Kidney sclerosis proportion analysis method and system based on MRI image

Abstract

The invention discloses a kidney sclerosis proportion analysis method and system based on an MRI image, and belongs to the technical field of medical image processing, and the method comprises the steps: carrying out the MRI imaging sequence selection according to the kidney lesion characteristics of a patient; performing image acquisition on the kidney of the patient based on the MRI imaging sequence to obtain different MRI imaging sequence images; the MRI imaging sequence images are fused, and a multi-modal image is obtained; preprocessing the multi-modal image, wherein the preprocessing comprises pixel-by-pixel signal intensity correction, histogram equalization, adaptive filtering and non-local mean denoising; carrying out image segmentation on the preprocessed multi-modal image to obtain a binary image comprising a hardened region and a normal region; and obtaining the kidney sclerosis proportion based on the binary image. According to the method, the recognition accuracy and the quantification precision of the kidney sclerosis area can be improved, the kidney sclerosis degree can be evaluated more accurately, powerful support is provided for clinical research, and the medical level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for analyzing kidney sclerosis ratio based on MRI images. Background Art

[0002] Among the existing medical imaging technologies, MRI (Magnetic Resonance Imaging) is an important means to evaluate the structure and function of the kidney. However, traditional MRI image analysis technology has certain limitations in identifying areas of renal sclerosis. Current technology mainly relies on image analysis of a single MRI sequence, which often cannot accurately distinguish between normal tissue and sclerotic tissue. In addition, the limitations of image processing algorithms also lead to low accuracy in segmenting renal sclerosis areas, and the presence of artifacts and noise further increases the difficulty of analysis. Therefore, there is an urgent need for a renal sclerosis ratio analysis method to improve image contrast, reduce noise interference, and accurately segment and quantify renal sclerosis areas. Summary of the invention

[0003] The purpose of this application is to overcome the defects of the prior art and provide a method and system for analyzing the proportion of renal sclerosis based on MRI images.

[0004] In a first aspect, the present application provides a method for analyzing the proportion of renal sclerosis based on MRI images, comprising the following steps:

[0005] MRI imaging sequence selection is based on the patient's renal lesion characteristics;

[0006] Acquiring images of the patient's kidneys based on the selected MRI imaging sequence to obtain different MRI imaging sequence images;

[0007] The obtained MRI imaging sequence images are fused to obtain a fused multimodal image, including: performing standardization processing on each of the MRI imaging sequence images to obtain a standardized MRI imaging sequence image; performing spatial alignment on the standardized different MRI imaging sequence images; performing weighted fusion on the aligned different MRI imaging sequence images to obtain the multimodal image;

[0008] Preprocessing the multimodal image includes: performing pixel-by-pixel signal intensity correction on the multimodal image so that each pixel of the multimodal image has the same signal intensity level; performing histogram equalization on the corrected multimodal image to enhance the contrast of the image; performing adaptive filtering on the multimodal image after histogram equalization to remove noise and retain edge details; performing non-local mean denoising on the multimodal image after adaptive filtering to obtain a preprocessed multimodal image;

[0009] Performing image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area;

[0010] The renal sclerosis ratio is obtained based on the binary image.

[0011] Optionally, the selected MRI imaging sequence includes: T1-weighted imaging sequence, T2-weighted imaging sequence, diffusion-weighted imaging sequence and specific kidney imaging sequence; the obtained MRI imaging sequence images include: T1-weighted imaging sequence images, T2-weighted imaging sequence images, diffusion-weighted imaging sequence images and specific kidney imaging sequence images.

[0012] Optionally, performing pixel-by-pixel signal intensity correction on the multimodal image so that each pixel of the multimodal image has the same signal intensity level includes:

[0013] Perform multiple scans on the same patient or standard phantom to obtain the signal intensity relationship between different multimodal images and obtain the correction coefficient;

[0014] The multimodal image is adjusted pixel by pixel using the correction coefficient so that each pixel of the multimodal image has the same signal intensity level.

[0015] Optionally, the image segmentation is performed on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area, including:

[0016] Constructing a data set, wherein the data set includes a training set, a validation set, and a test set;

[0017] Constructing a deep learning model, and training the deep learning model using the training set;

[0018] Using a cross entropy loss function to optimize the deep learning model;

[0019] Using the validation set to evaluate the accuracy of the trained deep learning model, and using the test set to test the generalization ability of the model, to obtain the final deep learning model;

[0020] The preprocessed multimodal image is segmented using the deep learning model to obtain a binary image including a hardened area and a normal area.

[0021] Optionally, the deep learning model includes a U-Net model.

[0022] Optionally, the hardened area includes a white pixel area; and the normal area includes a black pixel area.

[0023] Optionally, obtaining the renal sclerosis ratio based on the binary image includes:

[0024] Obtaining the number of pixels of the hardened area based on the binary image;

[0025] Obtaining the number of pixels of the entire kidney based on the binary image;

[0026] The ratio of the number of pixels in the hardened area to the number of pixels in the entire kidney is taken as the kidney hardening ratio.

[0027] In a second aspect, the present application further provides a renal sclerosis ratio analysis system based on MRI images, which is used to perform the renal sclerosis ratio analysis method based on MRI images as described in any one of the first aspects, including:

[0028] An MRI imaging sequence selection module is used to select an MRI imaging sequence according to the patient's renal lesion characteristics;

[0029] An image acquisition module, used for acquiring images of the patient's kidneys based on a selected MRI imaging sequence to obtain different MRI imaging sequence images;

[0030] A multimodal fusion module, used for fusing the obtained MRI imaging sequence images to obtain a fused multimodal image;

[0031] A preprocessing module, used for preprocessing the multimodal image;

[0032] An image segmentation module is used to perform image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area;

[0033] The sclerosis ratio calculation module is used to obtain the kidney sclerosis ratio based on the binary image.

[0034] The present application provides a method and system for analyzing the proportion of renal sclerosis based on MRI images. By selecting a suitable MRI imaging sequence according to the characteristics of the patient's renal lesions, more comprehensive and accurate renal image information can be obtained; by fusing images of different imaging sequences to obtain multimodal images, the advantages of each sequence can be combined; pixel-by-pixel signal intensity correction, histogram equalization, adaptive filtering, and non-local mean denoising in the preprocessing process can improve image quality and provide a good foundation for subsequent analysis; using a deep learning model to segment the preprocessed image can accurately distinguish between sclerosis areas and normal areas. The method and system of the present application can more accurately assess the degree of renal sclerosis, provide strong support for clinical research, and improve medical standards.

[0035] In order to make the above features and advantages of the invention more obvious and easy to understand, embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 This is a flow chart of a method for analyzing the proportion of renal sclerosis based on MRI images provided in one embodiment of the present application.

[0038] Figure 2 This is a flowchart of step S50 in the method for analyzing the proportion of renal sclerosis based on MRI images provided in one embodiment of the present application.

[0039] Figure 3 This is a schematic diagram of the structure of a renal sclerosis ratio analysis system based on MRI images provided in another embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose and technical solution of the embodiment of the present application clearer, the technical solution of the embodiment of the present application will be clearly and completely described in conjunction with the drawings of the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, not all of the embodiments. Based on the described embodiment of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0041] In one embodiment, see Figure 1 The present application provides a method for analyzing the proportion of renal sclerosis based on MRI images. The method for analyzing the proportion of renal sclerosis based on MRI images may include the following steps: S10~S60.

[0042] S10: Select MRI imaging sequences based on the patient's renal lesion characteristics.

[0043] S20: Capturing images of the patient's kidneys based on the selected MRI imaging sequence to obtain different MRI imaging sequence images.

[0044] S30: fusing the obtained MRI imaging sequence images to obtain a fused multimodal image, including: performing standardization processing on each of the MRI imaging sequence images to obtain a standardized MRI imaging sequence image; performing spatial alignment on the standardized different MRI imaging sequence images; and performing weighted fusion on the aligned different MRI imaging sequence images to obtain the multimodal image.

[0045] S40: Preprocessing the multimodal image, including: performing pixel-by-pixel signal intensity correction on the multimodal image so that each pixel of the multimodal image has the same signal intensity level; performing histogram equalization on the corrected multimodal image to enhance the contrast of the image; performing adaptive filtering on the histogram equalized multimodal image to remove noise and retain edge details; performing non-local mean denoising on the adaptively filtered multimodal image to obtain a preprocessed multimodal image.

[0046] S50: performing image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area.

[0047] S60: Obtaining a renal sclerosis ratio based on the binary image.

[0048] In the MRI image-based renal sclerosis ratio analysis method of the present application, by selecting MRI imaging sequences according to the characteristics of the patient's renal lesions, highly targeted images can be obtained; by fusing images of different imaging sequences into multimodal images, the advantageous information of each sequence can be integrated; preprocessing of multimodal images can improve image quality; image segmentation of the preprocessed multimodal images can accurately divide the sclerosis and normal areas; finally, the sclerosis ratio is obtained based on the binary image, which can accurately evaluate the degree of renal sclerosis and provide a reliable basis for clinical research.

[0049] In step S10, refer to Figure 1 In step S10, the MRI imaging sequence is selected according to the patient's renal lesion characteristics.

[0050] As an example, the MRI imaging sequence is selected based on the patient's kidney structure, function and lesion characteristics. The most appropriate MRI imaging sequence can be selected based on factors such as the patient's age, physical condition, the presence of other underlying diseases, and the type, location, suspected degree of sclerosis of the kidney lesion.

[0051] As an example, the selected MRI imaging sequence may include: T1-weighted imaging sequence, T2-weighted imaging sequence, and diffusion-weighted imaging sequence. These sequences can provide kidney images with different contrasts, which are helpful to clearly depict the anatomical structure of the kidney.

[0052] Specifically, the T1-weighted imaging sequence uses the difference in T1 relaxation time in tissue to generate contrast. T1 relaxation time refers to the time required for hydrogen nuclei in tissue to return from a high energy state to a low energy state. In images generated by T1-weighted imaging sequences, fat usually appears as a high signal, i.e., bright white, while water and liquid appear as a low signal, i.e., dark. T1-weighted imaging sequences can provide good soft tissue contrast and help distinguish normal tissue from sclerotic tissue.

[0053] Furthermore, the T2-weighted imaging sequence uses the difference in T2 relaxation time in tissue to generate contrast. T2 relaxation time refers to the time required for hydrogen nuclei in tissue to lose phase consistency at the same energy level. In the images generated by the T2-weighted imaging sequence, water and liquids usually appear as high signals, i.e. bright white, while fat and solid tissues appear as low signals, i.e. dark colors. The T2-weighted imaging sequence is more sensitive to the contrast between liquid and soft tissue, which helps to show the diseased areas of the kidney.

[0054] Furthermore, the diffusion-weighted imaging sequence is an imaging technique that detects the Brownian motion of water molecules in tissues. The diffusion-weighted imaging sequence generates images by measuring the degree of diffusion of water molecules after a gradient magnetic field is applied. In the images generated by the diffusion-weighted imaging sequence, areas where the diffusion of water molecules is restricted, such as tumors or infarcted tissues, will appear as high signals, i.e. bright white, while normal tissues appear as low signals, i.e. dark. The diffusion-weighted imaging sequence can be used for early detection of certain tumors. The diffusion-weighted imaging sequence can evaluate the microstructure of tissues by detecting the Brownian motion of water molecules, which is particularly useful for identifying hardened areas.

[0055] As an example, the MRI imaging sequence may also include: a specific kidney imaging sequence. The specific physiological characteristics of the kidney are optimized through the specific kidney imaging sequence to obtain more accurate information about the sclerosis area, further improving the specificity and sensitivity of the image. The specific physiological characteristics of the kidney may include: blood perfusion characteristics, metabolic function characteristics, tissue composition characteristics, functional reserve characteristics, etc.

[0056] As an example, the specific kidney imaging sequence may include: kidney filter imaging or kidney diffusion imaging. By selecting different MRI imaging sequences, the sclerotic area of ​​the kidney can be better highlighted, providing high-quality raw data for subsequent image analysis and quantification.

[0057] In step S20, refer to Figure 1 In step S20, images of the patient's kidneys are collected based on the selected MRI imaging sequence to obtain different MRI imaging sequence images.

[0058] As an example, when a doctor starts an MRI device, he or she may select one or more of the MRI imaging sequences, including: a T1-weighted imaging sequence, a T2-weighted imaging sequence, a diffusion-weighted imaging sequence, and a specific kidney imaging sequence, and adjust relevant parameters of the MRI device according to different MRI imaging sequences to perform image acquisition to obtain different MRI imaging sequence images.

[0059] As an example, the MRI imaging sequence images may include: T1-weighted imaging sequence images, T2-weighted imaging sequence images, diffusion-weighted imaging sequence images, and specific kidney imaging sequence images.

[0060] As an example, for T1-weighted imaging sequences, shorter repetition time TR and echo time TE can be selected. The repetition time TR can be set between 300-800 milliseconds and the echo time TE can be set between 10-30 milliseconds to highlight the T1 relaxation characteristics of the tissue and display the anatomical structure of the kidney. For T2-weighted imaging sequences, longer repetition time TR and echo time TE can be used. The repetition time TR can be set between 2000-5000 milliseconds and the echo time TE can be set between 80-120 milliseconds to display the T2 relaxation characteristics of the tissue and changes in the water content of the kidney, which is helpful for discovering kidney lesions. For diffusion-weighted imaging sequences, different diffusion sensitivity coefficient b values ​​can be set. The b value can be set between 0-1000 s / mm² to detect the diffusion movement of water molecules, which is helpful for microstructural changes in kidney tissue.

[0061] As an example, the acquired MRI imaging sequence images can be converted into a digital matrix, and each element in the digital matrix represents the signal intensity of the corresponding pixel in the MRI imaging sequence images. When the MRI imaging sequence images are acquired by the MRI device, the MRI imaging sequence images are usually in the form of pixels, and each acquired MRI imaging sequence image can be converted into a digital matrix, wherein each pixel of the MRI imaging sequence image corresponds to an element in the digital matrix.

[0062] In one example, a doctor collects a black-and-white image through an MRI device. Its size is 2×2 pixels. This black-and-white image is converted into a 2×2 digital matrix. The digital matrix is ​​represented as follows:

[0063]

[0064] Wherein, P1 represents the signal strength of the pixel in the upper left corner of the image, P2 represents the signal strength of the pixel in the upper right corner, P3 represents the signal strength of the pixel in the lower left corner, and P4 represents the signal strength of the pixel in the lower right corner. The digital matrix of the black and white image can be .

[0065] As an example, for a digital matrix of a black and white image, each element represents the signal strength of the corresponding pixel in the black and white image, and the signal strength is usually related to the brightness. The digital matrix of a color image is more complicated because each pixel of the color image includes three colors: red, green and blue. Each element of the digital matrix of the color image represents the signal strength of the three colors of red, green and blue of the corresponding pixel.

[0066] In step S30, refer to Figure 1 In step S30, the obtained MRI imaging sequence images are fused to obtain a fused multimodal image, including: standardizing each of the MRI imaging sequence images to obtain a standardized MRI imaging sequence image; spatially aligning the standardized different MRI imaging sequence images; and weighted fusion of the aligned different MRI imaging sequence images to obtain the multimodal image.

[0067] Specifically, firstly, each individual MRI imaging sequence image is subjected to standardization processing to obtain a standardized MRI imaging sequence image, so as to ensure the consistency of the MRI imaging sequence image in terms of spatial resolution and contrast.

[0068] As an example, the brightness and contrast of the images can be adjusted to make the images in different sequences visually consistent.

[0069] Furthermore, the standardized images of different MRI imaging sequences are spatially aligned using an image registration algorithm to facilitate subsequent fusion processing.

[0070] As an example, the image registration algorithm may include: a registration method based on mutual information. Specifically, the transformation parameters between different MRI imaging sequence images are calculated to achieve optimal alignment.

[0071] Furthermore, the aligned images of different MRI imaging sequences are weighted fused to obtain a multimodal image containing more information.

[0072] As an example, linear or nonlinear fusion technology can be used for weighted fusion to combine information from different MRI imaging sequence images to obtain a multimodal image that retains complementary information from each MRI imaging sequence image, making the hardened area more prominent and easier to identify.

[0073] In another example, a set of T1-weighted imaging sequence images and T2-weighted imaging sequence images are collected and fused to obtain a multimodal image, which can clearly identify the hardened area while showing the kidney structure, providing doctors with more accurate information. Fusion can improve the contrast and clarity of MRI imaging sequence images, thereby improving the recognition accuracy of kidney hardened areas, which can not only improve image quality, but also lay a solid foundation for subsequent image analysis and calculation of hardening ratio.

[0074] In step S40, refer to Figure 1In step S40, the multimodal image is preprocessed, including: performing pixel-by-pixel signal intensity correction on the multimodal image so that each pixel of the multimodal image has the same signal intensity level; performing histogram equalization on the corrected multimodal image to enhance the contrast of the image; performing adaptive filtering on the multimodal image after histogram equalization to remove noise and retain edge details; performing non-local mean denoising on the multimodal image after adaptive filtering to obtain a preprocessed multimodal image.

[0075] Specifically, the multimodal image is firstly subjected to pixel-by-pixel signal intensity correction to eliminate the image intensity deviation caused by inconsistent equipment or scanning parameters, so that each pixel of the multimodal image has the same signal intensity level. Specifically, the signal intensity relationship between different MRI imaging sequence images is obtained by performing multiple scans on the same patient or the same standard phantom, and a correction coefficient is obtained. The multimodal image is adjusted pixel by pixel using the correction coefficient so that each pixel of the multimodal image has the same signal intensity level.

[0076] Furthermore, advanced preprocessing techniques are performed on the corrected multimodal images to further improve the image quality.

[0077] As an example, the advanced preprocessing technology may include: histogram equalization, adaptive filtering. Specifically, histogram equalization is performed on the corrected multimodal image to enhance the contrast of the image; and adaptive filtering is performed on the multimodal image after histogram equalization to remove noise and retain edge details.

[0078] As an example, the advanced preprocessing technology may also include: non-local mean denoising. Specifically, non-local mean denoising is performed on the multimodal image after adaptive filtering to obtain a preprocessed multimodal image.

[0079] As an example, the histogram equalization changes the histogram distribution of the corrected multimodal image so that the brightness distribution of the corrected image is more uniform, thereby enhancing the contrast of the image and making the difference between the hardened area and the normal tissue more obvious.

[0080] As an example, the adaptive filtering adaptively adjusts the parameters of the filter according to the local features of the image to optimize the denoising effect, and can remove noise while retaining edges and details.

[0081] As an example, the non-local mean denoising takes into account repeated texture information in the image and performs weighted averaging on similar pixels in the image, thereby effectively reducing noise and protecting the structure of the image. It can further remove noise in the image and lay a solid foundation for subsequent image segmentation and analysis.

[0082] In another example, for a T1 weighted imaging sequence image after multimodal fusion, multiple scans are performed using signal intensity correction to obtain a correction coefficient, and the T1 weighted imaging sequence image is adjusted pixel by pixel based on the correction coefficient to ensure that different sequence images are compared at the same intensity level to achieve a consistent signal intensity level; the corrected T1 weighted imaging sequence image is further changed by histogram equalization to make the difference between the hardened area and the normal area more obvious; the T1 weighted imaging sequence image after histogram equalization is adaptively filtered to adaptively adjust the filter parameters according to the local features of the image to remove noise and retain edges and details; the T1 weighted imaging sequence image after adaptive filtering is subjected to non-local mean denoising, and similar pixels in the T1 weighted imaging sequence image are weighted averaged to further remove noise in the image to obtain the preprocessed T1 weighted imaging sequence image. After the above preprocessing operation, the contrast and clarity of the T1 weighted imaging sequence image can be significantly improved, and high-quality, high-signal-to-noise ratio images can be obtained, so that the subsequent image segmentation can more accurately identify and segment the hardened area of ​​the kidney, thereby calculating the hardening ratio.

[0083] In step S50, refer to Figure 1 In step S50, the pre-processed multimodal image is segmented to obtain a binary image including a hardened area and a normal area.

[0084] As an example, see Figure 2 , step S50 may include the following steps: S501~S505.

[0085] S501: Construct a data set, where the data set includes a training set, a validation set, and a test set.

[0086] S502: Construct a deep learning model, and train the deep learning model using the training set.

[0087] S503: Optimize the deep learning model using a cross entropy loss function.

[0088] S504: Use the validation set to evaluate the accuracy of the trained deep learning model, and use the test set to test the generalization ability of the model to obtain the final deep learning model.

[0089] S505: Use the deep learning model to perform image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area.

[0090] As an example, in step 501, the data set is constructed based on the historical kidney data acquired in advance, and the data set may include but is not limited to 1000 historical kidney images, each of which has a corresponding sclerosis area annotation that has been annotated in advance. The data set may include a training set, a validation set, and a test set.

[0091] Further, in step 502, a deep learning model is constructed, and the deep learning model may include an architecture based on a convolutional neural network.

[0092] As an example, a deep learning model based on U-Net is constructed. The U-Net deep learning model has symmetrical contraction paths and expansion paths, which can extract and fuse features at different levels. The jump connection in the model can combine high-resolution feature maps with upsampled feature maps, thereby retaining more detail information during segmentation.

[0093] As an example, the U-Net deep learning model may include: feature extraction, downsampling, upsampling, and jump connections. Specifically, the preprocessed multimodal image obtains useful features in the image through feature extraction; further extracts features through downsampling (i.e., contraction path) to obtain low-resolution, high-pass inverse feature maps, extracts low-resolution feature maps of different scales through multiple downsampling, and doubles the number of channels after each downsampling; restores the low-resolution feature map to a high-resolution feature map through upsampling (i.e., expansion path), restores low-resolution feature maps of different scales through multiple upsampling, and halves the number of channels after each upsampling, doubling the size of the feature map; uses jump connections to splice the low-resolution feature map of the corresponding level in the downsampling with the high-resolution feature map after upsampling, and then passes through two 3×3 convolutional layers for feature fusion and processing after splicing, so that the image resolution is gradually restored, and the detailed information retained in the jump connection is used to achieve accurate segmentation of the kidney image.

[0094] Furthermore, the deep learning model is trained using the training set. During the training process, the deep learning model is optimized using a cross entropy loss function, and the weights of the deep learning model are continuously adjusted through forward propagation and back propagation to minimize the loss function.

[0095] As an example, the cross entropy loss function measures the difference between the predicted segmentation map and the true label, which can be expressed as follows:

[0096]

[0097] Where L represents the cross entropy loss function; N is the total number of pixels in the image; K is the number of categories, namely the hardened area and the non-hardened area; y ik is the true label of the i-th pixel belonging to category k; is the probability that the i-th pixel of the predicted segmentation map belongs to category k.

[0098] Furthermore, in step 504, the accuracy of the trained deep learning model is evaluated using the validation set, and the generalization ability of the deep learning model is tested using the test set. The parameters of the deep learning model are fine-tuned according to the test results, and the fine-tuned deep learning model is retrained to finally obtain a deep learning model with the best performance. The final deep learning model can accurately segment the sclerosis area from the kidney image, realize image segmentation, and provide a reliable basis for the final calculation of the sclerosis ratio.

[0099] Furthermore, in step 505, the final deep learning model is used to perform image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area. The binary image can accurately identify the hardened boundary; the hardened area may include a white pixel area, and the normal area may include a black pixel area.

[0100] As an example, the binary image obtained can be further used for feature extraction and analysis to provide doctors with quantitative information. The renal sclerosis ratio can also be obtained by calculating the percentage of the sclerosis area to the entire kidney area, which helps monitor the progression of the disease.

[0101] In step S60, refer to Figure 1 In step S60, the kidney sclerosis ratio is obtained based on the binary image.

[0102] As an example, all pixels marked as sclerosis in the binary image are counted to obtain the number of pixels in all sclerosis areas, that is, the number of white pixels in the binary image. Then, based on the binary image, the number of pixels in the entire kidney is obtained, and the entire kidney includes the sclerosis area and the normal area, and the total number of all white pixels and black pixels in the binary image is obtained. Finally, based on the number of pixels in the sclerosis area and the number of pixels in the entire kidney, the kidney sclerosis ratio is obtained. The formula for the kidney sclerosis ratio is as follows:

[0103]

[0104] Among them, the renal sclerosis ratio is a dimensionless value, which directly reflects the proportion of the sclerosis area to the entire kidney.

[0105] As an example, image processing software or programming languages, such as Python with the OpenCV library, can be used to perform pixel counting.

[0106] In one example, a binary image is obtained through a deep learning model, and the binary image shows that the sclerosis area has 5000 pixels, while the entire kidney area has 20000 pixels, then the kidney sclerosis ratio can be obtained as , that is 25%.

[0107] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0108] In the kidney sclerosis ratio analysis method based on MRI images of the present application, different MRI imaging sequences are selected in combination with the characteristics of the patient's kidney lesions, and MRI imaging sequence images are obtained. The different imaging principles of each MRI imaging sequence are used to comprehensively present the kidney structure and lesions; multiple MRI imaging sequence images are fused to obtain multimodal images, which can integrate advantageous information and highlight the hardened area; the intensity deviation is eliminated by pixel-by-pixel correction, and then the image quality is further improved through histogram equalization, adaptive filtering and non-local mean denoising; a deep learning model is constructed based on U-Net for image segmentation, and the hardened and normal areas are accurately divided using contraction, expansion paths and jump connections; finally, the kidney sclerosis ratio is calculated based on the binary image. The present application method can improve the recognition accuracy and quantification accuracy of the hardened area, can accurately evaluate the degree of kidney sclerosis, and provide more reliable basis and quantitative information for clinical research.

[0109] In another embodiment, see Figure 3The present application also provides a kidney sclerosis ratio analysis system based on MRI images, which may include: an MRI imaging sequence selection module 10, an image acquisition module 20, a multimodal fusion module 30, a preprocessing module 40, an image segmentation module 50, and a hardening ratio calculation module 60. Among them, the MRI imaging sequence selection module 10 is used to select an MRI imaging sequence according to the characteristics of the patient's kidney lesions; the image acquisition module 20 is used to acquire images of the patient's kidneys based on the selected MRI imaging sequence to obtain different MRI imaging sequence images; the multimodal fusion module 30 is used to fuse the obtained MRI imaging sequence images to obtain a fused multimodal image; the preprocessing module 40 is used to preprocess the multimodal image; the image segmentation module 50 is used to segment the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area; the hardening ratio calculation module 60 is used to obtain the kidney hardening ratio based on the binary image.

[0110] In the above-mentioned kidney sclerosis ratio analysis system based on MRI images, the MRI imaging sequence selection module 10 accurately selects the imaging sequence according to the characteristics of the patient's kidney lesions, the image acquisition module 20 obtains multi-sequence images, the multimodal fusion module 30 integrates the image advantage information, the preprocessing module 40 further improves the image quality, the image segmentation module 50 accurately divides the sclerosis and normal areas, and the sclerosis ratio calculation module 60 obtains the sclerosis ratio to provide a more reliable basis for clinical research. The system of the present application can improve the recognition accuracy and quantification accuracy of the kidney sclerosis area, can provide a quantitative basis for the degree of kidney sclerosis, provide intuitive and reliable data for doctors, assist doctors to timely understand the patient's kidney lesions, and formulate scientific and reasonable response strategies.

[0111] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] Although the present application has been disclosed as above with the embodiments, it is not intended to limit the present application. Any person with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be determined by the scope of the attached patent application.

Claims

1. A method for analyzing the proportion of renal sclerosis based on MRI images, characterized in that: The following steps are involved: MRI imaging sequence selection is based on the patient's renal lesion characteristics; Acquiring images of the patient's kidneys based on the selected MRI imaging sequence to obtain different MRI imaging sequence images; The obtained MRI imaging sequence images are fused to obtain a fused multimodal image, including: performing standardization processing on each of the MRI imaging sequence images to obtain a standardized MRI imaging sequence image; performing spatial alignment on the standardized different MRI imaging sequence images; performing weighted fusion on the aligned different MRI imaging sequence images to obtain the multimodal image; Preprocessing the multimodal image includes: performing pixel-by-pixel signal intensity correction on the multimodal image so that each pixel of the multimodal image has the same signal intensity level; performing histogram equalization on the corrected multimodal image to enhance the contrast of the image; performing adaptive filtering on the multimodal image after histogram equalization to remove noise and retain edge details; performing non-local mean denoising on the multimodal image after adaptive filtering to obtain a preprocessed multimodal image; Performing image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area; The renal sclerosis ratio is obtained based on the binary image.

2. The method for analyzing the proportion of renal sclerosis based on MRI images according to claim 1, characterized in that: The selected MRI imaging sequences include: T1-weighted imaging sequence, T2-weighted imaging sequence, diffusion-weighted imaging sequence and specific kidney imaging sequence; the obtained MRI imaging sequence images include: T1-weighted imaging sequence images, T2-weighted imaging sequence images, diffusion-weighted imaging sequence images and specific kidney imaging sequence images.

3. The method for analyzing the proportion of renal sclerosis based on MRI images according to claim 1, characterized in that: The performing pixel-by-pixel signal intensity correction on the multimodal image so that each pixel of the multimodal image has the same signal intensity level includes: Perform multiple scans on the same patient or standard phantom to obtain the signal intensity relationship between different multimodal images and obtain the correction coefficient; The multimodal image is adjusted pixel by pixel using the correction coefficient so that each pixel of the multimodal image has the same signal intensity level.

4. The method for analyzing the proportion of renal sclerosis based on MRI images according to claim 1, characterized in that: The image segmentation of the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area includes: Constructing a data set, wherein the data set includes a training set, a validation set, and a test set; Constructing a deep learning model, and training the deep learning model using the training set; Using a cross entropy loss function to optimize the deep learning model; Using the validation set to evaluate the accuracy of the trained deep learning model, and using the test set to test the generalization ability of the model, to obtain the final deep learning model; The deep learning model is used to perform image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area.

5. The method for analyzing the proportion of renal sclerosis based on MRI images according to claim 4, characterized in that: The deep learning model includes a U-Net model.

6. The method for analyzing the proportion of renal sclerosis based on MRI images according to claim 4, characterized in that: The hardened area includes a white pixel area; and the normal area includes a black pixel area.

7. The method for analyzing the proportion of renal sclerosis based on MRI images according to claim 1, characterized in that: The step of obtaining the renal sclerosis ratio based on the binary image comprises: Obtaining the number of pixels of the hardened area based on the binary image; Obtaining the number of pixels of the entire kidney based on the binary image; The ratio of the number of pixels in the hardened area to the number of pixels in the entire kidney is taken as the kidney hardening ratio.

8. A renal sclerosis ratio analysis system based on MRI images, characterized in that: Used to perform the method for analyzing the proportion of kidney sclerosis based on MRI images according to any one of claims 1 to 7; the system for analyzing the proportion of kidney sclerosis based on MRI images comprises: An MRI imaging sequence selection module is used to select an MRI imaging sequence according to the patient's renal lesion characteristics; An image acquisition module, used for acquiring images of the patient's kidneys based on a selected MRI imaging sequence to obtain different MRI imaging sequence images; A multimodal fusion module, used for fusing the obtained MRI imaging sequence images to obtain a fused multimodal image; A preprocessing module, used for preprocessing the multimodal image; An image segmentation module is used to perform image segmentation on the preprocessed multimodal image to obtain a binary image including a hardened area and a normal area; The sclerosis ratio calculation module is used to obtain the kidney sclerosis ratio based on the binary image.

Images