CT Image Segmentation Processing Method and Device

Abstract

The present invention provides a CT image segmentation processing method and apparatus, which includes obtaining an original image; performing windowing preprocessing on the original image to obtain a first image; performing segmentation processing on the first image based on a first image segmentation model to obtain a segmented image, and performing denoising processing on the segmented image to determine a region of interest; using the region of interest to train a second image segmentation model, and segmenting a target region from the region of interest based on the trained second image segmentation model. The present invention can effectively solve the problems of insufficient segmentation or over-segmentation caused by adjacent organs, false positives, and abnormal regions during the image segmentation process, thereby achieving the technical effect of improving the segmentation accuracy.

Description

Technical Field

[0001] This application relates to the technical field of medical image processing. Specifically, it relates to a CT image segmentation processing method, apparatus, computer device, and storage medium. Background Art

[0002] Currently, computational methods for liver segmentation have been widely proposed. For example: an interactive method that combines the random walk model, superpixels, and active contour model; a method for automatic liver segmentation using threshold segmentation, generalized edge model, and statistical shape model; a method for liver segmentation using level sets, shape statistical models, and graph cut techniques. However, the above algorithms all have limitations to varying degrees. The main drawback of the active contour-based method is its dependence on the boundary starting point and operating parameters; the threshold-based method that uses the contrast between the liver and other abdominal organs for segmentation is prone to segmentation errors; intensity statistical distribution-based methods, such as level sets, shape statistical models, and graph cut, are difficult to apply to the boundary region, resulting in under-segmentation or over-segmentation.

[0003] The automatic liver segmentation process presents the following challenges: First, since the liver is a soft tissue organ, the shape of the liver is affected by adjacent organs and thus varies greatly. Second, the boundaries of adjacent organs are blurred, which makes visualization difficult, especially for abdominal organs such as the stomach or spleen. In addition, lesions can affect the appearance and shape of the liver, change its density, signal intensity, and thus distort its structure.

[0004] Therefore, there is an urgent need for an automatic segmentation method and apparatus that can effectively solve the above technical problems and thus improve the segmentation effect. Summary of the Invention

[0005] Embodiments of the present invention provide a CT image segmentation processing method, apparatus, computer device, and storage medium to solve the problem of under-segmentation or over-segmentation caused by adjacent organs, false positives, and abnormal regions during the image segmentation process.

[0006] To achieve the above object, in the first aspect of the embodiments of the present invention, there is provided a CT image segmentation processing method, including:

[0007] Obtain an original image;

[0008] Perform windowing preprocessing on the original image to obtain a first image;

[0009] Perform segmentation processing on the first image based on a first image segmentation model to obtain a segmented image, and perform denoising processing on the segmented image to determine a region of interest;

[0010] Use the region of interest to train the second image segmentation model, and segment the target region from the region of interest based on the trained second image segmentation model.

[0011] Optionally, in a possible implementation manner of the first aspect, the method further includes:

[0012] Use the trained third image segmentation model to segment the abnormal region from the region of interest;

[0013] Jointly reconstruct the abnormal region and the target region to obtain the final segmentation result.

[0014] Optionally, in a possible implementation manner of the first aspect, performing windowing preprocessing on the original image to obtain the first image includes:

[0015] Perform windowing processing on the voxel values in the original image whose CT values are in the range of [-100, 200];

[0016] Configure the voxel values less than -100 in the original image to a fixed CT value of -100, and configure the voxel values greater than 200 to a fixed CT value of 200.

[0017] Optionally, in a possible implementation manner of the first aspect, the method further includes:

[0018] Use abdominal CT slices to train the first image segmentation model;

[0019] Use the abnormal regions in the region of interest to train the third image segmentation model.

[0020] Optionally, in a possible implementation manner of the first aspect, performing denoising processing on the segmented image includes:

[0021] Based on the constructed diagonal histogram, remove the structures connected to the liver from the segmented image to obtain the region of interest containing the liver. The diagonal histogram consists of 100 bin widths, each bin width representing the number of corresponding pixels where the corresponding diagonal exists. The valley point of the diagonal histogram represents the cut-off point where the liver separates from other structures.

[0022] In the second aspect of the embodiments of the present invention, there is provided a CT image segmentation processing device, including:

[0023] An acquisition module, configured to acquire the original image;

[0024] A preprocessing module, configured to perform windowing preprocessing on the original image to obtain the first image;

[0025] The region of interest determination module is configured to perform segmentation processing on the first image based on the first image segmentation model to obtain a segmented image, perform denoising processing on the segmented image, and determine the region of interest;

[0026] The target region segmentation module is configured to use the region of interest to train the second image segmentation model, and segment the target region from the region of interest based on the trained second image segmentation model.

[0027] Optionally, in a possible implementation manner of the second aspect, the apparatus further includes:

[0028] The abnormal region segmentation module is configured to segment the abnormal region from the region of interest by using the trained third image segmentation model;

[0029] The region joint reconstruction module is configured to jointly reconstruct the abnormal region and the target region to obtain the final segmentation result.

[0030] Optionally, in a possible implementation manner of the second aspect, the region of interest determination module includes:

[0031] The connected organ removal unit is configured to remove the structures connected to the liver from the segmented image based on the constructed diagonal histogram, so as to obtain the region of interest including the liver. The diagonal histogram consists of 100 bin widths, each bin width represents the number of corresponding pixels existing in the corresponding diagonal, and the valley point of the diagonal histogram represents the cut-off point where the liver is separated from other structures.

[0032] In the third aspect of the embodiments of the present invention, there is provided a computer device, including a memory and a processor, where the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps in the above-mentioned various method embodiments are implemented.

[0033] In the fourth aspect of the embodiments of the present invention, there is provided a readable storage medium, where a computer program is stored in the readable storage medium, and when the computer program is executed by a processor, the steps of the method in the first aspect and various possible designs of the first aspect of the present invention are implemented.

[0034] The CT image segmentation processing method, device, computer device, and storage medium provided by the present invention obtain an original image; perform windowing preprocessing on the original image to obtain a first image; perform segmentation processing on the first image based on a first image segmentation model to obtain a segmented image, and perform denoising processing on the segmented image to determine a region of interest; use the region of interest to train a second image segmentation model, and segment a target region from the region of interest based on the trained second image segmentation model. The present invention can effectively solve the problems of insufficient segmentation or over-segmentation caused by adjacent organs, false positives, and abnormal regions during image segmentation, thereby achieving the technical effect of improving segmentation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of the first implementation manner of the CT image segmentation processing method;

[0036] Figure 2 It is a schematic diagram of windowing processing;

[0037] Figure 3 It is an architecture diagram of the LU-Net model;

[0038] Figure 4 It is a schematic diagram of eliminating false positives;

[0039] Figure 5 It is a schematic diagram of eliminating organs connected to the liver;

[0040] Figure 6 It is a schematic diagram of histogram construction;

[0041] Figure 7 It is a schematic diagram of dividing the region of interest;

[0042] Figure 8 It is a schematic diagram of initial segmentation;

[0043] Figure 9 It is a schematic diagram of abnormal region segmentation;

[0044] Figure 10 It is a schematic diagram of joint reconstruction;

[0045] Figure 11 Schematic diagram of eliminating false positives and filling holes;

[0046] Figure 12 It is a structural diagram of the first implementation manner of the CT image segmentation processing device. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0049] It should be understood that in various embodiments of the present invention, the magnitude of the serial numbers of the processes does not mean the sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0050] It should be understood that in the present invention, "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0051] It should be understood that in the present invention, "a plurality of" means two or more. "And / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects before and after. "Including A, B and C" and "including A, B, C" mean that all of A, B and C are included. "Including A, B or C" means including any one of A, B and C. "Including A, B and / or C" means including any one or any two or all three of A, B and C.

[0052] It should be understood that in the present invention, "B corresponding to A", "B corresponding to A relatively", "A corresponding to B relatively" or "B corresponding to A relatively" means that B is associated with A, and B can be determined according to A. Determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information. The matching of A and B means that the similarity between A and B is greater than or equal to a preset threshold.

[0053] Depending on the context, as used herein, "if" can be interpreted as "when" or "while" or "in response to determining" or "in response to detecting".

[0054] The technical solution of the present invention will be described in detail below with specific embodiments. These specific embodiments below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0055] The present invention provides a CT image segmentation processing method, as Figure 1 shown in its flowchart, including:

[0056] Step S110, obtaining an original image.

[0057] In this step, the original image is obtained from a pre-created CT image database, where the resolution of all images in the database in the axial plane is 512×512 pixels; during the creation of the CT image database, due to the differences in each hospital or medical institution, the images in the database are from different acquisition methods and different models of scanning devices, so the database covers images with different spatial resolutions and field of view angles. A large number of disease sources determine that the lesions in the database have different shapes, quantities, sizes, and contrasts, and there should also be cases of primary and secondary cancers.

[0058] Step S120, performing windowing preprocessing on the original image to obtain a first image.

[0059] In step S120, when preprocessing the original image (i.e., abdominal CT slice image), it is necessary to perform windowing processing on the voxel values within the window range of CT values [-100, 200] to enhance the contrast between the liver and surrounding organs and tissues; for voxels with CT values less than -100 or greater than 200, the values of -100 HU and 200 HU are set respectively, Figure 2 representing the result of the windowing processing of the CT slice image. Tissues within the range of -100 HU to +200 HU, a total of 300 HU, can be displayed, while tissues greater than +200 HU are displayed as white; tissues less than -100 HU are displayed as black. After the windowing processing, the voxel values are recalibrated between 0 and 1 to minimize stability problems in the deep neural network.

[0060] Step S130, performing segmentation processing on the first image based on the first image segmentation model to obtain a segmented image, and performing denoising processing on the segmented image to determine the region of interest.

[0061] In step S130, after windowing preprocessing, the liver region is segmented by a two-step cascaded method. The first step is to determine the region of interest (ROI), within which the liver region is included. The second step uses the ROI as the input to an image segmentation model to segment the liver region. This method can significantly reduce the number of voxels in the second step and enable more detailed liver segmentation.

[0062] Specifically, the first image segmentation model in the process of determining the ROI in step S130 refers to the LU-Net model. The second and third image segmentation models involved in subsequent steps are also LU-Net models, but the three are trained based on different training sets. Among them, the LU-Net model is a new medical image segmentation network model based on the application of CNN. In this technical solution, the traditional U-Net network is optimized, and its model architecture is as Figure 3 shown. LU-Net contains a contracting path (left side), which includes five convolutional blocks, using 32, 64, 128, 256, and 512 filters. The convolutional blocks are composed of a convolutional layer (kernel 3×3), an activation function Leaky Rectified linear unit (LeakyReLu), dropout regularization, and batch normalization. The expanding path (right side) of LU-Net contains four layers of transposed convolutions, which perform upsampling using 256, 128, 64, and 32 filters respectively. Subsequently, the transposed convolutions are cascaded with the corresponding feature maps of the convolutional blocks in the contracting path. The sigmoid function is used to unfold the last layer of the path, and the final result is presented. The Dice loss function is used for network training. As a CNN, the biggest advantage of the LU-Net architecture is that it can automatically learn the most representative image features without the need for feature engineering. In addition, the connection between the contracting and expanding paths helps with information transmission.

[0063] In step S130, the first image segmentation model uses lumbar CT slices with an original pixel size of 512×512 to train the LU-Net model. This model uses a 2.5D method, and the data input to the neural network is a stack of three axial slices: the central slice and its front and back slices. The output of LU-Net is the segmentation map of the central slice of the stack. By combining the segmentations of each slice, a 3D segmentation volume can be generated. "Denoising processing" is mainly divided into two aspects, as follows:

[0064] One is to reduce the influence of false positives. False positives refer to Figure 4 sporadic fragments not in the liver in Figure 4 b) in Figure 4 Remove the sporadic fragments not in the liver and retain the segmentation object with the largest area (

[0065] Second, after reducing false positives, there are still organs or tissues connected to the liver, such as the stomach and spleen close to the liver, as Figure 5 (a) shows that the liver segmentation is connected to other abdominal organs, so it is necessary to eliminate adjacent organs or tissues. The specific implementation steps for segmenting the connected organs are as follows:

[0066] In one embodiment, the denoising process of the segmented image includes: based on the constructed diagonal histogram, removing the structures connected to the liver from the segmented image to obtain the region of interest containing the liver. The diagonal histogram consists of 100 bins, and each bin represents the number of corresponding pixels existing on the corresponding diagonal. The valley point of the diagonal histogram represents the cut-off point where the liver is separated from other structures.

[0067] In this step, the denoising process is the specific implementation of "eliminating the interference of adjacent organs or tissues" in the above step S130. The elimination of these adjacent organs is based on the diagonal projection histogram of the segmented image in the axial plane. The choice of using the diagonal to construct the histogram is based on the analysis of all slice volumes, and it can be predicted that the liver is approximately or completely above the diagonal ( Figure 5 (d)), and other abdominal organs related to the segmentation are below the diagonal. Among them, the interval of the diagonal histogram is defined between diagonal 1 and diagonal 2, specifically as Figure 6 (a) shows that the two diagonals are 100 pixels apart, and this distance value is empirically defined through experiments by analyzing the results of the basic patients and the false positive reduction steps; the diagonal histogram consists of 100 bins, and each bin represents the number of corresponding pixels existing on the corresponding diagonal. The bottom of the histogram represents the cut-off point where the liver is separated from other organs, specifically as Figure 6 (b) shows, and then the liver is separated from other organs according to the bottom of the histogram, specifically as Figure 6 (c) shows.

[0068] After eliminating the interference of adjacent organs, first define the first bounding box in the obtained segmented image ( Figure 7 (a)); then add five pixels in each direction of the height and width of the first bounding box to obtain the second bounding box ( Figure 7 (b)); finally, apply the second bounding box to the abdominal CT slice ( Figure 7 (c)), thereby dividing the region of interest ( Figure 7 (d)).

[0069] Step S140: Use the region of interest to train the second image segmentation model, and segment the target region from the region of interest based on the trained second image segmentation model.

[0070] In step S140, the second image segmentation model is the LU-Net model, and its model training is carried out under the condition of delimiting the region of interest; this model uses a 2.5D method. The specific initial segmentation process is as Figure 8 shown. According to the trained second image segmentation model, the liver region is segmented from the delimited region of interest (such as Figure 8 (a)), and then the scattered fragments outside the liver are removed, and only the region object with the largest volume is retained (such as Figure 8 (c)), so that the obtained segmentation reduces false positives.

[0071] In one embodiment, the method further includes: using the trained third image segmentation model to segment the abnormal region from the region of interest; jointly reconstructing the abnormal region and the target region to obtain the final segmentation result.

[0072] In this step, in the CT images of patients with healthy livers or livers slightly affected by lesions, the above initial segmentation steps S110-S140 are sufficient to achieve; however, in the complex case of advanced liver lesions, due to the differences between the lesion textures and healthy liver tissues, steps S110-S140 may exclude the liver regions containing these lesions, such as Figure 8 (b), (c) shown, the segmented target liver does not contain abnormal regions. For this situation, it is necessary to first re-determine the abnormal regions in the region of interest; then train the third image segmentation model based on the abnormal regions, so as to use the trained third image segmentation model to segment the abnormal regions contained in the region of interest, such as Figure 9 shown; finally, the liver region obtained by the initial segmentation and the liver region obtained by the lesion segmentation are jointly reconstructed. This joint is made by a binary operation (OR) on the initial segmentation and the lesion segmentation masks, Figure 10 showing the process of this joint reconstruction. Since the joint reconstruction enables the recovery of the abnormal regions, the segmentation accuracy is also improved. This reconstruction step allows the recovery of abnormal regions with significant texture differences between the abnormal regions and healthy tissues. In addition, this kind of reconstruction does not affect the initial segmentation of patients with healthy livers or livers slightly affected by lesions.

[0073] In one embodiment, the method further includes: the image generated after the joint reconstruction may contain fragments outside the liver (false positive fragments), and there may also be unfilled holes. For this situation, the present invention further includes two steps of reducing false positives and filling holes after the joint reconstruction, such as Figure 11 shown, reducing false positives to only retain the object with the largest volume in the volume, removing the scattered fragments; and using a morphological closing operation with a 7×7 circular structuring element to reduce the holes existing in the final segmentation, and the parameters of this morphological closing operation are artificially defined after several experiments.

[0074] The CT image segmentation processing method provided by the present invention includes: obtaining an original image; performing windowing preprocessing on the original image to obtain a first image; performing segmentation processing on the first image based on a first image segmentation model to obtain a segmented image, and performing denoising processing on the segmented image to determine a region of interest; using the region of interest to train a second image segmentation model, and segmenting a target region from the region of interest based on the trained second image segmentation model. The present invention can effectively solve the problems of insufficient segmentation or over-segmentation caused by adjacent organs, false positives, and abnormal regions during image segmentation, thereby achieving the technical effect of improving the accuracy of liver segmentation.

[0075] Technical effects:

[0076] 1. The present invention applies a cascading method in the initial segmentation step, that is, the scope of the problem is narrowed down to the second step in the first step, effectively reducing errors in the initial segmentation;

[0077] 2. Diagonal projection histogram analysis eliminates false positives in previous segmentation steps and improves the accuracy of liver segmentation;

[0078] 3. In the initial segmentation step, the region-of-interest delimitation sub-step limits the scope of the CT image to the region containing the liver, excluding other image details, and excluding other organs and tissues in the abdominal region with features similar to the liver;

[0079] 4. The unique reconstruction step of the present invention restores the lesion regions excluded in the initial segmentation, thereby improving the segmentation result. This method can be applied to the livers of healthy people as well as livers with severe lesions, effectively solving the problem of uneven segmentation caused by different liver sizes and densities among different people;

[0080] 5. The present invention realizes both reducing false positives and excluding non-liver segments, and finally only retains the largest element in terms of volume. In the final segmentation step, a combination of reducing false positives and hole filling is adopted, improving the traditional segmentation result and enhancing the segmentation accuracy and integrity.

[0081] An embodiment of the present invention also provides a CT image segmentation processing device, as Figure 12 shown, including:

[0082] An acquisition module, configured to acquire an original image;

[0083] A preprocessing module, configured to perform windowing preprocessing on the original image to obtain a first image;

[0084] The region of interest determination module is configured to perform segmentation processing on the first image based on the first image segmentation model to obtain a segmented image, and perform denoising processing on the segmented image to determine the region of interest;

[0085] The target region segmentation module is configured to use the region of interest to train the second image segmentation model, and segment the target region from the region of interest based on the trained second image segmentation model.

[0086] In one embodiment, the apparatus further includes:

[0087] The abnormal region segmentation module is configured to segment the abnormal region from the region of interest by using the trained third image segmentation model;

[0088] The region joint reconstruction module is configured to jointly reconstruct the abnormal region and the target region to obtain a final segmentation result.

[0089] In one embodiment, the region of interest determination module includes:

[0090] The connected organ removal unit is configured to remove the structures connected to the liver from the segmented image based on the constructed diagonal histogram, so as to obtain the region of interest including the liver. The diagonal histogram consists of 100 bin widths, each bin width respectively representing the number of corresponding pixels where the corresponding diagonal exists. The valley point of the diagonal histogram represents the cut-off point where the liver is separated from other structures.

[0091] Wherein, the readable storage medium may be a computer storage medium or a communication medium. The communication medium includes any medium that facilitates the transfer of a computer program from one place to another. The computer storage medium may be any available medium accessible by a general-purpose or special-purpose computer. For example, the readable storage medium is coupled to the processor, so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium may also be a component of the processor. The processor and the readable storage medium may be located in an application specific integrated circuit (ASIC). Additionally, the ASIC may be located in the user equipment. Of course, the processor and the readable storage medium may also exist as discrete components in the communication device. The readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0092] The present invention also provides a program product, which includes execution instructions stored in a readable storage medium. At least one processor of the device can read the execution instructions from the readable storage medium, and the execution of the execution instructions by at least one processor enables the device to implement the methods provided by the above various embodiments.

[0093] In the above embodiments of the terminal or the server, it should be understood that the processor may be a central processing unit (CPU for short), or may also be other general-purpose processors, digital signal processors (DSP for short), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the present invention can be directly embodied as being completed by the execution of the hardware processor, or can be completed by a combination of hardware and software modules in the processor.

[0094] 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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A CT image segmentation processing method, characterized in that, it includes: Obtain the original image; Perform windowing preprocessing on the original image to obtain a first image; Based on the first image segmentation model, perform segmentation processing on the first image to obtain a segmented image, and perform denoising processing on the segmented image to determine the region of interest; Performing denoising processing on the segmented image includes: based on the constructed diagonal histogram, removing the structures connected to the liver from the segmented image, so as to obtain the region of interest containing the liver. The diagonal histogram consists of 100 class intervals, each class interval represents the number of corresponding pixels where the corresponding diagonal exists, and the valley point of the diagonal histogram represents the cut-off point where the liver is separated from other structures; Use the region of interest to train the second image segmentation model, and based on the trained second image segmentation model, segment the target region from the region of interest; Use the trained third image segmentation model to segment the abnormal region from the region of interest; jointly reconstruct the abnormal region and the target region to obtain the final segmentation result.

2. The CT image segmentation processing method according to claim 1, characterized in that, Performing windowing preprocessing on the original image to obtain a first image includes: Perform windowing processing on the voxel values in the range of [-100, 200] of the CT values in the original image; Configure the voxel values less than -100 in the original image to a fixed -100 CT value, and configure the voxel values greater than 200 to a fixed 200 CT value.

3. The CT image segmentation processing method according to claim 1, characterized in that, The method further includes: Use abdominal CT slices to train the first image segmentation model; Use the abnormal regions in the region of interest to train the third image segmentation model.

4. A CT image segmentation processing device, characterized in that, it includes: An acquisition module for acquiring the original image; A preprocessing module for performing windowing preprocessing on the original image to obtain a first image; A region of interest determination module for performing segmentation processing on the first image based on the first image segmentation model to obtain a segmented image, and performing denoising processing on the segmented image to determine the region of interest; A target region segmentation module for using the region of interest to train the second image segmentation model, and based on the trained second image segmentation model, segmenting the target region from the region of interest; The device further includes: An abnormal region segmentation module for using the trained third image segmentation model to segment the abnormal region from the region of interest; A region joint reconstruction module for jointly reconstructing the abnormal region and the target region to obtain the final segmentation result.

5. The CT image segmentation processing device according to claim 4, characterized in that, The region of interest determination module includes: An adjacent organ removal unit for removing structures connected to the liver from the segmented image based on the constructed diagonal histogram, so as to obtain a region of interest containing the liver. The diagonal histogram consists of 100 class intervals, each class interval representing the number of corresponding pixels where the corresponding diagonal exists. The valley point of the diagonal histogram represents the cut-off point where the liver is separated from other structures.

6. A computer device, comprising a memory and a processor, where the memory stores a computer program that can run on the processor, characterized in that when the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

7. A computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 3.

Images