means are difficult to achieve accurate cross-modal evaluation, which affects the accuracy and effectiveness of radiotherapy plans.

[0004] Finally, traditional cone-beam CT (CBCT)-based assessment methods, due to their radiation characteristics and imaging modality limitations, cannot provide real-time assessment of the 3D bladder topography and volume. This is extremely inconvenient in scenarios where real-time monitoring of bladder status is required to adjust radiotherapy plans, and it cannot provide timely, accurate, and dynamic bladder information for clinical treatment.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore includes information that does not constitute prior art known to ordinary technicians in this field. Summary of the Invention

[0006] The purpose of this application is to provide a system and method for assessing the transmembrane state of the three-dimensional bladder morphology, which, at least to a certain extent, overcomes the problems existing in the prior art. By sequentially pre-processing and enhancing the bladder ultrasound image, the image quality is improved and the bladder structure information is highlighted. The model is used to process multi-source data. For example, the physiological parameters of the urinary system are classified, feature extracted, normalized, correlated and analyzed, and fused to generate target physiological state characteristics; the ultrasound image and functional imaging data are feature mined, fused, and multi-step processed to generate predicted CT images. Finally, combined with the dynamic image of the bladder, the predicted CT image and the simulated positioning CT image are processed in multiple steps to complete the three-dimensional bladder morphology assessment. The entire process uses deep learning and multimodal data fusion technology to convert cross-modal evaluation into homomodal evaluation, significantly improving the intuitiveness and accuracy of the evaluation. It can effectively assist doctors in formulating precise radiotherapy plans, reduce the side effects of radiotherapy caused by changes in bladder filling, and provide strong support for pelvic tumor radiotherapy.

[0007] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0008] According to one aspect of the present application, a method for assessing the transmembrane state of the three-dimensional morphology of the bladder is provided, comprising: acquiring bladder ultrasound image information of a target patient, simulated positioning CT image information of a target patient, urinary system physiological parameter information of the target patient, functional imaging data, and bladder dynamic image information, wherein the bladder dynamic image information is used to characterize the morphological change pattern of the bladder under different physiological states; preprocessing the bladder ultrasound image information of the target patient to generate target ultrasound image sequence information; processing the target ultrasound image sequence information to generate a target enhanced ultrasound image; processing the patient's urinary system physiological parameter information based on a target bladder morphology assessment model to generate target physiological state characteristics; processing the target physiological state characteristics, target enhanced ultrasound image, and functional imaging data based on the target bladder morphology assessment model to generate a predicted CT image of the target patient; processing the predicted CT image of the target patient and the simulated positioning CT image information of the target patient based on the target bladder morphology assessment model in combination with the bladder dynamic image information to generate a three-dimensional bladder morphology assessment result.

[0009] Another aspect of the present application is a transmembrane state assessment device for a three-dimensional bladder morphology, characterized in that it includes: an acquisition module for acquiring bladder ultrasound image information of a target patient, simulated positioning CT image information of a target patient, urinary system physiological parameter information of a target patient, functional imaging data, and bladder dynamic image information, wherein the bladder dynamic image information is used to characterize the morphological change pattern of the bladder under different physiological states; a processing module for preprocessing the bladder ultrasound image information of the target patient to generate target ultrasound image sequence information; processing the target ultrasound image sequence information to generate a target enhanced ultrasound image; processing the patient's urinary system physiological parameter information based on a target bladder morphology assessment model to generate target physiological state characteristics; processing the target physiological state characteristics, target enhanced ultrasound image, and functional imaging data based on the target bladder morphology assessment model to generate a predicted CT image of the target patient; processing the predicted CT image of the target patient and the simulated positioning CT image information of the target patient based on the target bladder morphology assessment model in combination with the bladder dynamic image information to generate a three-dimensional bladder morphology assessment result.

[0010] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the method for evaluating the transmembrane state of the three-dimensional bladder morphology is implemented.

[0011] This application provides a system and method for assessing the transmembrane state of the bladder's three-dimensional morphology. The server acquires multi-source data, including bladder ultrasound images, simulated positioning CT images, urinary system physiological parameters, functional imaging data, and dynamic bladder images, providing a comprehensive basis for assessment. Bladder ultrasound images are sequentially preprocessed and enhanced to improve image quality and highlight bladder structural information. A model is used to process the multi-source data. For example, urinary system physiological parameters are classified, feature extracted, normalized, and correlated with each other, resulting in target physiological state features. Ultrasound images and functional imaging data are then feature mined, fused, and processed in multiple steps to generate predicted CT images. Finally, the predicted CT images and simulated positioning CT images are processed in multiple steps, combined with dynamic bladder images, to complete the three-dimensional bladder morphology assessment. The entire process utilizes deep learning and multimodal data fusion technologies to transform cross-modal assessment into homomodal assessment, significantly improving the intuitiveness and accuracy of the assessment. This can effectively assist physicians in developing precise radiotherapy plans, reduce radiotherapy side effects caused by bladder filling changes, and provide strong support for pelvic tumor radiotherapy.

[0012] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1A flow chart showing a method for evaluating the transmembrane state of a bladder three-dimensional morphology provided by one embodiment of the present application;

[0014] Figure 2 A schematic structural diagram of a transmembrane state assessment device for three-dimensional bladder morphology provided in one embodiment of the present application is shown. DETAILED DESCRIPTION

[0015] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0016] The following combination Figure 1 The following describes a method for evaluating the transmembrane state of the three-dimensional bladder morphology according to an exemplary embodiment of the present application. In one embodiment, the present application also provides a system and method for evaluating the transmembrane state of the three-dimensional bladder morphology. Figure 1 The following schematically shows a flow chart of a method for evaluating the transmembrane state of a bladder three-dimensional morphology according to an embodiment of the present application. Figure 1 As shown, the method is applied to the server and includes:

[0017] S101, obtaining bladder ultrasound image information of a target patient, simulated positioning CT image information of a target patient, urinary system physiological parameter information, functional imaging data, and bladder dynamic image information of a target patient.

[0018] In one embodiment, ultrasound diagnostic equipment is used to generate images by emitting ultrasound waves toward the patient's bladder and receiving reflected echoes. For example, the patient lies on an examination bed, and the doctor applies coupling gel to the ultrasound probe, places it on the patient's lower abdomen, and adjusts the probe's angle and position to obtain ultrasound images of different sections of the bladder. These images present the general shape of the bladder in real time, but may suffer from issues such as poor clarity and noise interference. For example, some tiny lesions may be difficult to clearly display. During the radiotherapy simulation positioning phase, the patient lies on a CT scanning bed, positions themselves according to the radiotherapy requirements, and then undergoes a CT scan. During the scan, X-rays penetrate the patient's body from different angles. The detector receives the attenuated radiation signals after passing through the human body and, through computer processing, reconstructs a cross-sectional image of the bladder. These CT images can accurately show the bladder's location, shape, size, and relationship to surrounding tissues, and are an important basis for radiotherapy planning. For example, doctors can use these images to determine the target area for radiotherapy.

[0019] Information on urinary system physiological parameters can be obtained through a variety of medical testing methods, including but not limited to laboratory tests and physiological function tests. For example, blood samples can be collected from patients to test for renal function indicators such as creatinine and urea nitrogen to assess whether renal excretion is normal. Urine samples can be collected for routine urinalysis to check for abnormalities such as red blood cells, white blood cells, and protein, which can help determine whether there is inflammation or infection in the urinary system. Cystometry and other methods can also be used to measure changes in bladder pressure during filling and urination to understand bladder function. Functional imaging data can be obtained using functional imaging techniques such as positron emission tomography (PET)-CT or magnetic resonance spectroscopy (MRS). For example, in PET-CT, the patient is first injected with a drug labeled with a radionuclide. These drugs are metabolized in the body and accumulate in metabolically active areas. A PET-CT scan is then performed. By detecting the gamma rays emitted by the radionuclide, metabolic information about bladder tissue is obtained, which can be used to assess bladder function. If a bladder tumor is present, the tumor site is typically more metabolically active than normal tissue, resulting in a different signal on the image.

[0020] Dynamic bladder image information can be obtained using real-time ultrasound dynamic monitoring or cystography dynamic imaging technology. During real-time ultrasound dynamic monitoring, the morphological changes of the bladder during different physiological processes such as filling and urination are continuously observed. The ultrasound equipment can continuously capture images to form a dynamic image sequence. Cystography dynamic imaging is to inject a contrast agent into the bladder and use X-ray fluoroscopy equipment to dynamically observe the changes in bladder morphology over time, such as the peristalsis of the bladder wall and whether the bladder empties smoothly, so as to fully understand the morphological changes of the bladder under different physiological states.

[0021] In another embodiment, obtaining a target bladder morphology assessment model includes obtaining a training sample set and a pre-set bladder morphology assessment model, wherein the training sample set includes multimodal data and corresponding 3D bladder morphology assessment results. The hospital collected data from 500 pelvic tumor patients over the past three years from its case database as the training sample set. This data includes multimodal information, specifically: bladder ultrasound images of each patient during radiotherapy simulation positioning, which record the real-time bladder morphology; high-resolution simulation positioning CT images, which accurately demonstrate the bladder's position, shape, size, and relationship with surrounding tissues; urinary system physiological parameters obtained through laboratory tests and physiological function tests, such as renal function indicators (creatinine, urea nitrogen, etc.), routine urine test results (red blood cell, white blood cell, protein, etc.), and cystometry data; functional imaging data obtained using PET-CT, which reflects the metabolic status of bladder tissue; and dynamic bladder images collected through real-time ultrasound dynamic monitoring or dynamic cystography, which demonstrate the morphological changes of the bladder under different physiological states. In addition, the hospital's medical imaging team, based on clinical experience and expertise, evaluated the 3D bladder morphology of each patient and incorporated these assessment results into the training sample set. In addition, a neural network based on the CycleGAN architecture is used as the preset bladder morphology assessment model. This model has powerful feature extraction and processing capabilities and is suitable for processing multimodal data.

[0022] The number of features in each modality of data in the training set is counted, and a sampling ratio is generated based on the balance of feature distribution. Data analysis tools are used to extract and count features from the multimodal data in the training set. For example, ultrasound image features include bladder wall thickness and internal bladder echogenicity; CT image features include bladder volume, shape index, and distance from surrounding organs; urinary system physiological parameter features cover the numerical range and trend of various indicators; functional imaging data features involve the distribution and intensity of metabolically active areas; and bladder dynamic image features include bladder wall peristalsis frequency and emptying rate. Statistics show that CT image data has a higher number of features, while functional imaging data has a relatively lower number of features. To ensure that data from each modality is fully utilized during training and to prevent data from certain modalities from dominating the training results, a sampling ratio is calculated based on the balance of feature distribution. Assuming that the number of CT image features is three times that of functional imaging data, to balance the impact of the two, the sampling ratio of the functional imaging data is set to three times that of the CT image data. The sampling ratios of the other modalities are determined similarly.

[0023] Based on the sampling ratio, the training sample set is subjected to multimodal stratified sampling processing to generate a preset number of sampling feature combinations. The training sample set is subjected to multimodal stratified sampling. Taking ultrasound images and CT images as an example, a certain number of samples are extracted from the ultrasound image data layer according to their sampling ratios, and samples are extracted from the CT image data layer according to the corresponding sampling ratios, and the extracted ultrasound image samples and CT image samples are combined. Similarly, similar operations are performed on the urinary system physiological parameters, functional imaging data, and bladder dynamic image information. Finally, the samples extracted from different modalities are combined together to form sampling feature combinations. Assuming that 1,000 sampling feature combinations are generated by default, through this stratified sampling method, each combination contains multimodal data, and the proportion of each modality data is relatively balanced, which can more comprehensively reflect the characteristics of the bladder.

[0024] Based on statistical tests performed on any key feature combined with each sampling feature, the data is divided into a response group and a poor response group. Each group contains a preset number of data samples, and at least one sample has lesion identification information. "Whether there is a tumor lesion in the bladder wall" is selected as the key feature for analysis. For each sampling feature combination, a statistical test is performed to determine the relationship between this combination and the bladder wall tumor lesion. For example, a chi-square test is used to analyze the correlation between features such as abnormal bladder wall echo in ultrasound images, bladder wall morphological changes in CT images, and metabolic abnormalities in functional imaging data and tumor lesions. Based on the test results, the data is divided into a response group and a poor response group. Assuming that each group contains 200 data samples, during the division process, it is ensured that each group has at least one sample with clear lesion identification information, such as a sample confirmed by pathological biopsy as bladder cancer. The purpose of this division is to enable the model to learn the different characteristics of normal and diseased bladder morphology during training.

[0025] The pre-set bladder morphology assessment model was iteratively trained based on the response group and the poor response group. The model parameters were optimized through cross-validation to generate the trained model and performance indicators. The data from the response group and the poor response group were input into the pre-set Cyclegan-based bladder morphology assessment model for training. During the training process, the model continuously adjusted its parameters to learn the features and patterns in the data. For example, the model learned the relationship between different echo features in ultrasound images and bladder lesions, as well as the relationship between bladder morphology parameters in CT images and bladder function. To avoid model overfitting, a cross-validation method was used to divide the training data into five parts. Four parts of the data were selected for training each time, and one part of the data was used for validation. This was repeated five times. The average of the five validation results was finally taken as the model's performance indicators, such as accuracy, recall rate, and F1 value. After multiple rounds of iterative training, the model parameters were continuously optimized and the performance gradually improved.

[0026] If the lesion identification information in the training results is recognized by the model as a key feature for morphological assessment, the trained model will be used as the target bladder morphological assessment model. After multiple rounds of training and verification, the model's learning results are analyzed. If the model can accurately identify the importance of lesion identification information to bladder morphological assessment when predicting the three-dimensional bladder morphology, that is, the lesion identification information plays a key role in the model's decision-making process, for example, the model can accurately judge the changing trend of bladder morphology based on the characteristics of bladder wall tumor lesions, then the model is considered to have learned effective features, and the trained model will be determined as the target bladder morphological assessment model. This model will be used for subsequent assessments of the three-dimensional bladder morphology of new patients, helping doctors to more accurately judge the patient's bladder status and formulate more precise radiotherapy plans.

[0027] S102 , preprocessing the bladder ultrasound image information of the target patient to generate target ultrasound image sequence information.

[0028] In one embodiment, bladder ultrasound images of a target patient are processed for denoising and contrast enhancement to generate an initial bladder image. The patient lies on an examination bed, and a physician uses ultrasound diagnostic equipment to obtain an ultrasound image of the bladder. Due to the principles of ultrasound imaging and factors such as ultrasound scattering and reflection by human tissue, the acquired original ultrasound images contain noise, and the contrast between the bladder and surrounding tissue is low, impairing the observation of bladder morphology. Technicians import these original ultrasound images into image processing software and perform denoising using a Gaussian filter algorithm. Gaussian filtering effectively removes random noise from the image by performing a weighted average of each pixel and its neighboring pixels, resulting in a smoother image. Next, a histogram equalization algorithm is applied to enhance image contrast. This algorithm redistributes the grayscale values of image pixels, expanding the dynamic range of the image's grayscale and enhancing the definition of the boundary between the bladder and surrounding tissue. After these two processing steps, an initial bladder image is generated. Compared to the original image, the noise level is significantly reduced, and the bladder's outline and internal structure are easier to observe.

[0029] The initial bladder image is subjected to bladder region segmentation processing to generate a target bladder image, wherein the target bladder image is an image containing only the bladder region. In order to more accurately analyze the bladder morphology, the bladder region needs to be separately segmented from the initial bladder image. The U-Net network model based on deep learning is used for bladder region segmentation. The U-Net network has an encoder and decoder structure. The encoder is responsible for extracting the features of the image, and the decoder uses these features to reconstruct and segment the image. The initial bladder image is input into the U-Net network. After training, the network learns the morphological features of the bladder, can accurately identify the bladder region, and segment it from the entire image to generate a target bladder image containing only the bladder region. This can eliminate the interference of surrounding tissues and provide a basis for subsequent more accurate evaluation of bladder morphology.

[0030] The target bladder image undergoes image quality assessment to generate quality defect information. This quality defect information characterizes quality issues such as image blur and artifacts, as well as their location and type. Although image quality has improved after the previous processing, some quality issues may still exist. The target bladder image is evaluated using a specialized image quality assessment algorithm. This algorithm analyzes each pixel in the image and its surrounding area to detect quality issues such as blur and artifacts. For example, blur is determined by calculating the image's gradient information. A low gradient value in a certain area indicates unclear edges, suggesting blur. Artifacts are also detected by analyzing abnormal changes in pixel values within the image. The algorithm records the location and type of these quality issues and generates quality defect information. For example, if blur is detected in the upper left corner of the image, this could be caused by uneven scanning angle or pressure of the ultrasound probe in that area. If there are also artifacts caused by ultrasound reflections in the image, the algorithm will accurately record detailed information such as the location, degree of blur, and type of these blurred areas and artifacts.

[0031] The quality defect information is processed to generate repair strategy parameters, which include blur processing parameters, denoising parameters, artifact repair parameters, and image enhancement parameters. Based on the generated quality defect information, the system automatically calculates the repair strategy parameters. For blur problems, the system determines the blur processing parameters based on factors such as the size of the blurred area and the degree of blur. For example, if the blurred area is small and the degree of blur is light, a smaller Gaussian kernel parameter may be set for re-filtering to improve the blur without losing too much image detail. For artifact problems, artifact repair parameters are determined based on the type and characteristics of the artifact, such as using a specific artifact removal algorithm and setting corresponding thresholds and other parameters. Based on the overall quality of the image, denoising parameters and image enhancement parameters are also determined to further optimize the image quality. Finally, repair strategy parameters including blur processing parameters, denoising parameters, artifact repair parameters, and image enhancement parameters are generated.

[0032] The target bladder image is processed based on the repair strategy parameters to generate target ultrasound image sequence information. According to the blur processing parameters, an appropriate filtering algorithm is used to process the blurred area to make it clear. Based on the denoising parameters, the image is denoised again to further reduce the impact of noise on the image. Based on the artifact repair parameters, a corresponding algorithm is used to remove artifacts from the image. Based on the image enhancement parameters, an image enhancement algorithm is used to optimize the image to improve its overall quality. After this series of processing, the processed images are arranged in chronological order or a specific logical sequence to generate target ultrasound image sequence information. This image sequence information more accurately reflects the true morphology of the bladder and provides high-quality data support for subsequent operations such as 3D bladder morphology assessment and generation of predictive CT images based on these images.

[0033] S103: Process the target ultrasound image sequence information to generate a target enhanced ultrasound image.

[0034] In one embodiment, the target ultrasound image sequence information is normalized to generate a normalized image sequence. From the target ultrasound image sequence information obtained in the previous step, an ultrasound image at a certain moment is selected as an example. The pixel value range of this frame of image may be between 0 and 255, but there are differences in brightness and contrast between different images, which is not conducive to subsequent unified processing. Through normalization, the image pixel values are mapped to the standard range of 0-1. The normalization formula used is:

[0035] (I(x,y) is the pixel value of the original image at coordinate (x,y), I min and I maxare the minimum and maximum pixel values in the original image, respectively. This process is repeated for each image in the target ultrasound image sequence to obtain a normalized image sequence. After normalization, the pixel value ranges of all images are unified, providing a consistent basis for subsequent processing.

[0036] Based on the preset enhancement algorithm, the normalized image sequence is contrast enhanced to generate an enhanced contrast image sequence. A preset contrast enhancement algorithm formula is used, such as the improved histogram equalization algorithm CLAHE (Contrast-Limited Adaptive Histogram Equalization). The CLAHE algorithm divides the image into multiple small blocks, performs histogram equalization on each small block, and limits the degree of contrast enhancement to avoid excessive amplification of noise. The CLAHE algorithm is applied to the normalized image sequence of this patient, setting the contrast limit parameter to 40 and the block size to 8x8 pixels. After processing, the contrast between the bladder tissue and the surrounding tissue in the image is significantly enhanced, and the structures that were originally difficult to distinguish under low contrast become clearer, generating an enhanced contrast image sequence.

[0037] The contrast-enhanced image sequence was subjected to noise filtering to generate a low-noise image sequence. Edge sharpening was then performed on the low-noise image sequence to generate a sharp-edge image sequence. Since noise is inevitably introduced during ultrasound imaging, it may become more noticeable after contrast enhancement. Therefore, the contrast-enhanced image sequence was subjected to noise filtering. A Gaussian filter algorithm was used, with the standard deviation of the Gaussian kernel set to 1.5. Gaussian filtering effectively smoothes the image by performing a weighted average of each pixel and its neighboring pixels, removing most of the noise. After processing, the noise in the image was suppressed, generating a low-noise image sequence. The bladder's outline and internal structure remained clear, with no significant information lost due to filtering. To highlight the edges of the bladder, the low-noise image sequence was subjected to edge sharpening. Edge sharpening was performed using the Laplacian operator, which detects second-order derivatives in an image to highlight edges and details. The Laplacian operator was applied to each image in the low-noise image sequence, and the result was then added to the original image to achieve edge sharpening. After processing, the edge of the bladder becomes sharper and the boundary with the surrounding tissue is clearer, generating a sharp-edge image sequence. This allows for more accurate identification of the bladder boundary during subsequent assessment of bladder morphology.

[0038] A Gaussian high-pass filter is convolved with the sharp-edge image sequence to generate a detail-enhanced image sequence. To further enhance detail in bladder images, a Gaussian high-pass filter is convolved with the sharp-edge image sequence to highlight high-frequency details. The Gaussian high-pass filter parameters are set to effectively enhance the subtle structures and textures in the bladder images. By performing a convolution operation on each image in the sharp-edge image sequence, the subtle structures within the bladder, such as mucosal folds, become more clearly visible, generating a detail-enhanced image sequence.

[0039] The detail-enhanced image sequence is fused based on the preset image fusion rules to generate a target enhanced ultrasound image. The detail-enhanced image sequence is fused based on the preset image fusion rules. The preset fusion rule can be based on a weighted averaging method, in which multiple images in the detail-enhanced image sequence are weighted averaged according to certain weights. For example, three adjacent detail-enhanced images are assigned weights of 0.3, 0.4, and 0.3, respectively, and then weighted averaged. Through this fusion process, the advantageous information in different images is integrated to generate a target enhanced ultrasound image. This image not only has clear edges and rich details, but also has a low noise level, which can more accurately reflect the true morphology of the bladder and provide high-quality image data for subsequent analysis based on the target bladder morphology assessment model.

[0040] S104: Process the patient's urinary system physiological parameter information based on the target bladder morphology assessment model to generate target physiological state characteristics.

[0041] In one embodiment, a target bladder morphology assessment model is used to classify and process a patient's urinary system physiological parameter information, generating abnormal and normal parameter information. The hospital obtains the patient's urinary system physiological parameter information, including renal function indicators such as creatinine and urea nitrogen from blood tests, urine routine indicators such as red blood cell, white blood cell, and protein content from urine tests, and bladder filling and urination pressure data from cystometry. These parameters are input into the trained target bladder morphology assessment model. The model classifies the parameters based on built-in classification rules and learned knowledge of normal ranges. For example, normal creatinine levels are 53-106 μmol / L for men and 44-97 μmol / L for women. If a patient's creatinine value falls outside this range, the model will classify it as abnormal. Similarly, normal urine should have very few red blood cells, and white blood cell counts are also within the normal range; if they fall outside this range, they are classified as abnormal. In this way, the model classifies the patient's urinary system physiological parameter information into two groups: abnormal and normal.

[0042] Based on the target bladder morphology assessment model, feature extraction is performed on the abnormal parameter information and normal parameter information to generate initial physiological state features. For the abnormal parameter information and normal parameter information obtained in the previous step, the target bladder morphology assessment model further performs feature extraction. Taking creatinine as an example, the model will extract features such as the specific numerical value of the creatinine value, the degree of deviation from the normal range, and the change trend of the recent test value; for white blood cells in urine, features such as the number of white blood cells and the fluctuation of the white blood cell count are extracted. By performing similar processing on all parameters, the key features of each parameter are extracted and combined to form the initial physiological state features. These features reflect the physiological condition of the patient's urinary system from different angles. For example, if the creatinine deviates greatly from the normal range, it may indicate that there are major problems with kidney function, affecting the normal metabolism and excretion function of the bladder.

[0043] Normalize the initial physiological state characteristics to generate a standardized feature vector. Since the dimensions and value ranges of different parameters vary greatly, in order to facilitate subsequent unified analysis, the initial physiological state characteristics are normalized. For example, the creatinine value may range from tens to hundreds, while the urine white blood cell count is usually in the single digit to dozens. It is meaningless to directly compare the two values. Use the normalization formula, such as (x is the original eigenvalue, x min and x max , which are the minimum and maximum values in the training sample set, map all initial physiological state feature values to a standard range of 0-1. After normalization, all features are on the same scale, forming a standardized feature vector. This makes different features comparable and provides a unified data foundation for subsequent analysis.

[0044] The initial physiological state characteristics are processed through correlation analysis to generate correlation information between parameters. The target bladder morphology assessment model performs correlation analysis on the standardized feature vectors to explore the intrinsic relationship between different parameters. For example, it is found that while the number of white blood cells in the patient's urine increases, the bladder pressure fluctuates abnormally during urination. The model concludes through data analysis that there is a positive correlation between the two, that is, the increase in white blood cells may affect the bladder's urination function, thereby causing abnormal bladder pressure. The model may also find the correlation between renal function indicators and routine urine indicators. For example, when creatinine increases, the protein content in the urine also tends to increase. By comprehensively analyzing the correlation between various parameters, the correlation information between parameters is generated. This information helps to gain a deeper understanding of the overall condition of the patient's urinary system and the interaction mechanism between various parameters.

[0045] The correlation information between parameters and the standardized feature vectors are fused to generate the target physiological state feature. The model will comprehensively consider the standardized values of each parameter and the correlation between them. For example, if a strong correlation is found between certain parameters and the standardized values of these parameters deviate from the normal range, the model will give these parameters a greater weight. Through a specific fusion algorithm, all information is integrated together to generate the target physiological state feature. This feature comprehensively reflects the comprehensive physiological state of the patient's urinary system, including both the abnormalities of a single parameter and the mutual influence between parameters. It provides an important basis for the subsequent three-dimensional bladder morphological assessment and prediction of CT images based on this feature, helping doctors to more accurately judge the current physiological state of the patient's bladder and provide strong support for the formulation of radiotherapy plans.

[0046] S105 , processing the target physiological state characteristics, the target enhanced ultrasound image, and the functional imaging data based on the target bladder morphology assessment model to generate a predicted CT image of the target patient.

[0047] In one embodiment, the target physiological state characteristics are feature-encoded based on the target bladder morphology assessment model to generate physiological state characteristic codes. Suppose there is a patient with a pelvic tumor, and after preliminary processing of the physiological parameter information of the urinary system, the target physiological state characteristics are obtained. These characteristics include information such as the patient's renal function indicators (such as abnormal creatinine and urea nitrogen), urine routine indicators (abnormal red blood cell and white blood cell counts), and abnormal bladder pressure. The target bladder morphology assessment model performs feature encoding on these target physiological state characteristics. The model converts each feature into a specific coding form. For example, for the feature that the creatinine value exceeds the normal range, the model encodes it into a set of numbers based on factors such as its degree of deviation and change trend. This set of numbers not only represents the abnormality of creatinine, but also contains potential correlation information with other physiological state characteristics. By encoding all target physiological state characteristics, physiological state characteristic codes are generated. These codes store the physiological information of the patient's urinary system in a compact and easy-to-process manner.

[0048] Feature extraction is performed on the target enhanced ultrasound image to generate ultrasound image features. The target enhanced ultrasound image of the patient was previously processed through a series of processes to obtain the target enhanced ultrasound image. The target bladder morphology assessment model extracts features from this image. The model identifies features such as bladder wall thickness, internal bladder echo uniformity, and bladder shape and contour. For example, if the model detects localized thickening of the bladder wall, it extracts information such as the location and extent of the thickening as features. For areas of uneven internal bladder echoes, it extracts features such as the extent and echo intensity differences. Through these operations, key information about bladder morphology and structure in the target enhanced ultrasound image is converted into ultrasound image features that can be understood and processed by the model. Feature mining is performed on the functional imaging data to generate functional imaging features. The patient also underwent a PET-CT examination, which generated functional imaging data. The target bladder morphology assessment model performs feature mining on this functional imaging data. PET-CT images reflect the metabolic status of bladder tissue. The model analyzes features such as the distribution and metabolic intensity of metabolically active areas. If a region within the bladder is found to have significantly higher metabolic activity than surrounding tissue, the model extracts information such as the location of that region and the ratio of metabolic activity intensity to normal tissue as functional imaging features. These characteristics provide information about bladder function at the metabolic level, helping to provide a more comprehensive understanding of bladder status.

[0049] Physiological status feature codes, ultrasound image features, and functional imaging features are fused to generate a fused feature vector. During the fusion process, the model assigns different weights to different types of features based on their importance and relevance. For example, given the potential impact of a patient's current renal function on bladder morphology, a higher weight may be given to the physiological status feature code; a relatively high weight is also assigned to bladder wall thickening, a key morphological feature in ultrasound images. These features are then combined using a specific fusion algorithm to generate a fused feature vector. This vector integrates information from multiple aspects, including physiology, morphology, and function, to comprehensively describe the patient's bladder status.

[0050] The fused feature vector is subjected to dimensionality reduction and reconstruction to generate initial prediction features, which are then convolved and pooled to generate target prediction features. The fused feature vector contains a large amount of information, but it may contain redundant and complex high-dimensional data, which is not conducive to subsequent processing. Therefore, the model performs dimensionality reduction and reconstruction on the fused feature vector. The model uses dimensionality reduction algorithms such as principal component analysis (PCA) to remove redundant information from the data, extract the main components, and convert the high-dimensional fused feature vector into low-dimensional initial prediction features. While reducing the dimensionality, the model uses reconstruction algorithms to retain key information as much as possible, ensuring that the initial prediction features not only reflect the main content of the original features but also have a more concise form to facilitate subsequent convolution and pooling operations.

[0051] The model performs convolution and pooling on the initial prediction features. The convolution operation slides on the initial prediction features through different convolution kernels to extract local features. For example, a convolution kernel of a specific size and weight is used to detect local information such as edges and textures related to bladder morphology in the prediction features, highlighting key morphological features. The pooling operation further compresses the convolved features, reducing the data dimension while retaining important features. For example, the maximum pooling method is used to select the maximum value in a certain area as the pooling result, reducing the amount of data while enhancing the representativeness of the features. After multiple rounds of convolution and pooling, the target prediction features are generated. These features are more focused on the key morphological and functional characteristics of the bladder, providing more accurate information for generating predicted CT images.

[0052] The target prediction features are upsampled and deconvolved to generate a predicted CT image for the target patient. The model performs upsampling and deconvolution on the target prediction features. The upsampling operation increases the data resolution through methods such as interpolation, gradually bringing the feature map size closer to that of the actual CT image. Deconvolution is the inverse of convolution, reconstructing the target prediction features into an image with richer details. During this process, the model gradually converts the target prediction features into a predicted CT image based on learned CT image features and patterns. For example, based on the learning of a large number of CT images during previous training, the model supplements missing details, making the predicted CT image closer to the actual CT image in terms of morphology, structure, and function, thereby generating a predicted CT image for the target patient. This predicted CT image provides important data support for subsequent comparison with simulated positioning CT images and evaluation of the three-dimensional bladder morphology, helping physicians more accurately assess the patient's actual bladder condition and formulate more precise radiotherapy plans.

[0053] S106 , processing the predicted CT image of the target patient and the simulated positioning CT image information of the target patient based on the target bladder morphology assessment model and the bladder dynamic image information to generate a bladder three-dimensional morphology assessment result.

[0054] In one embodiment, the target bladder morphology assessment model performs feature alignment on the predicted CT image and simulated positioning CT image information of the target patient to generate aligned image features. Suppose a patient with a pelvic tumor obtains a simulated positioning CT image during the radiotherapy simulation positioning phase. Simultaneously, the target bladder morphology assessment model generates a predicted CT image generated based on the target bladder morphology assessment model through the previous steps. The target bladder morphology assessment model performs feature alignment on these two images. The model first identifies key feature points of the bladder in both images, such as key points of the bladder's contour and landmark feature points of its internal structure. Then, using an image registration algorithm, the features of the predicted CT image are matched and aligned with those of the simulated positioning CT image. For example, key features such as the edges and corners of the bladder in the two images are spatially aligned as closely as possible. In this way, aligned image features are generated, laying the foundation for more accurate subsequent comparison and analysis of the differences between the two images, ensuring that bladder morphology is assessed under the same standards.

[0055] Based on the target bladder morphology assessment model, dynamic bladder images are preprocessed to generate bladder wall elasticity features. Real-time ultrasound dynamic monitoring is used to obtain dynamic bladder image information for the patient. The target bladder morphology assessment model preprocesses dynamic bladder images to generate bladder wall elasticity features. The model analyzes the deformation of the bladder wall during filling and emptying. By tracking the position changes of specific points on the bladder wall at different times, the strain and stress of the bladder wall are calculated. For example, when the bladder is full, the degree of stretching of a certain area of the bladder wall is observed, and the elastic modulus of this area is calculated based on mechanical principles as part of the bladder wall elasticity features. These elastic features can reflect the elasticity of the bladder wall and are important for assessing bladder function and morphological changes.

[0056] Based on the target bladder morphology assessment model, the feature vector of the bladder dynamic image is processed with the features of the aligned predicted CT image and the simulated positioning CT image to generate fused features. The bladder wall elasticity feature vector is fused with the aligned image features. The model performs weighted fusion of different features based on their importance and relevance. For example, areas of abnormal elasticity within the bladder wall elasticity feature are given a higher weight, as they may indicate bladder pathology or functional abnormality. Morphological features such as bladder shape and size in the predicted CT image and the simulated positioning CT image are also weighted based on their importance to bladder morphology assessment.

[0057] Different types of features have varying importance in reflecting bladder status, so the first step in a specific fusion algorithm is to determine the weights for each feature. For example, if a patient with a pelvic tumor has abnormal renal function indicators (such as creatinine and urea nitrogen), this may significantly impact bladder metabolism and excretion. Therefore, during fusion, the weights of physiological status features related to renal function will be increased accordingly. Key morphological features in ultrasound images, such as bladder wall thickening or abnormal echogenicity, will also be given a higher weight because they may be important indicators of bladder pathology. In dynamic bladder images, dynamic features such as bladder wall peristalsis frequency and emptying rate are also given higher weights if abnormal changes occur. The weighting method can be based on statistical analysis of large amounts of clinical data or optimized through expert experience and machine learning algorithms to ensure that the impact of important features on the fusion results is highlighted.

[0058] After weight determination, the feature fusion phase begins. Assume that the patient's physiological state feature code, ultrasound image features, functional imaging features, and bladder dynamic image feature vectors have already been extracted. Using a weighted summation fusion approach, these features are linearly combined according to the determined weights. For example, the physiological state feature code is represented by vector A, the ultrasound image features by vector B, the functional imaging features by vector C, and the bladder dynamic image feature vector by vector D. The corresponding weights are ω1, ω2, ω3, and ω4, respectively. The fused feature vector F can be expressed as F = ω1A + ω2B + ω3C + ω4D. This method organically combines features from different modalities, enabling the fused feature to comprehensively reflect multiple aspects of the bladder.

[0059] This fusion process integrates and complements information on bladder morphology, function, and dynamic changes. Physiological state feature encoding reflects the overall physiological status of the patient's urinary system. For example, abnormal renal function may indicate impaired bladder excretion. Ultrasound image features reveal the real-time morphology and structure of the bladder, such as bladder wall thickness and internal echogenicity, which can be used to determine whether morphological abnormalities exist. Functional imaging features provide metabolic information about the bladder's function, identifying potential areas of pathology. Bladder dynamic image features reveal changes in the bladder under different physiological conditions, such as bladder wall elasticity and peristalsis. Through fusion, these complementary features form a comprehensive description of bladder status. For example, if ultrasound images show localized bladder wall thickening, functional imaging features identify metabolic abnormalities in this thickened area, while dynamic bladder images also reveal abnormal peristalsis in this area during urination. By combining these features, the fused features can more accurately reflect potential bladder pathology and its impact on dynamic changes in bladder function and morphology.

[0060] After completing the above steps, a comprehensive fusion feature is ultimately generated. This fusion feature is no longer a single modality, but rather a complex feature vector encompassing multiple aspects of the bladder's status. It deeply fuses information from different sources, providing a rich and comprehensive data foundation for subsequent 3D bladder morphology assessment. In subsequent processing, based on this fusion feature, the model can more accurately enhance, map, and classify features, resulting in a more precise 3D bladder morphology assessment. This helps physicians gain a more comprehensive understanding of the patient's bladder status and develop more appropriate radiotherapy plans.

[0061] The fused features are enhanced to generate a fused enhanced feature sequence. The fused features are enhanced to highlight key information. The model uses some techniques from convolutional neural networks (CNN), such as using specific convolution kernels for convolution operations. These convolution kernels are designed to extract features related to bladder morphology and function, such as edge features and texture features. Through multiple convolutions and nonlinear activation functions, the key information in the fused features is enhanced and noise and irrelevant information is suppressed. For example, the detailed features of the edges and internal structures of the bladder wall are enhanced in contrast and clarity after the convolution operation. After a series of processing, a fused enhanced feature sequence is generated, making the features of the bladder more prominent and facilitating subsequent analysis and classification.

[0062] The fused enhanced feature sequence is subjected to feature mapping and classification processing to generate bladder morphology classification information. The model maps the fused enhanced feature sequence to a specific feature space, where different feature combinations correspond to different bladder morphology categories. Through training, the model learns the characteristic patterns of various categories, including normal bladder morphology, bladder morphology with varying degrees of abnormal filling, and bladder morphology with pathological changes. Then, based on the input fused enhanced feature sequence, the model uses a classification algorithm, such as a support vector machine (SVM) or the classification layer of a neural network, to determine the morphological category of the current bladder and generate bladder morphology classification information. For example, it can determine whether the bladder is in a normal filling state, overfilled, or has morphological abnormalities caused by a tumor.

[0063] The bladder morphology classification information is comprehensively evaluated and quantified to generate a 3D bladder morphology assessment result. The model comprehensively evaluates and quantifies the bladder morphology classification information. The model considers the various categories of bladder morphology classification information, as well as other relevant information obtained from previous processing, such as bladder wall elasticity characteristics and the degree of difference between the predicted CT image and the simulated positioning CT image. A comprehensive assessment of the 3D bladder morphology is performed using pre-set evaluation rules and quantitative indicators. For example, an assessment score is calculated based on factors such as the degree of similarity between the bladder morphology and normal morphology and the degree of abnormal bladder wall elasticity. A higher score indicates a closer-to-normal bladder morphology and a lower risk of adverse effects during radiotherapy. A lower score indicates a more severe abnormality and a greater likelihood of requiring adjustment to the radiotherapy plan. The resulting 3D bladder morphology assessment provides physicians with an accurate and quantitative basis for radiotherapy planning, helping them determine whether to adjust parameters such as the target volume and dose to mitigate radiotherapy side effects caused by changes in bladder filling.

[0064] The server acquires multi-source data, including bladder ultrasound images, simulated positioning CT images, urinary system physiological parameters, functional imaging data, and dynamic bladder images, providing a comprehensive basis for assessment. Bladder ultrasound images undergo sequential preprocessing and enhancement to improve image quality and highlight bladder structural information. Models are used to process multi-source data. For example, urinary system physiological parameters are classified, feature extracted, normalized, and correlated with each other to generate target physiological state features. Ultrasound images and functional imaging data are then feature mined, fused, and processed in multiple steps to generate predicted CT images. Finally, combined with dynamic bladder images, the predicted CT images and simulated positioning CT images undergo multi-step processing to complete the three-dimensional bladder morphology assessment. This entire process leverages deep learning and multimodal data fusion technologies to transform cross-modal assessment into homomodal assessment, significantly improving the intuitiveness and accuracy of the assessment. This effectively assists physicians in developing precise radiotherapy plans, reduces radiotherapy side effects caused by bladder filling changes, and provides strong support for pelvic tumor radiotherapy.

[0065] In one embodiment, Figure 2 As shown, the present application also provides a device for evaluating the transmembrane state of the bladder three-dimensional morphology, comprising:

[0066] Acquisition module 201 is used to acquire bladder ultrasound image information of a target patient, simulated positioning CT image information of a target patient, physiological parameter information of the urinary system of the target patient, functional imaging data, and dynamic bladder image information, wherein the dynamic bladder image information is used to characterize the morphological changes of the bladder under different physiological states;

[0067] The processing module 202 is used to preprocess the bladder ultrasound image information of the target patient to generate target ultrasound image sequence information; process the target ultrasound image sequence information to generate a target enhanced ultrasound image; process the patient's urinary system physiological parameter information based on the target bladder morphology assessment model to generate a target physiological state characteristic; process the target physiological state characteristic, the target enhanced ultrasound image, and the functional imaging data based on the target bladder morphology assessment model to generate a predicted CT image of the target patient; and process the predicted CT image of the target patient and the simulated positioning CT image information of the target patient based on the target bladder morphology assessment model in combination with the bladder dynamic image information to generate a three-dimensional bladder morphology assessment result.

[0068] Each embodiment of this application is described in a related manner. Similar portions between embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments. In particular, the embodiments of the method for assessing the transmembrane state of a bladder's three-dimensional morphology, electronic device, electronic device, and readable storage medium are generally similar to the aforementioned embodiment of the method for assessing the transmembrane state of a bladder's three-dimensional morphology, so the description is relatively simple. For relevant portions, reference can be made to the description of the aforementioned embodiment of the method for assessing the transmembrane state of a bladder's three-dimensional morphology.

Claims

1. A method for evaluating the transmembrane state of the three-dimensional bladder morphology, characterized in that: include: Obtaining bladder ultrasound image information, simulated positioning CT image information, urinary system physiological parameter information, functional imaging data, and bladder dynamic image information of the target patient, wherein the bladder dynamic image information is used to characterize the morphological changes of the bladder under different physiological states; Preprocessing the bladder ultrasound image information of the target patient to generate target ultrasound image sequence information; Processing target ultrasound image sequence information to generate a target enhanced ultrasound image; The patient's urinary system physiological parameter information is processed based on the target bladder morphology assessment model to generate target physiological state characteristics; Based on the target bladder morphology assessment model, the target physiological state characteristics, target enhanced ultrasound images, and functional imaging data are processed to generate a predicted CT image of the target patient; Based on the target bladder morphology assessment model and combined with the bladder dynamic image information, the predicted CT image of the target patient and the simulated positioning CT image information of the target patient are processed to generate the bladder three-dimensional morphology assessment result.

2. The method according to claim 1, wherein Preprocess the target patient's bladder ultrasound image information to generate target ultrasound image sequence information, including: Denoising and contrast enhancement are performed on the target patient's bladder ultrasound image information to generate an initial bladder image; performing bladder region segmentation processing on the initial bladder images respectively to generate target bladder images, wherein the target bladder images are images containing only the bladder region; Performing image quality assessment on the target bladder image to generate quality defect information, wherein the quality defect information is used to characterize quality problems such as image blur and artifacts as well as their locations and types; Processing the quality defect information to generate repair strategy parameters, wherein the repair strategy parameters include blur processing parameters, denoising parameters, artifact repair parameters, and image enhancement parameters; The target bladder image is processed based on the repair strategy parameters to generate target ultrasound image sequence information.

3. The method according to claim 2, wherein Processing target ultrasound image sequence information to generate target enhanced ultrasound images includes: Normalizing the target ultrasound image sequence information to generate a normalized image sequence; Performing contrast enhancement processing on the normalized image sequence based on a preset enhancement algorithm to generate an enhanced contrast image sequence; Perform noise filtering on the contrast-enhanced image sequence to generate a low-noise image sequence; Perform edge sharpening processing on the low-noise image sequence to generate a sharp-edge image sequence; Perform convolution processing based on Gaussian high-pass filter and sharp edge image sequence to generate detail enhanced image sequence; The detail-enhanced image sequence is fused based on preset image fusion rules to generate a target-enhanced ultrasound image.

4. The method according to claim 1, wherein Obtain a target bladder morphology assessment model, including: Obtaining a training sample set and a preset bladder morphology assessment model, wherein the training sample set includes multimodal data and corresponding bladder three-dimensional morphology assessment results; Count the number of features of each modal data in the training sample set and generate a sampling ratio based on the balance of feature distribution; Perform multimodal stratified sampling on the training sample set based on the sampling ratio to generate a preset number of sampling feature combinations; Based on a statistical test of any key feature combined with each sampling feature, the data is divided into a response group and a poor response group, each group contains a preset number of data samples, and at least one sample has lesion identification information; Iteratively train the pre-set bladder morphology assessment model based on the response group and the poor response group, optimize the model parameters through cross-validation, and generate the trained model and performance indicators; If the lesion identification information in the training results is recognized by the model as a key feature for morphological assessment, the trained model is used as the target bladder morphological assessment model.

5. The method according to claim 1, wherein The target bladder morphology assessment model is used to process the patient's urinary system physiological parameter information to generate target physiological state characteristics, including: Classify and process the patient's urinary system physiological parameter information based on the target bladder morphology assessment model to generate abnormal parameter information and normal parameter information; Based on the target bladder morphology assessment model, feature extraction and processing are performed on abnormal parameter information and normal parameter information to generate initial physiological state features; Normalize the initial physiological state characteristics to generate a standardized feature vector; Perform correlation analysis on the initial physiological state characteristics to generate correlation relationship information between parameters; The correlation information between parameters and the standardized feature vector are fused to generate the target physiological state characteristics.

6. The method according to claim 1, wherein Based on the target bladder morphology assessment model, the target physiological state characteristics, target enhanced ultrasound images, and functional imaging data are processed to generate a predicted CT image of the target patient, including: Performing feature coding processing on target physiological state features based on the target bladder morphology assessment model to generate physiological state feature codes; Performing feature extraction processing on the target enhanced ultrasound image to generate ultrasound image features; Perform feature mining on functional imaging data to generate functional imaging features; Fusing physiological state feature coding, ultrasound image features, and functional imaging features to generate a fused feature vector; Perform dimension reduction and reconstruction on the fused feature vector to generate initial prediction features; Perform convolution and pooling on the initial prediction features to generate target prediction features; The target prediction features are upsampled and deconvolved to generate the predicted CT image of the target patient.

7. The method according to claim 6, wherein Based on the target bladder morphology assessment model and combined with bladder dynamic image information, the predicted CT image of the target patient and the simulated positioning CT image information of the target patient are processed to generate a three-dimensional bladder morphology assessment result, including: Based on the target bladder morphology assessment model, feature alignment processing is performed on the predicted CT image and the simulated positioning CT image information of the target patient to generate aligned image features; The bladder dynamic image is preprocessed based on the target bladder morphology assessment model to generate bladder wall elasticity features; Based on the target bladder morphology assessment model, the feature vector of the bladder dynamic image is processed with the features of the aligned predicted CT image and the simulated positioning CT image to generate a fusion feature; Perform feature enhancement processing on the fused features to generate a fused enhanced feature sequence; Perform feature mapping and classification processing on the fused enhanced feature sequence to generate bladder morphology classification information; The bladder morphology classification information is comprehensively evaluated and quantified to generate the bladder three-dimensional morphology evaluation results.

8. A device for evaluating the transmembrane state of a bladder's three-dimensional morphology, characterized in that: The device comprises: An acquisition module is used to acquire bladder ultrasound image information of a target patient, simulated positioning CT image information of a target patient, physiological parameter information of the urinary system of the target patient, functional imaging data, and dynamic bladder image information, wherein the dynamic bladder image information is used to characterize the morphological changes of the bladder under different physiological states; The processing module is used to preprocess the bladder ultrasound image information of the target patient to generate target ultrasound image sequence information; process the target ultrasound image sequence information to generate a target enhanced ultrasound image; process the patient's urinary system physiological parameter information based on the target bladder morphology assessment model to generate target physiological state characteristics; process the target physiological state characteristics, target enhanced ultrasound image, and functional imaging data based on the target bladder morphology assessment model to generate a predicted CT image of the target patient; and process the target patient's predicted CT image and the target patient's simulated positioning CT image information based on the target bladder morphology assessment model in combination with bladder dynamic image information to generate a three-dimensional bladder morphology assessment result.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the method for assessing the transmembrane state of the three-dimensional bladder morphology according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the method for evaluating the transmembrane state of the three-dimensional bladder morphology according to any one of claims 1 to 7 is implemented.

Images