Kidney separation magnetic resonance image segmentation method

Abstract

The invention relates to the technical field of magnetic resonance image processing, in particular to a kidney-dividing magnetic resonance image segmentation method, which comprises the following steps of: acquiring double-kidney magnetic resonance scanning data of a child patient suffering from hydronephrosis through a T2WI sequence of magnetic resonance plain scanning and a DWI sequence of diffusion weighted imaging; a kidney region image segmentation model framework is established by combining a U-net network and a Mask R-CNN network structure; and performing kidney region image segmentation on the double-kidney magnetic resonance scanning data through a kidney region image segmentation model framework, and determining a hydrops side kidney region and an opposite side kidney region. According to the method, self-adaption of hydrops side kidney detail enhancement is realized in an artificial intelligence algorithm, hydrops side kidney detail enhancement is carried out under the condition that the segmentation performance can be improved, hydrops side kidney detail enhancement does not need to be carried out under the condition that the segmentation performance cannot be improved, invalid image enhancement calculation overhead is reduced, and the image enhancement efficiency is improved. The self-adaptive enhancement can achieve the balance between the segmentation precision and the segmentation efficiency, and can achieve the efficient and accurate kidney region image segmentation.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic resonance image processing, and in particular to a kidney magnetic resonance image segmentation method. Background Art

[0002] Hydronephrosis refers to kidney damage caused by obstruction of urine flow from the kidneys to the bladder. It is a common urological disease in children, with the most common cause being ureteropelvic junction obstruction (UPJO) due to intrinsic abnormalities of the proximal ureter. Hydronephrosis is not a static process but a dynamic one, and the optimal timing of its treatment remains controversial, especially in pediatric patients. Assessment of renal function is an important factor in surgical treatment, and previous reports have suggested that children with poor preoperative renal function have a poor prognosis. Therefore, there is a need to identify an imaging biomarker that can noninvasively and accurately assess renal function in children with hydronephrosis.

[0003] Assessing renal function using nuclear renal imaging is considered the reference standard and may aid in treatment decisions. However, nuclear renal imaging suffers from poor image quality and low soft tissue resolution. Furthermore, the use of radiotracers and ionizing radiation precludes frequent follow-up. Studies have reported moderate to excellent correlation between renal volume and renal function assessed using contrast-enhanced CT (CE-CT), non-contrast-enhanced CT (NCE-CT), and contrast-enhanced MRI (CE-MRI). However, CT examinations use ionizing radiation, and gadolinium-based contrast agents may increase the risk of nephrogenic systemic fibrosis. Non-contrast-enhanced magnetic resonance urography (NCE-MRU) is widely used in the evaluation of pediatric patients with hydronephrosis due to its advantages of high resolution, lack of radiation exposure, lack of contrast agent, and short scan time. Magnetic resonance urography (MRU) has been widely and routinely used for imaging pediatric patients with hydronephrosis, ensuring high soft tissue resolution and contrast without the use of contrast agents. In the paper "Non-contrast enhanced magnetic resonance urography for assessing split kidney function in pediatric patients with hydronephrosis: comparison with renal scintigraphy," our research team demonstrated a high correlation between split kidney volume and split kidney function, avoiding the use of radionuclides and ionizing radiation. Therefore, split kidney function can be assessed through magnetic resonance imaging analysis.

[0004] However, although the existing technology can determine renal morphological characteristics, diffusion-weighted characteristics and other features through magnetic resonance image analysis to evaluate renal function in hydronephrosis, there is a lack of efficient and accurate kidney region image segmentation methods in magnetic resonance image analysis, and insufficient capture of details of the affected kidney, which affects the accuracy and efficiency of renal feature extraction, and ultimately affects the efficiency and accuracy of renal function assessment. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for renal magnetic resonance image segmentation to solve the technical problem that the existing technology lacks an efficient and accurate kidney region image segmentation method in magnetic resonance image analysis, which affects the accuracy and efficiency of renal feature extraction and ultimately affects the efficiency and accuracy of renal function assessment.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions: A method for segmenting a kidney magnetic resonance image comprises the following steps: MRI data of both kidneys of children with hydronephrosis were obtained using plain T2WI and diffusion-weighted imaging (DWI) sequences. A kidney region image segmentation model framework established by combining the U-net network and the Mask R-CNN network structure; The kidney region image segmentation model framework is used to perform kidney region image segmentation on the bilateral kidney magnetic resonance scan data to determine the hydronephrosis side kidney region and the contralateral kidney region.

[0007] As a preferred embodiment of the present invention, the method for constructing the kidney region image segmentation model framework includes: A hydronephrosis-side kidney enhancement module is constructed using a U-net network for image enhancement of the hydronephrosis-side kidney in bilateral renal magnetic resonance scan data. The Mask R-CNN network is used to construct the first hydronephrosis kidney and contralateral kidney segmentation module for segmenting the hydronephrosis kidney region and the contralateral kidney region in the enhanced bilateral renal MRI scan data of the hydronephrosis kidney; A second hydronephrosis and contralateral kidney segmentation module is constructed using the Mask R-CNN network to segment the hydronephrosis and contralateral kidney regions in bilateral renal MRI scan data. The enhancement trigger module of the hydronephrosis side kidney enhancement module is regulated by feedback of the comparison results of the segmentation performance between the first hydronephrosis side kidney and the contralateral kidney segmentation module and the second hydronephrosis side kidney and the contralateral kidney segmentation module.

[0008] As a preferred embodiment of the present invention, the hydronephrosis-side kidney enhancement module uses a U-net network to segment the hydronephrosis-side kidney region in the bilateral renal MRI scan data to obtain a hydronephrosis-side kidney mask image, and is used to perform pixel superposition and fusion of the hydronephrosis-side kidney mask image and the bilateral renal MRI scan data to obtain enhanced bilateral renal MRI scan data of the hydronephrosis-side kidney; ; Where, This is the enhanced MRI scan data of both kidneys on the hydronephrosis side. This is the MRI scan data of both kidneys. is the mask image of the kidney on the hydronephrosis side obtained by segmentation using the U-net network, where U-net is the U-net network. The enhancement trigger module generates feedback to regulate the enhancement coefficient of the hydronephrosis-side kidney of the enhancement module on the hydronephrosis side.

[0009] As a preferred embodiment of the present invention, the first hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region and the contralateral kidney region in the bilateral renal magnetic resonance scan data enhanced with the hydronephrosis-side kidney using a Mask R-CNN network, thereby obtaining a hydronephrosis-side kidney mask-enhanced image and a contralateral kidney mask image. The structural expression of the first hydronephrosis kidney and contralateral kidney segmentation module is: ; Where, is the contralateral kidney mask image, is the mask-enhanced image of the kidney on the hydronephrosis side, and Mask R-CNN is the Mask R-CNN network.

[0010] As a preferred embodiment of the present invention, the second hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region and the contralateral kidney region in the bilateral renal magnetic resonance scan data using a Mask R-CNN network to obtain a hydronephrosis-side kidney mask image and a contralateral kidney mask image; The structural expression of the second hydronephrosis kidney and contralateral kidney segmentation module is: ; Where, is the contralateral kidney mask image, is the mask image of the kidney on the hydronephrosis side, and Mask R-CNN is the Mask R-CNN network.

[0011] As a preferred embodiment of the present invention, the enhancement coefficient of the hydronephrosis-side kidney fed back by the enhancement trigger module consists of two parts: a start / stop parameter for controlling the start / stop of the hydronephrosis-side kidney enhancement module, and an adaptive intensity parameter for controlling the enhancement intensity of the hydronephrosis-side kidney by the hydronephrosis-side kidney enhancement module, wherein: The enhancement coefficient of the kidney on the hydrops side is: ; Where, is the enhancement coefficient of the kidney on the hydronephrosis side, is the start and stop parameter, is the adaptive strength parameter.

[0012] As a preferred solution of the present invention, the start-stop parameters are: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, The mask enhanced image of the kidney on the hydronephrosis side is is the mask image of the kidney on the hydronephrosis side, is the Kullback-Leibler divergence operation, is the pixel superposition operation formula.

[0013] As a preferred solution of the present invention, the adaptive strength parameter is set based on the segmentation performance of the kidney on the hydronephrosis side in the kidney enhancement module on the hydronephrosis side. for: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, This is the mask image of the kidney on the hydronephrosis side obtained by U-net network segmentation.

[0014] As a preferred embodiment of the present invention, the adaptive strength parameter is set based on the segmentation performance difference between the first hydronephrosis side kidney and contralateral kidney segmentation module and the second hydronephrosis side kidney and contralateral kidney segmentation module. for: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, is the mask image of the kidney on the hydronephrosis side obtained by U-net network segmentation, is the Kullback-Leibler divergence operation.

[0015] As a preferred solution of the present invention, the mask-enhanced image of the hydronephrosis-side kidney and the mask image of the contralateral kidney output by the first hydronephrosis-side kidney and contralateral kidney segmentation module are used as the output of the kidney region image segmentation model framework, and the output of the kidney region image segmentation model framework is used as the kidney region segmentation result of the bilateral kidney magnetic resonance scanning data.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention obtains imaging data of magnetic resonance imaging (MRI) plain scan and diffusion-weighted imaging (DWI) of children with hydronephrosis, uses an artificial intelligence algorithm to perform renal image feature analysis on the imaging data, and implements adaptive detail enhancement of the hydronephrosis-side kidney in the artificial intelligence algorithm. When the segmentation performance can be improved, the detail enhancement of the hydronephrosis-side kidney is performed. When the segmentation performance cannot be improved, the detail enhancement of the hydronephrosis-side kidney is not required, thereby reducing ineffective image enhancement computational overhead. Adaptive enhancement can achieve a balance between segmentation accuracy and segmentation efficiency, and can achieve efficient and accurate kidney region image segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0018] Figure 1 A flow chart of a method for segmenting renal magnetic resonance images provided by an embodiment of the present invention; Figure 2 A schematic diagram of a kidney region image segmentation model framework provided by an embodiment of the present invention; Figure 3 This is the kidney region image segmentation result diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, the present invention provides a method for segmenting a kidney magnetic resonance image, comprising the following steps: MRI data of both kidneys of children with hydronephrosis were obtained using plain T2WI and diffusion-weighted imaging (DWI) sequences. The imaging scanning parameters are as follows: T2WI scanning parameters: 3.0 T MR scanner, 8-channel cardiac coil, TR5300ms, TE 66.55ms, slice thickness 3mm, FA 110°, DWI parameters: TR / TE (ms), 4000 / shortest TE; matrix 128×96, bandwidth, 250 kHz; slice thickness 3mm, b value (s / mm2) 0,800.

[0021] A kidney region image segmentation model framework established by combining the U-net network and the Mask R-CNN network structure; The kidney region image segmentation model framework is used to perform kidney region image segmentation on the bilateral renal MRI scan data to determine the hydronephrosis side kidney region and the contralateral kidney region, such as Figure 3 shown.

[0022] In order to obtain the morphological characteristics and diffusion-weighted characteristics of the hydronephrosis-side and contralateral kidneys, the present invention first needs to perform image segmentation of the hydronephrosis-side and contralateral (or normal-side) kidneys on bilateral kidney magnetic resonance scan data or nuclear magnetic resonance scan images, that is, segment the hydronephrosis-side and contralateral (or normal-side) kidney regions on the nuclear magnetic resonance scan image, and then extract the renal morphological characteristics and diffusion-weighted characteristics based on the segmented hydronephrosis-side and contralateral kidney regions. Therefore, the present invention constructs a kidney region image segmentation model framework for the hydronephrosis-side and contralateral kidney regions.

[0023] The method for constructing the kidney region image segmentation model framework includes: A hydronephrosis-side kidney enhancement module is constructed using a U-net network for image enhancement of the hydronephrosis-side kidney in bilateral renal magnetic resonance scan data. The Mask R-CNN network is used to construct the first hydronephrosis kidney and contralateral kidney segmentation module for segmenting the hydronephrosis kidney region and the contralateral kidney region in the enhanced bilateral renal MRI scan data of the hydronephrosis kidney; A second hydronephrosis and contralateral kidney segmentation module is constructed using the Mask R-CNN network to segment the hydronephrosis and contralateral kidney regions in bilateral renal MRI scan data. The enhancement trigger module of the hydronephrosis side kidney enhancement module, i.e., the compare module, is regulated by feedback of the comparison results of the segmentation performance between the first hydronephrosis side kidney and the contralateral kidney segmentation module and the second hydronephrosis side kidney and the contralateral kidney segmentation module.

[0024] The U-Net is a convolutional neural network for image segmentation. It is U-shaped and consists of two main parts: an encoder (Contracting Path) and a decoder (Expanding Path). The encoder progressively extracts features through convolutional and pooling layers, while the decoder gradually restores the image's resolution through upsampling and convolution operations. A unique feature of the U-Net is its skip connections, which pass feature maps from the encoder directly to the decoder, helping to recover more precise details. This design enables the U-Net to better preserve detailed information when handling image segmentation tasks, thereby improving segmentation accuracy and precision.

[0025] Mask R-CNN (Mask Region-based Convolutional Neural Network) is an object detection and instance segmentation model extended from Faster R-CNN. It not only detects the location and class of objects in an image, but also generates pixel-level segmentation masks for each detected object, enabling more refined object segmentation.

[0026] like Figure 2 As shown, the present invention constructs a kidney region image segmentation model framework for realizing the segmentation of the hydronephrosis side kidney region and the contralateral kidney region, which is composed of a hydronephrosis side kidney enhancement module based on the U-net network, and two hydronephrosis side kidney and contralateral kidney segmentation modules based on the Mask R-CNN network. Among them, the hydronephrosis side kidney enhancement module is used to first use the U-net network to preliminarily segment the hydronephrosis side kidney region on the original bilateral kidney magnetic resonance scan data, and superimpose the segmentation mask result of the hydronephrosis side kidney region on the original bilateral kidney magnetic resonance scan data to form the hydronephrosis side kidney enhanced bilateral kidney magnetic resonance scan data, which can enhance the specific area (hydronephrosis side kidney area) in the original bilateral kidney magnetic resonance scan data while keeping other areas unchanged, thereby achieving the effect of highlighting the hydronephrosis side kidney area and enhancing its details.

[0027] Since the present invention intends to evaluate the renal function of children with hydronephrosis, the kidney on the hydronephrosis side is more concerned in the evaluation process, that is, the renal morphological characteristics and diffusion-weighted characteristics of the kidney on the hydronephrosis side play a dominant role, that is, the accuracy of the extraction of its renal morphological characteristics and diffusion-weighted characteristics is required to be higher. In order to meet the accuracy requirements of the renal morphological characteristics and diffusion-weighted characteristics of the hydronephrosis side, the unilateral enhancement of the kidney area on the hydronephrosis side is adopted, while the renal morphological characteristics and weighted diffusion characteristics of the corresponding normal kidney play an auxiliary role, so that the requirements for the normal kidney can be lowered, and no enhancement processing is required. It can meet the requirements of renal morphological characteristics and weighted diffusion characteristics extraction while reducing unnecessary calculations.

[0028] Then, the bilateral renal MRI scan data with enhanced details of the hydronephrosis-side kidney were re-passed through the Mask R-CNN network structure (i.e., the first hydronephrosis-side kidney and contralateral kidney segmentation module) to perform more accurate segmentation of the hydronephrosis-side kidney area, and a more accurate segmentation result of the hydronephrosis-side kidney area was obtained through the enhanced details.

[0029] A Mask R-CNN network structure (i.e., the second hydronephrosis-side kidney and contralateral kidney segmentation module) is also set up in the kidney region image segmentation model framework to perform hydronephrosis-side kidney region segmentation on the original bilateral kidney MRI scan data. It is equivalent to a quantification component of the hydronephrosis-side kidney enhancement effect, that is, monitoring the hydronephrosis-side kidney region segmentation performance obtained on the original bilateral kidney MRI scan data, and comparing it with the hydronephrosis-side kidney region segmentation performance obtained on the bilateral kidney MRI scan data with enhanced details of the hydronephrosis-side kidney. When it is monitored that the former performance is lower than the latter, it will be fed back to the enhancement coefficient of the hydronephrosis-side kidney. On, that is, the start and stop parameters Trigger the start of the hydronephrosis side kidney enhancement module, so that the hydronephrosis side kidney enhancement module performs image enhancement on the hydronephrosis side kidney. That is, the segmentation of the first hydronephrosis side kidney and contralateral kidney segmentation module is based on the enhanced bilateral renal MRI scan data of the hydronephrosis side kidney. More enhanced detail information is extracted.

[0030] When the performance of the former is higher than or equal to the latter, it means that the enhancement of the kidney on the hydrops side will not bring about a positive effect of improving the segmentation performance, and will also be fed back to the enhancement coefficient of the kidney on the hydrops side. Up, that is, The enhancement module of the hydronephrosis side kidney is turned off, and the image enhancement of the hydronephrosis side kidney is not performed. Finally, the segmentation of the first hydronephrosis side kidney and the contralateral kidney segmentation module is based on the bilateral renal MRI scan data. The above is based on the original bilateral kidney MRI scan data. At this time, the regional enhancement operation is redundant and does not need to be performed. At the same time, the efficiency is improved while the accuracy is achieved. Therefore, the start and stop operation of the kidney enhancement module on the hydronephrosis side is controlled by the start and stop parameters to balance the image segmentation efficiency and accuracy.

[0031] At the same time, the enhancement coefficient of the kidney on the hydronephrosis side Adaptive strength parameters are also integrated , adaptive strength parameter There are two setting methods. The first is to use the initial segmentation performance of the kidney on the hydronephrosis side. Related, The lower the value, the better the initial segmentation performance of the kidney on the hydrops side. The higher the reliability of the initial segmentation result of the kidney on the hydrops side superimposed on the original bilateral renal MRI scan data should be. The higher the superposition degree is set, the better the detail enhancement effect will be. Adaptive Strength Parameters The integration of It is meaningful only when it is set , ,when The lower, The higher the value, the more reliable the superposition. Finally, the segmentation of the first hydronephrosis kidney and the contralateral kidney segmentation module is based on the enhanced bilateral renal MRI scan data of the hydronephrosis kidney. , adding redundant enhancement operations to reduce efficiency in exchange for improved accuracy.

[0032] The second is the difference in segmentation performance between the first hydronephrosis and contralateral kidney segmentation module and the second hydronephrosis and contralateral kidney segmentation module due to the adaptive strength parameter Only It is meaningful only when The condition has practical significance, indicating that the first hydronephrosis side kidney and contralateral kidney segmentation module requires the enhancement operation of the hydronephrosis side kidney enhancement module. Depending on the enhancement operation, the greater the difference in segmentation performance (equivalent to the difference in segmentation results between the first hydronephrosis side kidney and contralateral kidney segmentation module and the second hydronephrosis side kidney and contralateral kidney segmentation module) The larger the value is, the more it relies on the enhancement operation, and the more it is expected to catch up with the segmentation performance of the second hydronephrosis kidney and the contralateral kidney segmentation module. Therefore, the credibility of the hydronephrosis kidney preliminary segmentation result superimposed on the original bilateral renal MRI scan data should be higher, and the higher the superposition degree is, the higher the setting is. , ,when The higher, The higher the value, the more reliable the superposition. Finally, the segmentation of the first hydronephrosis kidney and the contralateral kidney segmentation module is based on the enhanced bilateral renal MRI scan data of the hydronephrosis kidney. , adding redundant enhancement operations to reduce efficiency in exchange for improved accuracy.

[0033] In summary, the present invention can achieve adaptive enhancement of the details of the kidney on the hydrops side. When the segmentation performance can be improved, the details of the kidney on the hydrops side are enhanced. When the segmentation performance cannot be improved, there is no need to enhance the details of the kidney on the hydrops side, thereby reducing the ineffective image enhancement calculation overhead. Adaptive enhancement can achieve a balance between segmentation accuracy and segmentation efficiency, thereby achieving efficient and accurate kidney area image segmentation.

[0034] The hydronephrosis-side kidney enhancement module uses a U-net network to segment the hydronephrosis-side kidney region in the bilateral renal MRI scan data to obtain a mask image of the hydronephrosis-side kidney, and is used to perform pixel superposition and fusion of the hydronephrosis-side kidney mask image and the bilateral renal MRI scan data to obtain enhanced bilateral renal MRI scan data of the hydronephrosis-side kidney. ; Where, This is the enhanced MRI scan data of both kidneys on the hydronephrosis side. This is the MRI scan data of both kidneys. is the mask image of the kidney on the hydronephrosis side obtained by segmentation using the U-net network, where U-net is the U-net network. The enhancement trigger module generates feedback to regulate the enhancement coefficient of the hydronephrosis-side kidney of the enhancement module on the hydronephrosis side.

[0035] The first hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region and the contralateral kidney region in the enhanced bilateral renal MRI scan data using the Mask R-CNN network, thereby obtaining a hydronephrosis-side kidney mask-enhanced image and a contralateral kidney mask image. The structural expression of the first hydronephrosis kidney and contralateral kidney segmentation module is: ; Where, is the contralateral kidney mask image, is the mask-enhanced image of the kidney on the hydronephrosis side, and Mask R-CNN is the Mask R-CNN network.

[0036] The second hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region and the contralateral kidney region in the bilateral renal MRI scan data using the Mask R-CNN network to obtain the hydronephrosis-side kidney mask image and the contralateral kidney mask image; The structural expression of the second hydronephrosis kidney and contralateral kidney segmentation module is: ; Where, is the contralateral kidney mask image, is the mask image of the kidney on the hydronephrosis side, and Mask R-CNN is the Mask R-CNN network.

[0037] The enhancement coefficient of the hydronephrosis-side kidney fed back by the enhancement trigger module consists of two parts: the start and stop parameters that control the start and stop of the hydronephrosis-side kidney enhancement module, and the adaptive intensity parameter that controls the enhancement intensity of the hydronephrosis-side kidney by the hydronephrosis-side kidney enhancement module. The enhancement coefficient of the kidney on the hydronephrosis side is: ; Where, is the enhancement coefficient of the kidney on the hydronephrosis side, is the start and stop parameter, is the adaptive strength parameter.

[0038] The start and stop parameters are: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, The mask enhanced image of the kidney on the hydronephrosis side is shown in Figure 2. is the mask image of the kidney on the hydronephrosis side, is the Kullback-Leibler divergence operation, is the pixel superposition operation formula.

[0039] The second hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region on the original bilateral renal MRI scan data. It is equivalent to a quantification component for the enhancement effect of the hydronephrosis-side kidney. That is, it monitors the hydronephrosis-side kidney region segmentation performance obtained on the original bilateral renal MRI scan data and compares it with the hydronephrosis-side kidney region segmentation performance obtained on the bilateral renal MRI scan data with enhanced details of the hydronephrosis-side kidney. When it is monitored that the former performance is lower than the latter, it will be fed back to the enhancement coefficient of the hydronephrosis-side kidney. On, that is, the start and stop parameters Trigger the start of the hydronephrosis side kidney enhancement module, so that the hydronephrosis side kidney enhancement module performs image enhancement on the hydronephrosis side kidney. That is, the segmentation of the first hydronephrosis side kidney and contralateral kidney segmentation module is based on the enhanced bilateral renal MRI scan data of the hydronephrosis side kidney. More enhanced detail information is extracted.

[0040] When the performance of the former is higher than or equal to the latter, it means that the enhancement of the kidney on the hydrops side will not bring about a positive effect of improving the segmentation performance, and will also be fed back to the enhancement coefficient of the kidney on the hydrops side. Up, that is, The enhancement module of the hydronephrosis side kidney is turned off, and the image enhancement of the hydronephrosis side kidney is not performed. Finally, the segmentation of the first hydronephrosis side kidney and the contralateral kidney segmentation module is based on the bilateral renal MRI scan data. The above is based on the original bilateral kidney MRI scan data. At this time, the regional enhancement operation is redundant and does not need to be performed. At the same time, the efficiency is improved while the accuracy is achieved. Therefore, the start and stop operation of the kidney enhancement module on the hydronephrosis side is controlled by the start and stop parameters to balance the image segmentation efficiency and accuracy.

[0041] Setting adaptive intensity parameters based on the segmentation performance of the hydronephrosis kidney in the hydronephrosis kidney enhancement module for: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, This is the mask image of the kidney on the hydronephrosis side obtained by U-net network segmentation.

[0042] Enhancement coefficient of the kidney on the hydronephrosis side Adaptive strength parameters are also integrated , adaptive strength parameter There are two setting methods. The first is to use the initial segmentation performance of the kidney on the hydronephrosis side. Related, The lower the value, the better the initial segmentation performance of the kidney on the hydrops side. The higher the reliability of the initial segmentation result of the kidney on the hydrops side superimposed on the original bilateral renal MRI scan data should be. The higher the superposition degree is set, the better the detail enhancement effect will be. Adaptive Strength Parameters The integration of It is meaningful only when it is set , ,when The lower, The higher the value, the more reliable the superposition. Finally, the segmentation of the first hydronephrosis kidney and the contralateral kidney segmentation module is based on the enhanced bilateral renal MRI scan data of the hydronephrosis kidney. , adding redundant enhancement operations to reduce efficiency in exchange for improved accuracy.

[0043] Adaptive strength parameters are set based on the difference in segmentation performance between the first hydronephrosis-side and contralateral kidney segmentation module and the second hydronephrosis-side and contralateral kidney segmentation module. for: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, is the mask image of the kidney on the hydronephrosis side obtained by U-net network segmentation, is the Kullback-Leibler divergence operation.

[0044] The second is the difference in segmentation performance between the first hydronephrosis and contralateral kidney segmentation module and the second hydronephrosis and contralateral kidney segmentation module due to the adaptive strength parameter Only It only has practical significance when The condition has practical significance, indicating that the first hydronephrosis side kidney and contralateral kidney segmentation module requires the enhancement operation of the hydronephrosis side kidney enhancement module. Depending on the enhancement operation, the greater the difference in segmentation performance (equivalent to the difference in segmentation results between the first hydronephrosis side kidney and contralateral kidney segmentation module and the second hydronephrosis side kidney and contralateral kidney segmentation module) The larger the value is, the more it relies on the enhancement operation, and the more it is expected to catch up with the segmentation performance of the second hydronephrosis kidney and the contralateral kidney segmentation module. Therefore, the credibility of the hydronephrosis kidney preliminary segmentation result superimposed on the original bilateral renal MRI scan data should be higher, and the higher the superposition degree is, the higher the setting is. , ,when The higher, The higher the value, the more reliable the superposition. Finally, the segmentation of the first hydronephrosis kidney and the contralateral kidney segmentation module is based on the enhanced bilateral renal MRI scan data of the hydronephrosis kidney. , adding redundant enhancement operations to reduce efficiency in exchange for improved accuracy.

[0045] The mask-enhanced image of the hydronephrosis-side kidney and the mask image of the contralateral kidney output by the first hydronephrosis-side kidney and contralateral kidney segmentation module are used as the output of the kidney region image segmentation model framework, and the output of the kidney region image segmentation model framework is used as the kidney region segmentation result of the bilateral kidney magnetic resonance scan data.

[0046] The present invention obtains imaging data of magnetic resonance imaging (MRI) plain scan and diffusion-weighted imaging (DWI) of children with hydronephrosis, uses an artificial intelligence algorithm to perform renal image feature analysis on the imaging data, and implements adaptive detail enhancement of the hydronephrosis-side kidney in the artificial intelligence algorithm. When the segmentation performance can be improved, the detail enhancement of the hydronephrosis-side kidney is performed. When the segmentation performance cannot be improved, the detail enhancement of the hydronephrosis-side kidney is not required, thereby reducing ineffective image enhancement computational overhead. Adaptive enhancement can achieve a balance between segmentation accuracy and segmentation efficiency, and can achieve efficient and accurate kidney region image segmentation.

[0047] The above embodiments are merely exemplary embodiments of the present application and are not intended to limit the scope of the present application. The scope of protection of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and scope of protection of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present application.

Claims

1. A method for segmenting a kidney magnetic resonance image, characterized in that: The following steps are involved: MRI data of both kidneys of children with hydronephrosis were obtained using plain T2WI and diffusion-weighted imaging (DWI) sequences. A kidney region image segmentation model framework established by combining the U-net network and the Mask R-CNN network structure; The kidney region image segmentation model framework is used to perform kidney region image segmentation on the bilateral kidney magnetic resonance scan data to determine the hydronephrosis side kidney region and the contralateral kidney region.

2. The method for segmenting a kidney magnetic resonance image according to claim 1, wherein: The method for constructing the kidney region image segmentation model framework includes: A hydronephrosis-side kidney enhancement module is constructed using a U-net network for image enhancement of the hydronephrosis-side kidney in bilateral renal magnetic resonance scan data. The Mask R-CNN network is used to construct the first hydronephrosis kidney and contralateral kidney segmentation module for segmenting the hydronephrosis kidney region and the contralateral kidney region in the enhanced bilateral renal MRI scan data of the hydronephrosis kidney; A second hydronephrosis and contralateral kidney segmentation module is constructed using the Mask R-CNN network to segment the hydronephrosis and contralateral kidney regions in bilateral renal MRI scan data. The enhancement trigger module of the hydronephrosis side kidney enhancement module is regulated by feedback of the comparison results of the segmentation performance between the first hydronephrosis side kidney and the contralateral kidney segmentation module and the second hydronephrosis side kidney and the contralateral kidney segmentation module.

3. The method for segmenting a kidney magnetic resonance image according to claim 2, wherein: The hydronephrosis-side kidney enhancement module uses a U-net network to segment the hydronephrosis-side kidney region in the bilateral kidney MRI scan data to obtain a hydronephrosis-side kidney mask image, and is used to perform pixel superposition and fusion of the hydronephrosis-side kidney mask image and the bilateral kidney MRI scan data to obtain enhanced bilateral kidney MRI scan data of the hydronephrosis-side kidney; ； Where, This is the enhanced MRI scan data of both kidneys on the hydronephrosis side. This is the MRI scan data of both kidneys. is the mask image of the kidney on the hydronephrosis side obtained by segmentation using the U-net network, where U-net is the U-net network. The enhancement trigger module generates feedback to regulate the enhancement coefficient of the hydronephrosis-side kidney of the enhancement module on the hydronephrosis side.

4. The method for segmenting a kidney magnetic resonance image according to claim 3, wherein: The first hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region and the contralateral kidney region in the enhanced bilateral renal magnetic resonance scan data of the hydronephrosis-side kidney using a Mask R-CNN network to obtain a hydronephrosis-side kidney mask-enhanced image and a contralateral kidney mask image; The structural expression of the first hydronephrosis kidney and contralateral kidney segmentation module is: ； Where, is the contralateral kidney mask image, is the mask-enhanced image of the kidney on the hydronephrosis side, and Mask R-CNN is the Mask R-CNN network.

5. The method for segmenting a kidney magnetic resonance image according to claim 4, wherein: The second hydronephrosis-side kidney and contralateral kidney segmentation module is used to segment the hydronephrosis-side kidney region and the contralateral kidney region in the bilateral renal magnetic resonance scan data using a Mask R-CNN network to obtain a hydronephrosis-side kidney mask image and a contralateral kidney mask image; The structural expression of the second hydronephrosis kidney and contralateral kidney segmentation module is: ； Where, is the contralateral kidney mask image, is the mask image of the kidney on the hydronephrosis side, and Mask R-CNN is the Mask R-CNN network.

6. The method for segmenting a kidney magnetic resonance image according to claim 5, wherein: The enhancement coefficient of the hydronephrosis-side kidney fed back by the enhancement trigger module consists of two parts: a start / stop parameter for controlling the start / stop of the hydronephrosis-side kidney enhancement module, and an adaptive intensity parameter for controlling the enhancement intensity of the hydronephrosis-side kidney by the hydronephrosis-side kidney enhancement module, wherein: The enhancement coefficient of the kidney on the hydrops side is: ； Where, is the enhancement coefficient of the kidney on the hydronephrosis side, is the start and stop parameter, is the adaptive strength parameter.

7. The method for segmenting a kidney magnetic resonance image according to claim 6, wherein: The start and stop parameters are: ； Where GT is the true mask image of the kidney area on the hydronephrosis side, The mask enhanced image of the kidney on the hydronephrosis side is shown in Figure 2. is the mask image of the kidney on the hydronephrosis side, is the Kullback-Leibler divergence operation, is the pixel superposition operation formula.

8. The method for segmenting a kidney magnetic resonance image according to claim 7, wherein: The adaptive strength parameter is set based on the segmentation performance of the hydronephrosis side kidney in the hydronephrosis side kidney enhancement module. for: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, This is the mask image of the kidney on the hydronephrosis side obtained by U-net network segmentation.

9. The method for segmenting a kidney magnetic resonance image according to claim 7, wherein: The adaptive strength parameter is set based on the segmentation performance difference between the first hydronephrosis-side kidney and contralateral kidney segmentation module and the second hydronephrosis-side kidney and contralateral kidney segmentation module. for: ; Where GT is the true mask image of the kidney area on the hydronephrosis side, is the mask image of the kidney on the hydronephrosis side obtained by U-net network segmentation, is the Kullback-Leibler divergence operation.

10. The method for segmenting a kidney magnetic resonance image according to claim 2, wherein: The mask-enhanced image of the hydronephrosis-side kidney and the mask image of the contralateral kidney output by the first hydronephrosis-side kidney and contralateral kidney segmentation module are used as the output of the kidney region image segmentation model framework, and the output of the kidney region image segmentation model framework is used as the kidney region segmentation result of the bilateral kidney magnetic resonance scanning data.

Images