Kidney function prediction method and device, electronic equipment and medium

Abstract

The invention discloses a kidney function prediction method and device, electronic equipment and a medium. The method comprises the steps that a kidney cortical medullary enhancement period image of a target patient is acquired; inputting a kidney cortical medullary enhancement stage image of the target patient into the trained segmentation model to segment a cortical region in the cortical medullary enhancement stage image to obtain a cortical region image of the kidney of the target patient; extracting radiation group features of the cortex region image; and inputting the radiation group features of the cortex region image into the trained renal function staging prediction model, and predicting the renal function staging of the target patient. The invention assesses renal function by a non-invasive and reliable alternative method.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a method and device for predicting split renal function, an electronic device, and a medium. Background Art

[0002] Obstructive Nephropathy (ON) is a kidney disease caused by urinary tract obstruction, which leads to impaired urine flow and gradually damages renal function. This obstruction may result from internal factors such as ureteral stones or stenosis, or external compression caused by tumors, cysts, etc. Accurate assessment of ON is crucial for guiding treatment decisions, and imaging diagnosis plays a central role in the assessment of ON. Among various imaging techniques, multi-phase contrast-enhanced Computed Tomography Urography (CTU) is widely used because it can accurately locate the position and severity of urinary tract obstruction through imaging phases such as non-contrast scan, corticomedullary phase, nephrographic phase, and excretory phase.

[0003] Many indicators are used to evaluate renal function. Glomerular Filtration Rate (GFR) is an important indicator for evaluating renal function, but its accurate measurement still faces challenges. Although formulas such as MDRD-7 and CKD-EPI are widely used to estimate GFR, their performance varies among different patient groups, which may lead to inaccuracies and thus limit their clinical utility. Emission Computed Tomography (ECT) is the gold standard for evaluating renal function, but its invasiveness and dependence on radioactive drugs limit its clinical application, especially in patients with impaired renal function. In addition, ECT requires intravenous injection of radioactive drugs for imaging, which increases the economic burden on patients.

[0004] The renal cortex is a key site for filtration and is extremely sensitive to structural and functional changes caused by ON. Conditions such as hydronephrosis or compression by adjacent lesions can lead to cortical thinning, which is closely related to the decline in renal function. Radiologists usually rely on the cortical morphology shown in CTU images to infer functional damage. However, this assessment highly depends on individual experience, is highly subjective, and has poor inter-observer consistency. In recent years, deep learning techniques, especially three-dimensional convolutional neural networks (3D CNNs), have shown significant potential in medical image tasks. These techniques have shown good application prospects in fields such as organ segmentation, disease detection, and functional analysis. However, many existing models are limited by small dataset sizes, limited attention to clinical utility, and failure to fully integrate into the actual diagnosis process. Considering the above problems, there is an urgent need for a non-invasive and reliable alternative method to evaluate renal function. Summary of the Invention

[0005] The present invention provides a method, device, electronic device and medium for predicting split renal function, which are used for a non-invasive and reliable alternative method to evaluate renal function.

[0006] According to an aspect of the present invention, there is provided a method for predicting split renal function, including:

[0007] Obtain the corticomedullary enhancement phase image of the target patient;

[0008] Input the corticomedullary enhancement phase image of the target patient into the trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image, and obtain the cortical region image of the target patient's kidney;

[0009] Extract the radiomic features of the cortical region image;

[0010] Input the radiomic features of the cortical region image into the trained renal function staging prediction model to predict the renal function stage of the target patient.

[0011] Optionally, before inputting the corticomedullary enhancement phase image of the target patient into the trained segmentation model, it further includes:

[0012] Obtain the corticomedullary enhancement phase images of several patients;

[0013] Divide the corticomedullary enhancement phase image into two images each having a unilateral kidney;

[0014] Label the cortical region in the corticomedullary enhancement phase image and construct the first data set;

[0015] Train the segmentation model based on the first data set to obtain the trained segmentation model.

[0016] Optionally, after obtaining the corticomedullary enhancement phase images of several patients, it further includes

[0017] Preprocess the corticomedullary enhancement phase image, and the preprocessing at least includes data cleaning, data cropping, resampling and data normalization processing.

[0018] Optionally, the segmentation model adopts the nnUNet v2 neural network segmentation model, selects the network configuration of 3d_fullres, and uses PlainConvUNet as the basic network structure; then the specific network parameters of the segmentation model include:

[0019] It consists of 7 different feature stages, and the number of feature channels in each stage is 32, 64, 128, 256, 320, 320, 320 respectively;

[0020] 3D convolution is used in each stage. The convolution kernels in the first stage are 1, 3, 3; the convolution kernels in the second stage are 1, 3, 3; the convolution kernels from the third stage to the seventh stage are 3, 3, 3;

[0021] The stride settings for each stage are as follows: the stride settings in the first stage are 1, 1, 1; the stride settings in the second stage are 1, 2, 2; the stride settings in the third stage are 1, 2, 2; the stride settings in the fourth stage are 2, 2, 2; the stride settings in the fifth stage are 2, 2, 2; the stride settings in the sixth stage are 1, 2, 2; the stride settings in the seventh stage are 1, 2, 2;

[0022] There are 2 convolutional layers in each stage of the encoder and decoder of the segmentation model, namely 2, 2, 2, 2, 2, 2, 2; the Dropout layer is not used in the segmentation model.

[0023] Optionally, the loss function of the segmentation model is:

[0024] Loss total = Loss dice + Loss CE

[0025]

[0026] Loss CE = -∑(Ylog(P)+(1 - Y)log(1 - p))

[0027] In the formula, Loss dice is the dice loss; Loss CE is the cross - entropy loss; Loss to tal is the total loss; I is the total number of pixel points; K is the total number of samples; u is the Softmax probability output; v is the hard - coded true annotation; Y represents the true label; P represents the predicted probability of the model.

[0028] Optionally, after inputting the kidney corticomedullary enhancement - phase image of the target patient into the trained segmentation model to segment the cortical area in the corticomedullary enhancement - phase image, it further includes:

[0029] Compare the points in the probability map containing the cortical area segmented by the segmentation model with a preset first threshold respectively, and convert the probability map into a binary mask map according to the comparison results;

[0030] Perform dilation and erosion operations on the binary mask map to remove artifacts and noise in the binary mask map;

[0031] Perform connected component analysis on the binary mask image after dilation and erosion operations to remove the connected regions smaller than a preset second threshold in the binary mask image, thereby obtaining the cortical region image.

[0032] Optionally, before inputting the radiomics features of the cortical region image into the trained renal function staging prediction model, it further includes:

[0033] Obtain the cortical region images of the kidneys of a number of patients;

[0034] Extract the radiomics features of the cortical region images, and construct a second data set with the radiomics features of a group of cortical region images as one sample;

[0035] Train the renal function staging prediction model based on the second data set to obtain the trained renal function staging prediction model.

[0036] According to another aspect of the present invention, there is provided a device for predicting split renal function, including:

[0037] An acquisition unit for acquiring the corticomedullary enhancement phase image of the kidney of a target patient;

[0038] A segmentation unit for inputting the corticomedullary enhancement phase image of the kidney of a target patient into the trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image, thereby obtaining the cortical region image of the kidney of the target patient;

[0039] A feature extraction unit for extracting the radiomics features of the cortical region image;

[0040] A prediction unit for inputting the radiomics features of the cortical region image into the trained renal function staging prediction model to predict the renal function stage of the target patient.

[0041] According to another aspect of the present invention, there is provided an electronic device, the electronic device includes:

[0042] At least one processor; and

[0043] A memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the split renal function prediction method according to any embodiment of the present invention.

[0045] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement the split renal function prediction method according to any embodiment of the present invention when executed.

[0046] The technical solution of the embodiment of the present invention automatically segments the cortical region of the enhanced-phase image of the renal cortex and medulla in the medical image through a trained segmentation model, extracts the radiomic features of the cortical region image, and then inputs the radiomic features of the cortical region image into the trained renal function staging prediction model, which can quickly predict the stage of nephropathy patients, provide valuable guidance for clinical decision-making, and help clinicians better evaluate and manage obstructive nephropathy patients.

[0047] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 is a flowchart of the split renal function prediction method provided in Embodiment 1 of the present invention;

[0050] Figure 2 is a flowchart of the split renal function prediction method provided in Embodiment 2 of the present invention;

[0051] Figure 3 is a structural diagram of a split renal function prediction device provided in Embodiment 2 of the present invention;

[0052] Figure 4 is a schematic structural diagram of an electronic device for implementing the split renal function prediction method of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0054] It should be noted that the terms "first", "second", etc. in the specification, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0055] Embodiment 1

[0056] Figure 1 FIG. is a flowchart of a method for predicting split renal function provided in Embodiment 1 of the present invention. As Figure 1 shown, the method includes:

[0057] S101. Obtain the renal corticomedullary enhancement phase image of the target patient.

[0058] Among them, the renal corticomedullary enhancement phase image is used to observe the enhancement performance of the renal cortex and medulla after injecting contrast agent, helping doctors judge whether the structure and function of the kidney are normal, and diagnosing kidney diseases. In this embodiment, multi-phase enhanced computed tomography urography (CTU) can be used, which can accurately locate the position and severity of urinary tract obstruction through imaging stages such as plain scan, corticomedullary phase, renal parenchymal phase and excretion phase. The renal corticomedullary enhancement phase CTU image of the patient can be obtained, and the obtained renal corticomedullary enhancement phase CTU image can be preprocessed to obtain an image that meets the input format of the segmentation model. For example, the renal corticomedullary enhancement phase CTU image can be converted into an image with a collection slice thickness of 5 mm and a resolution of 512×512 pixels; and the image can be data-cleaned to ensure the quality and reliability of the image.

[0059] S102. Input the renal corticomedullary enhancement phase image of the target patient into the trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image, and obtain the cortical region image of the target patient's kidney.

[0060] It should be noted that the segmentation model is a medical image segmentation neural network model, which is used to segment the cortical region in the enhanced-phase image of the renal cortex and medulla, and generate a cortical region image. The cortical region image is a mask image that only contains the cortical region, so as to be used for subsequent analysis and processing of the mask image of the cortical region, and for staging the renal function of the target patient.

[0061] S103. Extract the radiomic features of the cortical region image.

[0062] Among them, radiomic features refer to a series of quantitative features extracted from medical imaging images. These features can reflect the imaging manifestations of tissues or lesions, and are helpful for the diagnosis, treatment, and prognosis evaluation of diseases. Common radiomic features include morphological features, density features, texture features, histogram features, etc. Among them, morphological features can include the size, shape, and edge of the target in the image; density features include average density, density uniformity, and CT value; texture features include gray-level co-occurrence matrix (GLCM); run-length matrix (RLM) features; local binary pattern (LBP) features; histogram features include gray-level histogram and normalized histogram.

[0063] It should be noted that some features of the renal cortex can assist in staging the renal function. These features include: renal cortex thickness, renal cortex echo, renal cortex blood perfusion, and renal cortex morphology, etc.

[0064] S104. Input the radiomic features of the cortical region image into the trained renal function staging prediction model to predict the renal function stage of the target patient.

[0065] It should be noted that the renal function staging prediction model can adopt a machine learning model, taking various radiomic features as input and the renal function stage in the ECT diagnosis report as the target output feature, so as to predict the renal function stage of the patient.

[0066] In this embodiment, various radiomic features of the renal cortex region image in the cortical region image can be extracted as the input of the renal function staging prediction model, and then the renal function stage of the target patient can be predicted.

[0067] The technical solution of the embodiment of the present invention can automatically segment the cortical region of the enhanced-phase image of the renal cortex and medulla in the medical image through the trained segmentation model, extract the radiomic features of the cortical region image, and then input the radiomic features of the cortical region image into the trained renal function staging prediction model, which can quickly predict the stage of nephropathy patients, provide valuable guidance for clinical decision-making, and help clinicians better evaluate and manage obstructive nephropathy patients.

[0068] Embodiment 2

[0069] Figure 2 This is a flowchart of a method for predicting split renal function provided in the second embodiment of the present invention. As Figure 2 shown, the method includes:

[0070] S201. Obtain the renal cortex-medulla enhancement-phase image of the target patient.

[0071] The target patient is the patient to be measured. In this embodiment, the renal cortex-medulla enhancement-phase CTU image of the target patient can be obtained, and the renal cortex-medulla enhancement-phase CTU image is converted into an image with a slice thickness of 5 mm and a resolution of 512×512 pixels; and the image can be subjected to data cleaning to ensure the quality and reliability of the image.

[0072] It should be noted that the renal cortex-medulla enhancement-phase CTU image of the target patient can also be divided into two images each having a unilateral kidney, so as to perform cortical segmentation processing on the images containing the unilateral kidney respectively.

[0073] S202. Input the renal cortex-medulla enhancement-phase image of the target patient into the trained segmentation model to segment the cortical region in the cortex-medulla enhancement-phase image.

[0074] Before step S202, it further includes:

[0075] S2021. Obtain the renal cortex-medulla enhancement-phase images of several patients.

[0076] S2022. Perform preprocessing on the cortex-medulla enhancement-phase image, and the preprocessing at least includes data cleaning, data cropping, resampling, and data normalization processing.

[0077] It should be noted that the renal cortex-medulla enhancement-phase CTU images of multiple patients that have been obtained can be obtained, and all the renal cortex-medulla enhancement-phase CTU images are converted into images with a slice thickness of 5 mm and a resolution of 512×512 pixels, and the images are preprocessed. The preprocessing includes data cleaning, data cropping, resampling, and data normalization processing. Among them, data cleaning is to perform data cleaning on the image data to remove abnormal data in the image; data cropping is to crop the non-zero region in the image data to reduce the consumption of computing resources. Resampling is to normalize the intervals of all images to the median interval of the dataset by default for the different voxel intervals existing in the image dataset, and the original image data is resampled using cubic spline interpolation, while the mask data uses the nearest neighbor interpolation method; data normalization processing is to perform z-score normalization processing on the image data of each patient to standardize the data differences between different patients and enhance the robustness of the model.

[0078] S2023. Divide the corticomedullary enhancement phase image into two images each having a unilateral kidney.

[0079] It should be noted that, for better segmentation of the patient's kidneys, in this embodiment, the corticomedullary enhancement phase image of the patient's kidneys can be divided into images with unilateral kidneys, so as to facilitate subsequent segmentation and prediction of each kidney separately.

[0080] S2024. Mark the cortical region in the corticomedullary enhancement phase image and construct a first data set.

[0081] In this embodiment, the 3D Slicer software can be used to manually mark the cortical structure of each side of the kidney layer by layer to generate a corresponding mask file as a label. Using the bilateral corticomedullary enhancement phase CTU images of each patient as a sample, a first data set is constructed for subsequent training of the segmentation model.

[0082] It should be noted that in this embodiment, data augmentation processing can also be performed on the corticomedullary enhancement phase CTU image. For example, operations such as random rotation, random scaling, random elastic transformation, gamma correction, and mirroring can be performed on the image to generate various image data for training the segmentation model to improve the generalization ability of the model.

[0083] S2025. Train a segmentation model based on the first data set to obtain the trained segmentation model.

[0084] During the training process, the first data set can be divided into a training set and a test set according to a ratio of 4:1. Using the corticomedullary enhancement phase CTU image as the input and the marked cortical region as the label, and adopting the five-fold cross-validation method to train the segmentation model. When the model parameters of the segmentation model converge or reach the preset number of training times, the training process is completed to obtain the trained segmentation model.

[0085] In one embodiment, the segmentation model uses the nnUNet v2 neural network segmentation model, selects the network configuration of 3d_fullres, and uses PlainConvUNet as the basic network structure; the specific network parameters of the segmentation model include: consisting of 7 different feature stages, and the number of feature channels in each stage is 32, 64, 128, 256, 320, 320, 320 respectively; 3D convolution is used in each stage, the convolution kernel in the first stage is 1, 3, 3; the convolution kernel in the second stage is 1, 3, 3; the convolution kernels in the third stage to the seventh stage are 3, 3, 3; the stride settings in each stage: the stride setting in the first stage is 1, 1, 1; the stride setting in the second stage is 1, 2, 2; the stride setting in the third stage is 1, 2, 2; the stride setting in the fourth stage is 2, 2, 2; the stride setting in the fifth stage is 2, 2, 2; the stride setting in the sixth stage is 1, 2, 2; the stride setting in the seventh stage is 1, 2, 2; there are 2 convolutional layers in each stage of the encoder and decoder of the segmentation model, that is, 2, 2, 2, 2, 2, 2, 2; the segmentation model does not use the Dropout layer. And in this embodiment, the specified non-linear activation function is torch.nn.LeakyReLU, and Leaky ReLU can effectively avoid the problem of gradient disappearance and enhance the non-linear expression ability of the model. The non-linear activation function is set to calculate in-place to save memory. And deep supervision is set. In addition, it is possible to ensure that more than 1 / 3 of the pixels in a batch are foreground class pixels during Patch sampling, thereby enhancing the stability of the network.

[0086] Among them, the loss function of the segmentation model is:

[0087] Loss total = Loss dice + Loss CE

[0088]

[0089] Loss CE = -∑(Ylog(P)+(1 - Y)log(1 - p))

[0090] In the formula, Loss dice is the dice loss; Loss CE is the cross-entropy loss; Loss total is the total loss; I is the total number of pixel points; K is the total number of samples; u is the Softmax probability output; v is the hard-coded true annotation ground truth; Y represents the true label; P represents the predicted probability of the model.

[0091] The parameter update method of the segmentation model is:

[0092]

[0093] Among them, θ t+1 represents the updated weight parameter; θ t represents the convolutional weight parameter before update; l r is the current learning rate; represents the first-order moment estimate after bias correction; is the second-order moment estimate after bias correction; ε is a hyperparameter.

[0094] S203. Post-process the probability map of the segmented cortical region to obtain a cortical region image.

[0095] In one embodiment, step S203 includes:

[0096] S2031. Compare the points in the probability map of the cortical region segmented by the segmentation model with a preset first threshold respectively, and convert the probability map into a binary mask map according to the comparison results.

[0097] S2032. Perform dilation and erosion operations on the binary mask map to remove artifacts and noise in the binary mask map.

[0098] S2033. Perform connected component analysis on the binary mask map after dilation and erosion operations to remove connected regions smaller than a preset second threshold in the binary mask map, and obtain the cortical region image.

[0099] Among them, the generated probability map is converted into a binary mask through a threshold (i.e., the segmentation of the background and the foreground), and dilation and erosion operations are used to remove small artifacts and noise in the binary mask map. Among them, the dilation operation can connect separated regions, while the erosion operation can remove small objects. Analyze the connected components in the segmentation result, and remove the connected regions smaller than the preset second threshold in the binary mask map (remove regions smaller than a certain volume), so as to smooth the segmentation result.

[0100] S204. Extract the radiomic features of the cortical region image.

[0101] After the renal cortical region is segmented, the PyRadiomics package can be used to extract radiomic features from the mask.

[0102] It should be noted that multiple radiomic features can be extracted from each cortical region image, and the features include morphological features, density features, texture features, histogram features, etc.

[0103] Among them, morphological features may include the size, shape, and edges of the target in the image. Specifically, size features include measurement values such as the major axis, minor axis, area, and volume of the tumor or lesion, which can be used to evaluate the progression of the lesion; shape features include shapes such as round, oval, and irregular, and certain specific shapes may be related to the type or nature of the disease; edge features include features such as the smoothness of the edge, whether there are spicules or lobulations, which are helpful for judging the benign or malignant nature of the lesion. For example, the edges of malignant tumors are often irregular and may have spiculation signs.

[0104] Density features include average density, density uniformity, and CT value. Among them, the average density reflects the overall density of the tissue or lesion. There are certain differences in the density of different tissues in imaging, which helps to distinguish different types of tissues; the density uniformity feature includes whether the density inside the lesion is uniform. Uneven density may indicate necrosis, cystic change, or a mixture of different components in the lesion; the CT value feature includes that in CT images, the CT value range of a specific tissue or lesion can be used as one of its features, which is helpful for differential diagnosis. For example, the CT value of adipose tissue is relatively low, while the CT value of bone is relatively high.

[0105] Texture features include gray-level co-occurrence matrix (GLCM); run-length matrix (RLM) features; local binary pattern (LBP) features; among them, the gray-level co-occurrence matrix (GLCM) features are obtained by calculating the occurrence frequency of pixel pairs with different gray levels in the image, and features such as energy, entropy, contrast, and correlation can be obtained, which reflect the coarseness and complexity of the image texture. For example, a high energy value indicates that the image texture is relatively regular, and a high entropy value indicates complex texture; run-length matrix (RLM) features include calculations based on the continuous run lengths of pixels with the same gray value in the image, such as short-run emphasis, long-run emphasis, and gray-level inhomogeneity, which can describe the directionality and uniformity of the texture; local binary pattern (LBP) features include comparing each pixel in the image with its neighboring pixels and generating a binary pattern according to the comparison results, which can be used to analyze the local texture features of the image and have a certain invariance to image rotation and illumination changes.

[0106] Histogram features include gray-level histogram and normalized histogram. Among them, the gray-level histogram shows the frequency distribution of different gray values in the image, and parameters such as its peak value, mean value, and standard deviation can reflect the overall gray features of the image. For example, a higher mean value indicates that the image is brighter overall, and a larger standard deviation indicates a more dispersed gray value distribution; the normalized histogram is obtained by normalizing the gray-level histogram so that its sum is 1, which is convenient for comparing the gray distribution features of different images and can be used to detect abnormal regions or lesions in the image.

[0107] S205. Input the radiomic features of the cortical region image into the trained renal function staging prediction model to predict the renal function stage of the target patient.

[0108] In one embodiment, before step S205, it further includes:

[0109] S2051. Obtain the cortical region images of the kidneys of several patients.

[0110] S2052. Extract the radiomic features of the cortical region images, and construct a second data set with the radiomic features of a group of cortical region images as a sample.

[0111] S2053. Train the renal function staging prediction model based on the second data set to obtain the trained renal function staging prediction model.

[0112] In this embodiment, the original corticomedullary enhanced-phase CTU images of the kidneys of multiple patients that have been collected can be obtained, the cortical region images in the original corticomedullary enhanced-phase CTU images are extracted, and the cortical region images are separated from the bilateral split kidney images at the midline to generate a mask file of the cortical structure. Among them, the method for extracting the cortical region images in the original corticomedullary enhanced-phase CTU images can include the method of manual extraction, or the original corticomedullary enhanced-phase CTU images can be input into a segmentation model to extract the cortical region images of the bilateral kidneys of the patient.

[0113] By extracting 128 radiomic features of each patient's cortical region image as a sample, a second data set is constructed. The second data set is divided into a training set and a test set according to a ratio of 4:1. With 128 radiomic features as the input and the renal function staging in the ECT diagnosis report as the target output feature, a 5-fold cross-validation strategy is used to train the renal function staging prediction model to ensure the robustness and generalization of the model.

[0114] The technical solution of the embodiment of the present invention uses the nnUNet v2 deep adaptive neural network to segment the bilateral renal cortex, thereby realizing the automatic segmentation of the high-precision renal cortex, and further extracting features based on radiomics and training the XGBoost regression model to predict the split renal function. This method not only improves the recognition accuracy of the renal cortical region, but also can accurately and non-invasively predict the renal function staging of patients with obstructive nephropathy. It can provide valuable guidance for clinical decision-making and help clinicians better evaluate and manage patients with obstructive nephropathy.

[0115] Embodiment III

[0116] Figure 3 It is a schematic structural diagram of a split renal function prediction device provided by Embodiment III of the present invention. As Figure 3 shown, the device includes:

[0117] An acquisition unit 301, configured to acquire the corticomedullary enhanced-phase image of the kidney of a target patient;

[0118] A segmentation unit 302 for inputting the corticomedullary enhancement phase image of a target patient into a trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image, thereby obtaining a cortical region image of the target patient's kidney;

[0119] A feature extraction unit 303 for extracting radiomic features of the cortical region image;

[0120] A prediction unit 304 for inputting the radiomic features of the cortical region image into a trained renal function staging prediction model to predict the renal function stage of the target patient.

[0121] Example 4

[0122] Figure 4 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, for example, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, for example, a personal digital processor, a cellular phone, a smart phone, a wearable device (such as a helmet, glasses, a watch, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0123] As Figure 4 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0124] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0125] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a method for predicting split renal function.

[0126] In some embodiments, a method for predicting split renal function can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for predicting split renal function described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute a method for predicting split renal function in any other suitable manner (e.g., by means of firmware).

[0127] The various embodiments of the systems and technologies described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: implemented in one or more computer programs, the one or more computer programs can be executed and / or interpreted on a programmable system including at least one programmable processor, the programmable processor can be a dedicated or general-purpose programmable processor, can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0128] A computer program for implementing the method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, a special purpose computer, or other programmable data processing device, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine, or entirely on a remote machine or server.

[0129] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0130] In order to provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0131] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected with each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0132] A computing system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0133] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0134] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting split renal function, characterized in that, Including: Obtain the corticomedullary enhancement phase image of the target patient; Input the corticomedullary enhancement phase image of the target patient into the trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image, and obtain the cortical region image of the target patient's kidney; Extract the radiomic features of the cortical region image; Input the radiomic features of the cortical region image into the trained renal function staging prediction model to predict the renal function stage of the target patient.

2. The method for predicting split renal function according to claim 1, wherein Before inputting the corticomedullary enhancement phase image of the target patient into the trained segmentation model, it further includes: Obtain the corticomedullary enhancement phase images of several patients; Divide the corticomedullary enhancement phase images into two images each with a unilateral kidney; Label the cortical region in the corticomedullary enhancement phase image and construct the first dataset; Train the segmentation model based on the first dataset to obtain the trained segmentation model.

3. The method for predicting split renal function according to claim 2, wherein After obtaining the corticomedullary enhancement phase images of several patients, it further includes Preprocess the corticomedullary enhancement phase images, and the preprocessing at least includes data cleaning, data cropping, resampling, and data normalization processing.

4. The method for predicting split renal function according to claim 2, wherein The segmentation model uses the nnUNetv2 neural network segmentation model, selects the network configuration of 3d_fullres, and uses PlainConvUNet as the basic network structure; then the specific network parameters of the segmentation model include: Composed of 7 different feature stages, and the number of feature channels in each stage is 32, 64, 128, 256, 320, 320, 320 respectively; 3D convolution is used in each stage. The convolution kernel in the first stage is 1, 3, 3; the convolution kernel in the second stage is 1, 3, 3; the convolution kernels in the third to seventh stages are 3, 3, 3; The stride settings in each stage: the stride setting in the first stage is 1, 1, 1; the stride setting in the second stage is 1, 2, 2; the stride setting in the third stage is 1, 2, 2; the stride setting in the fourth stage is 2, 2, 2; the stride setting in the fifth stage is 2, 2, 2; the stride setting in the sixth stage is 1, 2, 2; the stride setting in the seventh stage is 1, 2, 2; There are 2 convolutional layers in each stage in the encoder and decoder of the segmentation model, that is, 2, 2, 2, 2, 2, 2, 2; the segmentation model does not use the Dropout layer.

5. The method for predicting split renal function according to claim 4, characterized in that, The loss function of the segmentation model is: Loss total = Loss dice + Loss CE Loss CE = -∑(Y log(P) + (1 - Y) log(1 - p)) Where Loss dice is the dice loss; Loss CE is the cross-entropy loss; Loss total is the total loss; I is the total number of pixels; K is the total number of samples; u is the Softmax probability output; v is the hard-coded true annotation; Y represents the true label; P represents the predicted probability of the model.

6. The method for predicting split renal function according to claim 1, wherein After inputting the corticomedullary enhancement phase image of the target patient into the trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image, it further includes: Compare the points in the probability map containing the cortical region segmented by the segmentation model with a preset first threshold respectively, and convert the probability map into a binary mask map according to the comparison results; Perform dilation and erosion operations on the binary mask map to remove the artifacts and noises in the binary mask map; Perform connected component analysis on the binary mask map after the dilation and erosion operations to remove the connected regions smaller than a preset second threshold in the binary mask map, and obtain the cortical region image.

7. The method for predicting split renal function according to claim 1, wherein Before inputting the radiomic features of the cortical region image into the trained renal function staging prediction model, the following steps are also included: Obtain cortical region images of the kidneys of a number of patients; Extract the radiomic features of the cortical region images, and construct a second data set with the radiomic features of a group of the cortical region images as a sample; Train the renal function staging prediction model based on the second data set to obtain the trained renal function staging prediction model.

8. A split renal function prediction device, characterized in that, It includes: An acquisition unit for acquiring the corticomedullary enhancement phase image of the kidney of a target patient; A segmentation unit for inputting the corticomedullary enhancement phase image of the kidney of the target patient into the trained segmentation model to segment the cortical region in the corticomedullary enhancement phase image to obtain the cortical region image of the kidney of the target patient; A feature extraction unit for extracting the radiomic features of the cortical region image; A prediction unit for inputting the radiomic features of the cortical region image into the trained renal function staging prediction model to predict the renal function stage of the target patient.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the split renal function prediction method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a processor to execute the split renal function prediction method according to any one of claims 1-7 when executed.

Images