A pancreas segmentation method and system for CT images

Abstract

The present invention relates to the technical field of CT image analysis, and in particular to a pancreas segmentation method and system for CT images. The present invention first performs abdominal window linear transformation, liver window linear transformation, and high-contrast window linear transformation on a CT image, and then performs Fourier transformation to obtain spectrum information of each linear transformation image. Each linear transformation image is then partitioned to obtain low-frequency spectrum information of the abdominal window linear transformation image, mid-frequency spectrum information of the liver window linear transformation image, and high-frequency spectrum information of the high-contrast window linear transformation image. A dynamic weight function is then set according to each region. Finally, feature fusion is performed based on the dynamic weight function and the partitioned spectrum information of each transformation image, so that the fused image has both global structure and local details, retains the overall outline of the pancreas and liver, avoids anatomical structure disruption caused by narrow window width, and enhances blood vessels and boundaries, especially improving the contrast between the pancreas and surrounding tissues.

Description

Technical Field

[0001] The present invention relates to the technical field of CT image analysis, and in particular to a pancreas segmentation method and system for CT images. Background Art

[0002] Pancreatic cancer is a highly malignant tumor. Early diagnosis and treatment are crucial for improving patient survival. CT (computed tomography) imaging is one of the commonly used techniques in pancreatic cancer diagnosis. CT scans allow doctors to obtain detailed images of the pancreas and surrounding tissues. These images allow doctors to diagnose the presence of pancreatic cancer, assess the size and location of the tumor, and develop treatment plans.

[0003] However, the segmentation of pancreatic CT images faces many challenges: low contrast: the grayscale value difference between pancreatic tissue and surrounding tissues (such as liver, spleen, intestines, etc.) is small, resulting in unclear boundaries of the pancreatic region and difficulty in accurate segmentation; noise and artifacts: CT images may contain noise and artifacts. These interference factors will reduce image quality and affect the accuracy of segmentation.

[0004] To address these issues, segmentation methods based on image preprocessing have gained increasing attention in recent years. The goal of image preprocessing is to optimize image quality and enhance the features of regions of interest through a series of algorithms, thereby improving the performance of segmentation algorithms. However, traditional CT image preprocessing methods may fail to simultaneously preserve pancreatic soft tissue structure, vascular information, and lesion characteristics, resulting in low accuracy and limited visibility of segmented pancreatic CT images. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a pancreas segmentation method and system for CT images, which are used to solve the problems existing in the prior art.

[0006] According to one aspect of the present invention, a pancreas segmentation method for CT images is provided, comprising the following steps:

[0007] S1: Acquire a CT image containing the pancreas;

[0008] S2: performing a preprocessing operation on the CT image to obtain a preprocessed CT image;

[0009] The S2 is specifically:

[0010] S2.1: performing linear transformations of different window widths and window levels on the CT image to generate a plurality of transformed images, wherein the linear transformations of different window widths and window levels include an abdominal window linear transformation, a liver window linear transformation, and a high-contrast window linear transformation;

[0011] S2.2: Performing Fourier transform on the multiple transformed images to obtain frequency spectrum information of each transformed image;

[0012] S2.3: performing a partitioning operation on the spectrum information of each transformed image to obtain low-frequency spectrum information of the abdominal window linear transformation image, mid-frequency spectrum information of the liver window linear transformation image, and high-frequency spectrum information of the high-contrast window linear transformation image;

[0013] S2.4: Calculate the fusion weight function of the spectrum information of each transformed image;

[0014] S2.5: Performing spectrum fusion based on the fusion weight functions of the low-frequency region, the intermediate-frequency region, and the high-frequency region and the low-frequency spectrum information of the abdominal window linear transformation image, the intermediate-frequency spectrum information of the liver window linear transformation image, and the high-frequency spectrum information of the high-contrast window linear transformation image;

[0015] S2.6: Perform inverse Fourier transform on the fused spectrum information to obtain a fused image;

[0016] S3: Inputting the preprocessed CT image into a pancreatic image segmentation model to implement pancreatic segmentation.

[0017] Preferably, the S2.3 specifically comprises: taking the center of the spectrum information as the origin, and calculating the radius r of each frequency point (u, v) in the spectrum information;

[0018] ;

[0019] Among them, the low frequency area: ;Intermediate frequency area: ;

[0020] High frequency area: ; r max is the maximum radius from each frequency point to the origin;

[0021] Then, the abdominal window linear transformation image, liver window linear transformation image and high contrast window linear transformation image are partitioned respectively to obtain the low frequency area of ​​the abdominal window linear transformation image, the intermediate frequency area of ​​the liver window linear transformation image and the high frequency area of ​​the high contrast window linear transformation image.

[0022] Preferably, in S2.4, the fusion weight function of the low-frequency region of the abdominal window linear transformation image, the intermediate-frequency region of the liver window linear transformation image, and the high-frequency region of the high-contrast window linear transformation image is:

[0023] ;

[0024] Where, is the low-frequency weight function, is the mid-frequency weight function, is the high frequency area weight function; σ L and σ H is the control parameter.

[0025] Preferably, in S2.3, the linear transformations of different window widths and window levels are abdominal window linear transformation, liver window linear transformation, and high contrast window linear transformation.

[0026] Preferably, the linear transformation formula is:

[0027]

[0028] Where, I windowed is the grayscale value after linear transformation, I is the grayscale value before linear transformation, WL is the window level, and WW is the window width.

[0029] Preferably, the window width of the abdominal window linearly transformed is 350, and the window level is 40; the window width of the liver window linearly transformed is 150, and the window level is 30; and the window width of the high contrast window linearly transformed is 100, and the window level is 20.

[0030] Preferably, the window width and window level of the high contrast window are determined as follows:

[0031] Sa: dividing the CT image into multiple local image regions;

[0032] Sb: Calculate the mean grayscale value of each local image area;

[0033] Sc: screening out the local image region of interest according to the mean grayscale value of each local image region;

[0034] Sd: Calculate the average value μ1 of the grayscale value means of the local image region of interest and the average value μ of the grayscale value means of the multiple local image regions;

[0035] Se: Determine the window width and window level of the high contrast window according to the average value μ1 of the grayscale value means of the local image region of interest and the average value μ of the grayscale value means of the multiple local image regions.

[0036] Preferably, the window level WL of the high contrast window is:

[0037] ;

[0038] Where a is the coefficient;

[0039] Wherein, the window width WW of the high contrast window is:

[0040] ;

[0041] The calculation formula of the coefficient a is:

[0042] ;

[0043] Where, is the target contrast-to-noise ratio, It is the contrast-to-noise ratio between the target tissue and the background tissue obtained by preprocessing the CT image using the initial window width and initial window level.

[0044] Preferably, in S2.5, the image fusion formula is:

[0045]

[0046] Where, is the spectrum information of the fused image, is the low-frequency weight function,

[0047] is the mid-frequency weight function, is the high frequency area weight function; is the low-frequency spectrum information of the abdominal window linear transformation image; is the spectrum information of the intermediate frequency region of the liver window linear transformation image; It is the high-frequency spectrum information of the high-contrast window linearly transformed image.

[0048] According to another aspect of the present invention, a pancreas segmentation system for CT images is provided. The system adopts the above-mentioned pancreas segmentation method for CT images. The system comprises:

[0049] An image acquisition module, used for acquiring a CT image containing the pancreas;

[0050] A preprocessing module, configured to perform a preprocessing operation on the CT image to obtain a preprocessed CT image;

[0051] The image segmentation module is used to input the preprocessed CT image into a pancreas image segmentation model to implement pancreas segmentation.

[0052] The present invention has the following technical effects:

[0053] The present invention proposes a pancreatic CT image segmentation method. When preprocessing the CT image, the CT image is first subjected to abdominal window linear transformation, liver window linear transformation and high contrast window linear transformation. Then, the multiple linear transformation images are respectively subjected to Fourier transformation to obtain the spectrum information of each linear transformation image. Then, the spectrum information of each linear transformation image is partitioned to obtain the low-frequency spectrum information of the abdominal window linear transformation image, the intermediate-frequency spectrum information of the liver window linear transformation image and the high-frequency spectrum information of the high contrast window linear transformation image. Then, a dynamic weight function is set according to each zone. Finally, feature fusion is performed based on the dynamic weight function and the partitioned spectrum information of each transformation image. The introduction of the high contrast window and the feature fusion scheme of the dynamic weight function and the partitioned spectrum information of each transformation image in the present invention enable the fused image to have both global structure and local details, retain the overall contours of organs such as the pancreas and liver, avoid anatomical structure breaks caused by narrow window width, and enhance blood vessels and boundaries, especially improve the contrast between the pancreas and surrounding tissues.

[0054] At the same time, when determining the window width and window position for high-contrast window linear transformation of the CT image, the CT image is first divided into multiple local image areas, and then the local image areas of interest that may be related to pancreatic tissue are screened out; then the average value μ1 of the grayscale value mean of the local image area of ​​interest and the average value μ of the grayscale value mean of the local image area are calculated respectively; and then the high-contrast window width and window position are determined. In this way, when determining the high-contrast window width and window position, the influence of the pancreatic tissue area on the window width and window position setting is highlighted; so that the fine details of the pancreatic area in the CT image after high-contrast window linear change preprocessing are more; at the same time, the present technical solution is particularly suitable for the setting of high-contrast window width and window position for linear transformation of CT images of pancreatic tissue under different states (normal or different pathological states), and has the effect of adaptively adapting to pancreatic tissue in different states. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 is a flow chart of a pancreas segmentation method for CT images provided by an embodiment of the present invention;

[0057] Figure 2 is a flow chart of performing a preprocessing operation on the CT image to obtain a preprocessed CT image provided by an embodiment of the present invention;

[0058] Figure 3 This is a flow chart of a method for determining the window width and window level of a high-contrast window provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

[0060] Example 1, attached Figure 1 A flow chart of a pancreas segmentation method for CT images is shown in FIG. Figure 1 As shown, a pancreas segmentation method for CT images includes the following steps:

[0061] S1: Acquire a CT image containing the pancreas;

[0062] CT (computed tomography) image acquisition is a crucial step in medical imaging. Its purpose is to obtain high-quality CT images using specialized scanning equipment, providing fundamental data for subsequent diagnosis and analysis. In this embodiment, CT images of the pancreas and surrounding tissues are acquired using a CT scanner, which may be a spiral CT, multislice spiral CT, or dual-source CT, among others, though this embodiment does not specifically limit this.

[0063] Before performing a CT scan, the CT scanning equipment needs to be calibrated daily. The daily calibration includes: mechanical calibration to ensure the movement accuracy of the scanning bed and the rotation accuracy of the scanning gantry; image quality calibration: using a standard phantom to test image quality, including spatial resolution, contrast resolution, noise level, etc.

[0064] After completing the above steps, it is necessary to set the CT scanning equipment parameters and scanning parameters. In this embodiment, the CT scanning equipment parameters include tube voltage, tube current, etc.; the scanning parameters include slice thickness, slice spacing, scanning range, scanning direction, etc.

[0065] The tube voltage and tube current are set according to the weight of the object to be scanned. In this embodiment, for pancreatic scanning, the tube voltage and tube current are 120 kV and 200-300 mA, respectively.

[0066] Specifically, in order to improve the resolution of CT images, the layer thickness is set to 3 mm; the layer spacing is set to be smaller than the layer thickness to avoid missing important information; the scanning range is from the top of the diaphragm to the lower pole of the kidney, ensuring that the entire pancreas and its surrounding tissues are included; the scanning direction is a cross-sectional scan, and a coronal or sagittal scan can be performed if necessary.

[0067] After the scan is completed, this step also includes storing the scanned CT image in a computer workstation and backing it up.

[0068] S2: performing a preprocessing operation on the CT image to obtain a preprocessed CT image;

[0069] Wherein, the preprocessing operation is to perform a preprocessing operation on the CT image using a multi-window frequency domain fusion method;

[0070] Among them, Figure 2 As shown, the multi-window frequency domain fusion method is used to pre-process the CT image as follows:

[0071] S2.1: performing linear transformations of different window widths and window levels on the CT image to generate a plurality of transformed images, wherein the linear transformations of different window widths and window levels include an abdominal window linear transformation, a liver window linear transformation, and a high-contrast window linear transformation;

[0072] In CT (computed tomography) images, window width (WW) and window level (WL) are key parameters used to adjust the image display range, aiming to optimize the contrast visualization of different tissues. Window width (WW) represents the display range of CT values ​​and determines image contrast. A narrow window width (e.g., 150 HU) enhances contrast between soft tissues and is suitable for observing tissues of similar density (e.g., liver / pancreas). A wide window width (e.g., 600 HU) displays a wider range of CT values ​​and is suitable for observing high-contrast structures (e.g., bones or air-containing organs). Window level (WL) represents the central CT value (Houns field unit, HU) of the window width and determines which tissues are displayed. For example, WL = 40 HU indicates that soft tissue (e.g., muscle) is centered on the image. Existing designs include abdominal and liver window transformations and image fusion. The abdominal window is used to display the overall contour of the pancreas and the surrounding fat space, while the liver window is used to enhance the contrast between blood vessels and the adjacent areas of the liver and pancreas. For example, Table 1 shows the main uses and display effects of abdominal and liver windows. As shown in Table 1,

[0073] Table 1 Main uses and display effects of abdominal windows and liver windows

[0074]

[0075] However, during the preprocessing of CT images of the pancreas, due to the small size of the pancreas, the boundary between it and the surrounding tissues is unclear and the boundaries are blurred. The image fusion effect of the above-mentioned abdominal window transformation and liver window transformation alone is not obvious, and the segmentation accuracy of the pancreas is still low. Therefore, according to the above-mentioned technical status, this embodiment adds a high-contrast linear transformation on the basis of the abdominal window linear transformation and the liver window linear transformation. The high-contrast linear transformation is used to improve the contrast between the pancreas and the surrounding tissues in the CT image, thereby improving the accuracy of CT image segmentation of the pancreas.

[0076] The linear transformation formula is:

[0077] ;

[0078] Where, I windowed is the grayscale value after linear transformation, I is the grayscale value before linear transformation, WL is the window level, and WW is the window width;

[0079] The linear transformation of the abdominal window has a window width of 350 and a window level of 40; the linear transformation of the liver window has a window width of 150 and a window level of 30; the linear transformation of the high contrast window has a window width of 100 and a window level of 20;

[0080] Furthermore, in addition to the window width and window position of the high contrast window linear transformation directly given above, as shown in the attached Figure 3 As shown, this embodiment also includes a method for determining the window width and window level of a high contrast window, specifically:

[0081] Sa: dividing the CT image into multiple local image regions;

[0082] The CT image is divided into a plurality of local regions using a sliding window method, wherein in this embodiment, the window size of the sliding window is 3×3;

[0083] Prior art image preprocessing schemes exist that divide CT images into multiple local image regions and set a window width and window level for each local image region to perform linear transformation. However, this scheme is complex and time-consuming, and the preprocessing effect is unsatisfactory. This may be because different linear transformation parameters are used for the same tissue at the intersection of local image regions, affecting the normal CT expression of the tissue. Therefore, this embodiment abandons this operational approach and proposes a new scheme for determining the window level and window width.

[0084] Sb: Calculate the mean grayscale value of each local image area;

[0085] The calculation formula for the mean grayscale value of each local image area is:

[0086] ;

[0087] Where μ j is the mean gray value of the jth local image area, R j It is j local image regions, N j is the number of pixels in the jth local image area, I i is the grayscale value of the i-th pixel.

[0088] Sc: screening out the local image region of interest according to the mean grayscale value of each local image region;

[0089] The local image region of interest is a region in the local image region that may contain pancreatic tissue or may contain part of pancreatic tissue;

[0090] Since the pancreas in the CT image may be normal pancreatic tissue or diseased pancreatic tissue, this step screens out the local image region of interest that may contain or partially contain the pancreas. This is only used to determine the subsequent linear transformation parameters; that is, the local image region with a grayscale value mean greater than a preset threshold is regarded as the local image region of interest.

[0091] Sd: Calculate the average value μ1 of the grayscale value means of the local image region of interest and the average value μ of the grayscale value means of the multiple local image regions;

[0092] In this step, the average value of the grayscale value means of the local image region of interest and the average value of the grayscale value means of multiple local image regions are calculated for subsequent determination of the linear transformation parameters of the CT image.

[0093] Se: determining the window width and window level of the high contrast window according to the average value μ1 of the grayscale value mean of the local image region of interest and the average value μ of the grayscale value mean of the multiple local image regions;

[0094] Wherein, the window level WL of the high contrast window is:

[0095] ;

[0096] Where a is the coefficient;

[0097] Wherein, the window width WW of the high contrast window is:

[0098] ;

[0099] Furthermore, the coefficient a is determined as follows:

[0100] Performing a linear transformation preprocessing operation on the CT image using an initial window width and an initial window position;

[0101] Wherein, the initial window width is 100, and the initial window level is 20;

[0102] Calculate the contrast-to-noise ratio (CNR) of the target tissue and background tissue in the preprocessed CT image;

[0103] wherein, after preprocessing the CT image, the CT image is input into an image segmentation model to segment the pancreatic tissue and background tissue; and then the contrast-to-noise ratio of the pancreatic tissue to the background tissue in the preprocessed CT image is calculated;

[0104] Determine coefficients a1 and a2 according to the contrast-to-noise ratio of the target tissue to the background tissue and the target contrast-to-noise ratio;

[0105] The target contrast-to-noise ratio is a preset value.

[0106] The calculation formula of the coefficient a is:

[0107] ;

[0108] Where, is the target contrast-to-noise ratio, It is the contrast-to-noise ratio between the target tissue and the background tissue obtained by preprocessing the CT image using the initial window width and initial window level.

[0109] Through the above scheme, when determining the window width and window level for high-contrast window linear transformation of a CT image, the CT image is first divided into multiple local image areas, and then the local image areas of interest that may be related to pancreatic tissue are screened out; then the average value μ1 of the grayscale value mean of the local image area of ​​interest and the average value μ of the grayscale value mean of the local image area are calculated respectively; and then the high-contrast window width and window level are determined. In this way, when determining the high-contrast window width and window level, the influence of the pancreatic tissue area on the window width and window level setting is highlighted; so that the fine details of the pancreatic area in the CT image after high-contrast window linear change preprocessing are more; at the same time, this technical scheme is particularly suitable for the setting of high-contrast window width and window level for linear transformation of CT images of pancreatic tissue under different states (normal or different pathological states), and has the effect of adaptively adapting to pancreatic tissue in different states.

[0110] S2.2: Performing Fourier transform on the multiple transformed images to obtain frequency spectrum information of each transformed image;

[0111] Among them, by performing fast Fourier transform on the abdominal window linear transformation image, liver window linear transformation image and high contrast linear transformation image, the abdominal window linear transformation image mainly retains the morphological information of the pancreas, the liver window linear transformation image mainly enhances the blood vessels and boundary information, and the high contrast linear transformation image mainly displays the fine details of the pancreas and possible pancreatic lesion location information.

[0112] S2.3: performing a partitioning operation on the spectrum information of each transformed image to obtain low-frequency spectrum information of the abdominal window linear transformation image, mid-frequency spectrum information of the liver window linear transformation image, and high-frequency spectrum information of the high-contrast window linear transformation image;

[0113] The center of the spectrum information is taken as the origin, and the radius r of each frequency point (u, v) in the spectrum information is calculated;

[0114] ;

[0115] Among them, the low frequency area: ;Intermediate frequency area: ;

[0116] High frequency area: ; r max is the maximum radius from each frequency point to the origin,

[0117] The above-mentioned formulas for determining the low-frequency area, the intermediate-frequency area and the high-frequency area are used to partition the abdominal window linear transformation image, the liver window linear transformation image and the high-contrast window linear transformation image respectively, so as to obtain the low-frequency area of ​​the abdominal window linear transformation image, the intermediate-frequency area of ​​the liver window linear transformation image and the high-frequency area of ​​the high-contrast window linear transformation image.

[0118] S2.4: Calculate the fusion weight function of the spectrum information of each transformed image;

[0119] The fusion weight function of the low-frequency region, the medium-frequency region and the high-frequency region is:

[0120] ;

[0121] Where, is the low-frequency weight function, is the mid-frequency weight function, is the high frequency area weight function; σ L and σ H is the control parameter;

[0122] S2.5: Performing spectrum fusion based on the fusion weight functions of the low-frequency region, the intermediate-frequency region, and the high-frequency region and the low-frequency spectrum information of the abdominal window linear transformation image, the intermediate-frequency spectrum information of the liver window linear transformation image, and the high-frequency spectrum information of the high-contrast window linear transformation image;

[0123] Among them, the formula for image fusion is:

[0124]

[0125] Where, is the spectrum information of the fused image, is the low-frequency weight function, is the mid-frequency weight function, is the high frequency area weight function; is the low-frequency spectrum information of the abdominal window linear transformation image; is the spectrum information of the intermediate frequency region of the liver window linear transformation image; The high-frequency spectrum information of the image is linearly transformed with a high-contrast window.

[0126] S2.6: Perform inverse Fourier transform on the fused spectrum information to obtain a fused image.

[0127] In this step, the CT image is first subjected to abdominal window linear transformation, liver window linear transformation and high contrast window linear transformation, and then the multiple linear transformation images are respectively subjected to Fourier transformation to obtain the spectrum information of each linear transformation image, and then the spectrum information of each linear transformation image is partitioned to obtain the low-frequency region spectrum information of the abdominal window linear transformation image, the medium-frequency region spectrum information of the liver window linear transformation image and the high-frequency region spectrum information of the high contrast window linear transformation image, and then a dynamic weight function is set according to each region, and finally feature fusion is performed based on the dynamic weight function and the partitioned spectrum information of each transformation image. The introduction of the high contrast window and the feature fusion scheme of the dynamic weight function and the partitioned spectrum information of each transformation image of the present invention enable the fused image to have both global structure and local details, retain the overall contours of organs such as the pancreas and liver, avoid anatomical structure breaks caused by narrow window width, and at the same time enhance blood vessels and boundaries, especially improve the contrast between the pancreas and surrounding tissues.

[0128] S3: Inputting the preprocessed CT image into a pancreatic image segmentation model to implement pancreatic segmentation.

[0129] Wherein, the pancreatic image segmentation model is a U-Net model;

[0130] U-Net is a classic convolutional neural network (CNN) originally designed for biomedical image segmentation tasks, especially cell segmentation. Due to its unique architecture and efficient performance, it has been widely used in the field of medical image segmentation, including pancreas segmentation.

[0131] The U-Net model consists of an encoder, a decoder, and an output layer. The encoder is mainly responsible for extracting image features and consists of multiple convolutional layers and pooling layers. Each convolutional layer contains two convolution operations. The convolution kernel size is usually 3×3, the stride is 1, and the padding is 1 to maintain the size of the feature map. The pooling layer usually uses maximum pooling with a pooling kernel size of 2×2 and a stride of 2 for downsampling, reducing the size of the feature map while extracting higher-level features.

[0132] The decoder's function is to gradually upsample the features extracted by the encoder, restoring them to the same size as the input image and generating a segmentation result. The decoder consists of multiple upsampling and convolutional layers. The upsampling layers typically employ transposed convolution or nearest neighbor interpolation. At each decoder stage, the upsampled feature map is fused with the feature map from the encoder stage via a skip connection. The skip connection transfers low-level features (such as edges and texture information) from the encoder stage to the decoder stage, helping to preserve image detail. The output layer is a 1×1 convolutional layer, used to map the decoder's feature map to the segmentation result. The number of output channels equals the number of categories of the segmentation target. In this embodiment, since this is pancreas segmentation, it has two channels, meaning the output result is the pancreas and background.

[0133] In Example 2, the present invention further provides a pancreas segmentation system for CT images. The system adopts the pancreas segmentation method for CT images of Example 1. The system includes:

[0134] An image acquisition module, used for acquiring a CT image containing the pancreas;

[0135] A preprocessing module, configured to perform a preprocessing operation on the CT image to obtain a preprocessed CT image;

[0136] The image segmentation module is used to input the preprocessed CT image into a pancreas image segmentation model to implement pancreas segmentation.

[0137] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.

[0138] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0139] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions to implement a pancreas segmentation method for CT images and / or other desired functions of any embodiment of the present application described above. Various contents such as initial external parameters and thresholds may also be stored in the computer-readable storage medium.

[0140] In one example, the electronic device may further include an input device and an output device, these components interconnected via a bus device and / or other connection mechanism (not shown). The input device may include, for example, a keyboard, a mouse, etc. The output device may output various information to the outside, including warning information, braking force, etc. The output device may include, for example, a display, a speaker, a printer, a communication network, and remote output devices connected thereto.

[0141] Of course, for the sake of simplicity, components such as buses, input / output interfaces, etc. are omitted. In addition, the electronic device may further include any other appropriate components according to specific application scenarios.

[0142] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to implement the functions of a pancreas segmentation method for CT images provided by any embodiment of the present application.

[0143] The computer program product may be written in any combination of one or more programming languages ​​to implement the program code for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0144] In addition, an embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, the processor implements a pancreas segmentation method for CT images provided by any embodiment of the present application.

[0145] The computer-readable storage medium may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor device, apparatus, or component, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A pancreas segmentation method for CT images, characterized in that: include: S1: Acquire a CT image containing the pancreas; S2: performing a preprocessing operation on the CT image to obtain a preprocessed CT image; The S2 is specifically: S2.1: performing linear transformations of different window widths and window levels on the CT image to generate a plurality of transformed images, wherein the linear transformations of different window widths and window levels include an abdominal window linear transformation, a liver window linear transformation, and a high-contrast window linear transformation; S2.2: Performing Fourier transform on the multiple transformed images to obtain frequency spectrum information of each transformed image; S2.3: performing a partitioning operation on the spectrum information of each transformed image to obtain low-frequency spectrum information of the abdominal window linear transformation image, mid-frequency spectrum information of the liver window linear transformation image, and high-frequency spectrum information of the high-contrast window linear transformation image; S2.4: Calculate the fusion weight function of the spectrum information of each transformed image; S2.5: Performing spectrum fusion based on the fusion weight functions of the low-frequency region, the intermediate-frequency region, and the high-frequency region and the low-frequency spectrum information of the abdominal window linear transformation image, the intermediate-frequency spectrum information of the liver window linear transformation image, and the high-frequency spectrum information of the high-contrast window linear transformation image; S2.6: Perform inverse Fourier transform on the fused spectrum information to obtain a fused image; S3: Inputting the preprocessed CT image into a pancreatic image segmentation model to implement pancreatic segmentation.

2. The pancreas segmentation method for CT images according to claim 1, characterized in that: The S2.3 specifically comprises: taking the center of the spectrum information as the origin, and calculating the radius r of each frequency point (u, v) in the spectrum information; ； Among them, the low frequency area: ;Intermediate frequency area: ; High frequency area: ; r max is the maximum radius from each frequency point to the origin; Then the abdominal window linear transformation image, liver window linear transformation image and high contrast window linear transformation image are partitioned respectively to obtain the low frequency area of ​​the abdominal window linear transformation image, the intermediate frequency area of ​​the liver window linear transformation image and the high frequency area of ​​the high contrast window linear transformation image.

3. The pancreas segmentation method for CT images according to claim 2, characterized in that: In S2.4, the fusion weight function of the low-frequency region of the abdominal window linear transformation image, the intermediate-frequency region of the liver window linear transformation image, and the high-frequency region of the high-contrast window linear transformation image is: ； Where, is the low-frequency weight function, is the mid-frequency weight function, is the high frequency area weight function; σ L and σ H is the control parameter.

4. The pancreas segmentation method for CT images according to claim 1, characterized in that: The formula for the linear transformation is: ; Where, I windowed is the grayscale value after linear transformation, I is the grayscale value before linear transformation, WL is the window level, and WW is the window width.

5. The pancreas segmentation method for CT images according to claim 1 or 4, characterized in that: The window width of the abdominal window linearly transformed is 350, and the window level is 40; the window width of the liver window linearly transformed is 150, and the window level is 30; the window width of the high contrast window linearly transformed is 100, and the window level is 20.

6. The pancreas segmentation method for CT images according to claim 1, characterized in that: The method for determining the window width and window level of a high contrast window is: Sa: dividing the CT image into multiple local image regions; Sb: Calculate the mean grayscale value of each local image area; Sc: screening out the local image region of interest according to the mean gray value of each local image region; Sd: Calculate the average value μ1 of the grayscale value means of the local image region of interest and the average value μ of the grayscale value means of the multiple local image regions; Se: Determine the window width and window level of the high contrast window according to the average value μ1 of the grayscale value means of the local image region of interest and the average value μ of the grayscale value means of the multiple local image regions.

7. The pancreas segmentation method for CT images according to claim 6, characterized in that: The window level WL of the high contrast window is: ； Where a is the coefficient; Wherein, the window width WW of the high contrast window is: ； The calculation formula of the coefficient a is: ； Where, is the target contrast-to-noise ratio, It is the contrast-to-noise ratio between the target tissue and the background tissue obtained by preprocessing the CT image using the initial window width and initial window level.

8. The pancreas segmentation method for CT images according to claim 2, characterized in that: In S2.5, the image fusion formula is: Where, is the spectrum information of the fused image, is the low-frequency weight function, is the mid-frequency weight function, is the high frequency area weight function; is the low-frequency spectrum information of the abdominal window linear transformation image; is the spectrum information of the intermediate frequency region of the liver window linear transformation image; It is the high-frequency spectrum information of the high-contrast window linearly transformed image.

9. A pancreas segmentation system for CT images, characterized in that: The system adopts the pancreas segmentation method for CT images according to any one of claims 1 to 8, and the system comprises: An image acquisition module, used for acquiring a CT image containing the pancreas; A preprocessing module, configured to perform a preprocessing operation on the CT image to obtain a preprocessed CT image; The image segmentation module is used to input the preprocessed CT image into a pancreas image segmentation model to implement pancreas segmentation.

Images