Region of interest segmentation method and device, computer device and storage medium
By performing interval sampling and neural network segmentation on the initial medical images, combined with interpolation processing, the problem of doctors manually outlining regions of interest, which is time-consuming and labor-intensive, is solved, improving segmentation efficiency and saving computing resources.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNITED IMAGING RES INST OF INTELLIGENT IMAGING
- Filing Date
- 2021-12-23
- Publication Date
- 2026-07-21
AI Technical Summary
In current technology, doctors manually delineate regions of interest, which is time-consuming, labor-intensive, and inefficient, especially when there are many MRI slices, resulting in low efficiency in region of interest segmentation.
By acquiring an initial medical image, performing interval sampling to obtain an intermediate medical image, and inputting it into a trained neural network model for segmentation, combined with interpolation processing, the segmentation result of the region of interest is determined, reducing the manual delineation steps.
This approach saves manpower and time, improves the segmentation efficiency of regions of interest, reduces the computational load of the segmentation network, and enhances overall segmentation efficiency.
Smart Images
Figure CN116363147B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical image processing technology, and in particular to a method, apparatus, computer device, storage medium, and computer program product for segmenting a region of interest. Background Technology
[0002] During radiotherapy in hospitals, the delineation of regions of interest (ROIs) is often involved, including the delineation of the target area and organs at risk. Accurate delineation of ROIs is crucial in radiotherapy for primary and metastatic tumors, and the determination and implementation of patient treatment plans are based on accurate ROI delineation. For example, the hippocampus, located between the thalamus and the medial temporal lobe of the brain, is part of the limbic system and is mainly responsible for long-term memory storage and orientation. In cranial radiotherapy, reducing the dose to the hippocampus, which affects organs at risk, can mitigate radiotherapy-induced neurological damage. Accurate hippocampal delineation is key to benefiting neurological function and reducing the progression of intracranial diseases.
[0003] Currently, doctors manually delineate regions of interest layer by layer on a sequence of magnetic resonance imaging (MRI) slices using a radiotherapy planning system, relying on their professional knowledge and experience. Due to the large number of MRI slices, the workload for doctors in manually delineating these regions is substantial. Therefore, manually delineating regions of interest is time-consuming, labor-intensive, and inefficient. Summary of the Invention
[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can segment regions of interest without manually delineating them, thereby saving manpower, reducing segmentation time, and improving segmentation efficiency.
[0005] Firstly, this application provides a method for segmenting regions of interest. The method includes:
[0006] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0007] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0008] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0009] In one embodiment, the method further includes:
[0010] Based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain a second segmentation result of the region of interest within the other image slices;
[0011] Based on the first segmentation result and the second segmentation result, the segmentation result of the region of interest in the initial medical image is determined.
[0012] In one embodiment, the step of performing interval sampling on the plurality of image slices to obtain an intermediate medical image includes:
[0013] The intermediate medical image is obtained by sampling the plurality of image slices at intervals of N slices, where N is a positive integer.
[0014] In one embodiment, the method further includes:
[0015] Obtain the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample;
[0016] The training sample set is obtained by performing interval sampling on the plurality of image slice samples every M image slice samples; wherein M is a positive integer;
[0017] The initial segmentation network is trained based on the training sample set to obtain the segmentation network.
[0018] In one embodiment, training a preset initial segmentation network based on the training sample set to obtain the segmentation network includes:
[0019] Using image slice samples from the training sample set as input to the initial segmentation network, the predicted segmentation result of the region of interest within the image slice samples from the training sample set is obtained;
[0020] Based on the predicted segmentation results and the labels of the true regions of interest within the image slice samples in the training sample set, the initial segmentation network is trained to obtain the segmentation network.
[0021] In one embodiment, after determining the segmentation result of the region of interest in the initial medical image, the method further includes:
[0022] The initial medical image, with the segmentation results of the region of interest identified, is registered with the scanned image of the second modality to obtain the target image;
[0023] Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained.
[0024] In one embodiment, after determining the segmentation result of the region of interest in the initial medical image, the method further includes:
[0025] The system receives correction instructions from the user, and in response to the correction instructions, adjusts the segmentation results of the region of interest in the initial medical image, and outputs the adjusted segmentation results.
[0026] Secondly, this application also provides a region of interest segmentation device. The device includes:
[0027] The first acquisition module is used to acquire an initial medical image of a first modality, wherein the initial medical image includes multiple image slices;
[0028] The first sampling processing module is used to perform interval sampling processing on the multiple image slices to obtain an intermediate medical image;
[0029] The segmentation module is used to input the intermediate medical image into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein the segmentation network is obtained by training a neural network model.
[0030] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0031] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0032] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0033] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0034] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0035] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0036] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0037] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0038] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:
[0039] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0040] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0041] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0042] The aforementioned method, apparatus, computer device, storage medium, and computer program product for segmenting regions of interest (ROIs) acquire an initial medical image of a first modality, perform interval sampling on multiple image slices to obtain an intermediate medical image, and then input the intermediate medical image into a segmentation network to obtain a first segmentation result of the ROI within the intermediate medical image. Since this embodiment obtains an intermediate medical image by performing interval sampling on multiple image slices and uses a neural network model to determine the first segmentation result of the ROI, it eliminates the need for doctors to manually delineate the ROI, thereby saving manpower and time required for ROI segmentation and improving the segmentation efficiency. Furthermore, this embodiment inputs the intermediate medical image obtained from interval sampling of multiple image slices into the segmentation network, rather than inputting all multiple image slices into the segmentation network; therefore, it can further reduce the computational load of the segmentation network, thereby further improving the segmentation efficiency of the ROI. Attached Figure Description
[0043] Figure 1 This is a flowchart illustrating a method for segmenting a region of interest according to an embodiment of this application;
[0044] Figure 2 This is a flowchart illustrating a method for determining the segmentation result of a region of interest in an initial medical image, as provided in an embodiment of this application.
[0045] Figure 3 This is a schematic diagram of a region of interest segmentation process provided in an embodiment of this application;
[0046] Figure 4 This is a flowchart illustrating a method for obtaining a segmentation network according to an embodiment of this application;
[0047] Figure 5 This is a flowchart illustrating another method for obtaining a segmentation network provided in an embodiment of this application;
[0048] Figure 6 This is a schematic flowchart of a method for obtaining the segmentation result of a region of interest in a target image according to an embodiment of this application;
[0049] Figure 7 This is a schematic diagram of a region of interest segmentation device provided in an embodiment of this application;
[0050] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0051] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0052] Reference Figure 1 , Figure 1 This is a flowchart illustrating a method for segmenting a region of interest according to an embodiment of this application. The method can be applied to a computer device and includes the following steps:
[0053] S101. Acquire an initial medical image of the first modality, wherein the initial medical image includes multiple image slices. The first modality may be an image acquired by an imaging device selected from, for example, magnetic resonance imaging (MR), computed tomography (CT), single-photon emission computed tomography (SPECT), positron emission tomography (PET), ultrasound (US), and X-ray (XR). In this embodiment, the initial medical image is selected as an MR type T1 weighted image. In this embodiment, the type of the initial medical image is not limited; the initial medical image may also be an MR diffusion-weighted image, a planar echo image, a susceptibility-weighted imaging (SWI) image, a magnetic resonance water image, or a fat image, etc.
[0054] In this embodiment, the initial medical image can be a thin-slice medical image with a thickness between 0.5 mm and 1.5 mm, or a thick-slice medical image with a thickness between 1.5 mm and 3 mm. It should be noted that the initial medical image can also be a medical image with other slice thicknesses; this embodiment does not impose specific limitations on the slice thickness of the medical image. The initial medical image can include CT medical images, MRI medical images, PET images, etc. The initial medical image can also be a three-dimensional medical image formed by multiple two-dimensional medical images.
[0055] S102. Perform interval sampling on multiple image slices to obtain an intermediate medical image.
[0056] In this embodiment, for example, a portion of the image slices at the beginning of a plurality of image slices can be sampled at intervals of two image slices, and the other image slices in the plurality of image slices can be sampled at intervals of three image slices to obtain an intermediate medical image. An intermediate medical image refers to an image slice obtained by sampling at intervals from a plurality of image slices.
[0057] S103. Input the intermediate medical image into the segmentation network to obtain the first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0058] It should be noted that the segmentation network can be a network trained on a neural network model based on a training sample set. The training sample set can include original medical image samples and labels of real regions of interest within the original medical image samples, or it can include image slice samples obtained after sampling and processing the original medical image samples, and labels of real regions of interest within the image slice samples.
[0059] The region of interest can be the target area or an organ at risk. The target area corresponds to the area to be irradiated or the tumor area, while the organ at risk is a radiation-sensitive area that is not the target area. For example, the organ at risk may be located around the target area and could be the heart, brain, or prostate. The organ at risk could be the hippocampus in cranial radiotherapy. Reducing the radiation dose to the hippocampus can lessen radiation-induced neurological damage. Accurate hippocampal delineation is crucial for the benefit of neurological function and for reducing the progression of intracranial disease. Although prophylactic brain irradiation can cause adverse reactions such as neurocognitive impairment, these cognitive side effects can be reduced by using prophylactic brain irradiation that protects the hippocampus.
[0060] Traditionally, regions of interest (ROIs) are delineated manually by physicians using a radiotherapy planning system, layer by layer, on a sequence of MRI slices based on their expertise and experience. Because MRI slices are numerous, the manual delineation process is quite labor-intensive, resulting in time-consuming, labor-intensive, and inefficient methods.
[0061] The region of interest (ROI) segmentation method provided in this embodiment acquires an initial medical image of the first modality and performs interval sampling on multiple image slices to obtain an intermediate medical image. This intermediate medical image is then input into a segmentation network to obtain the first segmentation result of the ROI within the intermediate medical image. Because this embodiment obtains the intermediate medical image by performing interval sampling on multiple image slices and uses a neural network model to determine the first segmentation result of the ROI, it eliminates the need for doctors to manually delineate the ROI, thereby saving manpower and time required for ROI segmentation and improving the segmentation efficiency. Furthermore, this embodiment inputs the intermediate medical image obtained from interval sampling of multiple image slices into the segmentation network, rather than inputting all multiple image slices into the segmentation network. Therefore, it can further reduce the computational load of the segmentation network, thereby further improving the segmentation efficiency of the ROI.
[0062] Reference Figure 2 , Figure 2 This is a flowchart illustrating a method for determining the segmentation result of a region of interest in an initial medical image according to an embodiment of this application. This embodiment relates to an optional implementation of how to determine the segmentation result of a region of interest in an initial medical image based on a first segmentation result. Based on the above embodiment, the method includes the following steps:
[0063] S201. Based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain the second segmentation result of the region of interest within the other image slices.
[0064] In this process, for example, nearest neighbor interpolation or linear interpolation can be used to interpolate other image slices in the initial medical image, excluding the intermediate medical image, to obtain a second segmentation result of the region of interest within the other image slices.
[0065] For example, refer to Figure 3 , Figure 3This is a schematic diagram of a region of interest (ROI) segmentation process provided in an embodiment of this application. For example, for hundreds of thin-slice MRI images of the head, the hundreds of thin-slice MRI images are sampled at random intervals of N layers to obtain "thick-slice" MRI images, i.e., intermediate medical images. The intermediate medical images are input into a segmentation network to obtain "thick-slice" segmentation results, which refer to the first segmentation results. Then, interpolation methods are used to interpolate other image slices to obtain the second segmentation results of the ROI within other image slices, thereby obtaining the segmentation results of the ROI in each thin-slice MRI image. In this embodiment, a "thin-slice" MRI image indicates that the spacing (slice thickness) between adjacent MRI slices meets a set range, and a "thick-slice" MRI image indicates that the spacing between adjacent MRI slices is greater than the set range, i.e., the spacing between adjacent MRI slices is larger.
[0066] S202. Based on the first segmentation result and the second segmentation result, determine the segmentation result of the region of interest in the initial medical image.
[0067] In this embodiment, after obtaining the first segmentation result and the second segmentation result, the segmentation result of the region of interest of each image slice in the initial medical image is obtained.
[0068] It should be noted that in clinical applications, for thin-slice image sections, it is generally necessary to segment the regions of interest of each thin-slice image section, while for thicker-slice image sections, only the first segmentation result needs to be obtained.
[0069] In this embodiment, based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain the second segmentation result of the region of interest in the other image slices. According to the first segmentation result and the second segmentation result, the segmentation result of the region of interest in the initial medical image is determined. It is not necessary to input each image slice into the segmentation network to obtain the segmentation result of each image slice. Instead, the second segmentation result of the region of interest in the other image slices is completed by interpolation method, thereby realizing the segmentation of the region of interest in each image slice.
[0070] Optionally, step S102 above, which involves performing interval sampling on multiple image slices to obtain an intermediate medical image, can be achieved in the following way:
[0071] Multiple image slices are sampled at intervals of N slices to obtain an intermediate medical image, where N is a positive integer.
[0072] In this embodiment, multiple sets of intermediate medical images with different slice thicknesses are obtained by setting the sampling interval of image slices. For example, N can be equal to a positive integer such as 1, 2, or 3. This embodiment does not limit the specific value of N, and the value of N can be set according to clinical needs.
[0073] Reference Figure 4 , Figure 4 This is a flowchart illustrating a method for obtaining a segmentation network according to an embodiment of this application. This embodiment relates to an optional implementation of how to obtain a segmentation network. Based on the above embodiment, the method includes the following steps:
[0074] S401. Obtain the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample.
[0075] Among them, the marking of the true region of interest can be the accurate marking of the region of interest manually drawn by the doctor.
[0076] S402. Perform interval sampling on multiple image slice samples every M image slice samples to obtain a training sample set; where M is a positive integer.
[0077] During training, the value of M can be equal to or different from the value of N mentioned above. It should be noted that the value of M in each interval sampling process can be equal or different. For example, M may be 2 in the first iteration of training and 3 in the second iteration.
[0078] S403. Train the preset initial segmentation network based on the training sample set to obtain the segmentation network.
[0079] In this embodiment, by acquiring the original medical image samples and the labels of the real regions of interest within multiple image slice samples in the original medical image samples, and performing interval sampling processing on the multiple image slice samples every M image slice samples, a training sample set is obtained. Then, a preset initial segmentation network is trained based on the training sample set to obtain the segmentation network. Thus, the segmentation result of the region of interest in the image slice can be obtained based on the segmentation network, thereby improving the segmentation efficiency of the region of interest segmentation result.
[0080] Reference Figure 5 , Figure 5 This is a flowchart illustrating another method for obtaining a segmentation network provided in this application embodiment. This embodiment relates to an optional implementation of how to train a preset initial segmentation network based on a training sample set to obtain the segmentation network. Based on the above embodiment, S403 includes the following steps:
[0081] S501. Use the image slice samples in the training sample set as the input of the initial segmentation network to obtain the predicted segmentation results of the regions of interest within the image slice samples in the training sample set.
[0082] S502. Based on the predicted segmentation results and the labels of the real regions of interest within the image slice samples in the training sample set, train the initial segmentation network to obtain the segmentation network.
[0083] The computer device can calculate the error between the predicted segmentation result of the initial segmentation network and the actual labeled region of interest as the training loss of the segmentation network, and optimize the parameters of the initial segmentation network based on the training loss until the training loss is less than a certain threshold, or until the training loss no longer decreases, thereby obtaining the segmentation network.
[0084] In this embodiment, image slice samples from the training sample set are used as input to the initial segmentation network to obtain the predicted segmentation results of the regions of interest (ROIs) within the image slice samples in the training sample set. Based on the predicted segmentation results and the labels of the actual ROIs within the image slice samples in the training sample set, the initial segmentation network is trained to obtain the segmentation network. Thus, the segmentation results of the ROIs in the sampled image slices can be obtained based on the segmentation network, thereby improving the segmentation efficiency of the ROIs.
[0085] Optionally, the segmentation network includes one or more segmentation models based on Convolutional Neural Networks (CNNs), such as UNet, VNet, FCN, SegNet, and DeepLab. FCN is a semantic segmentation network based on fully convolutional networks. SegNet is a deep network proposed by Cambridge University to solve image semantic segmentation problems for autonomous driving or intelligent robots. SegNet is a semantic segmentation network based on FCN, modified from the VGG-16 network, and has two versions: SegNet and Bayesian SegNet. DeepLab combines deep convolutional neural networks and probabilistic graphical models.
[0086] Reference Figure 6 , Figure 6 This is a schematic flowchart illustrating a method for obtaining the segmentation result of a region of interest in a target image according to an embodiment of this application. This embodiment relates to an optional implementation of how to obtain the segmentation result of the region of interest in a target image. Based on the above embodiment, the method includes the following steps:
[0087] S601. Register the initial medical image with the segmentation result of the region of interest to the second modality scan image to obtain the target image.
[0088] The type of scanned image in the second modality differs from the initial medical image in the first modality. For example, if the initial medical image in the first modality is an MRI image, the scanned image in the second modality is a CT image; or if the initial medical image in the first modality is a PET image, the scanned image in the second modality is an X-ray image.
[0089] In this embodiment, the initial medical image with the determined region of interest (ROI) segmentation result is registered with a computed tomography (CT) image to obtain the target image. Since the initial medical image and the CT image are essentially fused into a single target image after registration, the ROI in the target image can be segmented based on the ROI segmentation result in the initial medical image, thus obtaining the ROI segmentation result in the target image. Because the target image incorporates the CT image, the ROI segmentation result from the CT image is obtained. The CT image reflects tissue density (electron density), which is the basis for accurate radiation dose calculation. Therefore, after obtaining the ROI segmentation result from the CT image, the radiation dose can be calculated more accurately based on the ROI segmentation result.
[0090] S602. Based on the segmentation results of the region of interest in the initial medical image, obtain the segmentation results of the region of interest in the target image.
[0091] The method provided in this embodiment obtains a target image by registering an initial medical image with the segmentation results of the region of interest determined and a second-modality scan image. Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained. Thus, based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the computed tomography scan image can be obtained. Therefore, more accurate radiodoscopy calculations can be performed based on the segmentation results of the region of interest in the computed tomography scan image, achieving better clinical application results.
[0092] In one embodiment, after S202 described above, the following steps may also be included:
[0093] It receives correction instructions from the user, and in response to the correction instructions, adjusts the segmentation results of the region of interest in the initial medical image, and outputs the adjusted segmentation results.
[0094] In this embodiment, the doctor user can modify the displayed segmentation result of interest. The computer device receives the corresponding modification instruction and, in response to the modification instruction, adjusts the segmentation result of the region of interest in the initial medical image and outputs the adjusted segmentation result, thereby ensuring that the final segmentation result is more accurate.
[0095] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0096] This application also provides a radiotherapy planning method, including:
[0097] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0098] The multiple image slices are sampled at intervals to obtain an intermediate medical image; the intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model;
[0099] The initial medical image, with the segmentation results of the region of interest identified, is registered with the scanned image of the second modality to obtain the target image;
[0100] Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained;
[0101] A radiotherapy plan for the subject is formulated based on the segmentation results of the region of interest in the target image.
[0102] The technical solution of this invention is applicable to intensity-modulated radiotherapy (IMRT) systems, tomotherapy (TOMO) systems, and volumetric modulated arc therapy (VMAT) systems.
[0103] In one embodiment, the radiotherapy system can acquire the patient's CT and MR data, and determine the segmentation result of the region of interest in the target image based on the two modalities of data (in this embodiment, the hippocampus, a critical organ). Further, based on the density distribution image converted from the CT data, the radiotherapy plan is determined by combining the density distribution image and the segmentation result of the region of interest.
[0104] The patient undergoing radiotherapy does not need to move on the treatment bed and can be directly transported to the position to receive radiotherapy. The radiotherapy equipment executes the radiotherapy plan, and during the execution, the machine record information of the radiotherapy equipment is acquired. The machine record information includes: the gantry angle information of the radiotherapy equipment (usually an accelerator), the position information of the inner end face of the multi-leaf collimator, the position information of the inner end face of the secondary collimator, the generation time information of the machine record information, and the dose reading of the dose monitoring device, etc.
[0105] In one embodiment, the following steps may also be performed:
[0106] The radiation field intensity distribution data is reconstructed based on the machine-recorded information; the actual radiation dose distribution is calculated based on the radiation field intensity distribution data; and the dose corresponding to the region of interest in the target image is determined to be within a set threshold range based on the actual radiation dose distribution.
[0107] Based on the same inventive concept, this application also provides a region-of-interest (ROI) segmentation apparatus for implementing the above-described region-of-interest segmentation method. The solution provided by this apparatus is similar to the implementation described in the above-described method; therefore, the specific limitations in one or more ROI segmentation apparatus embodiments provided below can be found in the limitations of the region-of-interest segmentation method described above, and will not be repeated here.
[0108] In one embodiment, refer to Figure 7 , Figure 7 This is a schematic diagram of a region of interest segmentation device provided in an embodiment of this application. The device 700 includes: a first acquisition module 701, a sampling processing module 702, and a segmentation module 703.
[0109] The first acquisition module 701 is used to acquire an initial medical image of a first modality, wherein the initial medical image includes multiple image slices;
[0110] The first sampling processing module 702 is used to perform interval sampling processing on the multiple image slices to obtain an intermediate medical image;
[0111] The segmentation module 703 is used to input the intermediate medical image into the segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein the segmentation network is obtained by training a neural network model.
[0112] The region-of-interest (ROI) segmentation device provided in this embodiment acquires an initial medical image of a first modality and performs interval sampling on multiple image slices to obtain an intermediate medical image. This intermediate medical image is then input into a segmentation network to obtain the first segmentation result of the RIO within the intermediate medical image. Because this embodiment obtains the intermediate medical image by performing interval sampling on multiple image slices and uses a neural network model to determine the first segmentation result of the RIO, it eliminates the need for doctors to manually delineate the RIO, thereby saving manpower and time and improving the segmentation efficiency of the RIO. Furthermore, this embodiment inputs the intermediate medical image obtained from interval sampling of multiple image slices into the segmentation network, rather than inputting all multiple image slices into the segmentation network. Therefore, it can further reduce the computational load of the segmentation network, thereby further improving the segmentation efficiency of the RIO.
[0113] In one embodiment, the device 700 may further include:
[0114] An interpolation processing module is used to perform interpolation processing on other image slices in the initial medical image, excluding the intermediate medical image, based on the first segmentation result, to obtain a second segmentation result of the region of interest within the other image slices;
[0115] The determination module is used to determine the segmentation result of the region of interest in the initial medical image based on the first segmentation result and the second segmentation result.
[0116] In one embodiment, the first sampling processing module 702 is specifically used to sample the plurality of image slices every N image slices to obtain the intermediate medical image, wherein N is a positive integer.
[0117] In one embodiment, the device 700 may further include:
[0118] The second acquisition module is used to acquire the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample;
[0119] The second sampling processing module is used to perform interval sampling processing on the plurality of image slice samples every M image slice samples to obtain the training sample set; wherein, M is a positive integer;
[0120] The training module is used to train a preset initial segmentation network based on the training sample set to obtain the segmentation network.
[0121] In one embodiment, the training module is specifically used to take image slice samples from the training sample set as input to the initial segmentation network to obtain a predicted segmentation result of the region of interest within the image slice samples in the training sample set; and to train the initial segmentation network based on the predicted segmentation result and the label of the real region of interest within the image slice samples in the training sample set to obtain the segmentation network.
[0122] In one embodiment, the segmentation network includes one or more segmentation models based on convolutional neural networks (CNNs), such as UNet, VNet, FCN, SegNet, and DeepLab.
[0123] In one embodiment, the device 700 may further include:
[0124] The first acquisition module is used to register the initial medical image with the segmentation results of the region of interest determined with the scan image of the second modality to obtain the target image;
[0125] The second obtaining module is used to obtain the segmentation result of the region of interest in the target image based on the segmentation result of the region of interest in the initial medical image.
[0126] In one embodiment, the device 700 may further include:
[0127] The adjustment module is used to receive correction instructions from the user, and in response to the correction instructions, adjust the segmentation results of the region of interest in the initial medical image, and output the adjusted segmentation results.
[0128] Each module in the aforementioned region of interest segmentation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the operations corresponding to each module.
[0129] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, Figure 8This is an internal structural diagram of a computer device in one embodiment. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores image data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a region-of-interest (ROI) segmentation method.
[0130] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0131] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0132] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0133] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0134] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0135] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0136] Based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain a second segmentation result of the region of interest within the other image slices;
[0137] Based on the first segmentation result and the second segmentation result, the segmentation result of the region of interest in the initial medical image is determined.
[0138] Specifically, based on the grayscale value of the first segmentation result of the region of interest in the intermediate medical image, interpolation processing can be performed on other image slices in the initial medical image other than the intermediate medical image to obtain the second segmentation result of the region of interest in the other image slices.
[0139] For example, based on the grayscale value of the first segmentation result of the region of interest in two adjacent intermediate medical images, interpolation processing can be performed on other image slices located between the two intermediate medical images to obtain the second segmentation result of the region of interest in the other image slices.
[0140] For example, based on the grayscale value of the first segmentation result of the region of interest in the intermediate medical image closest to another image slice, interpolation processing can be performed on that other image slice to obtain the second segmentation result of the region of interest within that other image slice.
[0141] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0142] The intermediate medical image is obtained by sampling the plurality of image slices at intervals of N slices, where N is a positive integer.
[0143] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0144] Obtain the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample;
[0145] The training sample set is obtained by performing interval sampling on the plurality of image slice samples every M image slice samples; wherein M is a positive integer;
[0146] The initial segmentation network is trained based on the training sample set to obtain the segmentation network.
[0147] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0148] Using image slice samples from the training sample set as input to the initial segmentation network, the predicted segmentation result of the region of interest within the image slice samples from the training sample set is obtained;
[0149] Based on the predicted segmentation results and the labels of the true regions of interest within the image slice samples in the training sample set, the initial segmentation network is trained to obtain the segmentation network.
[0150] In one embodiment, the segmentation network includes one or more segmentation models based on Convolutional Neural Networks (CNNs), such as UNet, VNet, FCN, SegNet, and DeepLab.
[0151] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0152] The initial medical image, with the segmentation results of the region of interest identified, is registered with the scanned image of the second modality to obtain the target image;
[0153] Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained.
[0154] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0155] The system receives correction instructions from the user, and in response to the correction instructions, adjusts the segmentation results of the region of interest in the initial medical image, and outputs the adjusted segmentation results.
[0156] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0157] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0158] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0159] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0160] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0161] Based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain a second segmentation result of the region of interest within the other image slices;
[0162] Based on the first segmentation result and the second segmentation result, the segmentation result of the region of interest in the initial medical image is determined.
[0163] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0164] The intermediate medical image is obtained by sampling the plurality of image slices at intervals of N slices, where N is a positive integer.
[0165] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0166] Obtain the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample;
[0167] The training sample set is obtained by performing interval sampling on the plurality of image slice samples every M image slice samples; wherein M is a positive integer;
[0168] The initial segmentation network is trained based on the training sample set to obtain the segmentation network.
[0169] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0170] Using image slice samples from the training sample set as input to the initial segmentation network, the predicted segmentation result of the region of interest within the image slice samples from the training sample set is obtained;
[0171] Based on the predicted segmentation results and the labels of the true regions of interest within the image slice samples in the training sample set, the initial segmentation network is trained to obtain the segmentation network.
[0172] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0173] The segmentation network includes one or more segmentation models based on Convolutional Neural Networks (CNNs), such as UNet, VNet, FCN, SegNet, and DeepLab.
[0174] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0175] The initial medical image, with the segmentation results of the region of interest identified, is registered with the scanned image of the second modality to obtain the target image;
[0176] Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained.
[0177] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0178] The system receives correction instructions from the user, and in response to the correction instructions, adjusts the segmentation results of the region of interest in the initial medical image, and outputs the adjusted segmentation results.
[0179] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0180] Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices;
[0181] The multiple image slices are sampled at intervals to obtain an intermediate medical image;
[0182] The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model.
[0183] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0184] Based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain a second segmentation result of the region of interest within the other image slices;
[0185] Based on the first segmentation result and the second segmentation result, the segmentation result of the region of interest in the initial medical image is determined.
[0186] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0187] The intermediate medical image is obtained by sampling the plurality of image slices at intervals of N slices, where N is a positive integer.
[0188] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0189] Obtain the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample;
[0190] The training sample set is obtained by performing interval sampling on the plurality of image slice samples every M image slice samples; wherein M is a positive integer;
[0191] The initial segmentation network is trained based on the training sample set to obtain the segmentation network.
[0192] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0193] Using image slice samples from the training sample set as input to the initial segmentation network, the predicted segmentation result of the region of interest within the image slice samples from the training sample set is obtained;
[0194] Based on the predicted segmentation results and the labels of the true regions of interest within the image slice samples in the training sample set, the initial segmentation network is trained to obtain the segmentation network.
[0195] In one embodiment, the segmentation network includes one or more segmentation models based on Convolutional Neural Networks (CNNs), such as UNet, VNet, FCN, SegNet, and DeepLab.
[0196] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0197] The initial medical image, with the segmentation results of the region of interest identified, is registered with the scanned image of the second modality to obtain the target image;
[0198] Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained.
[0199] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:
[0200] The system receives correction instructions from the user, and in response to the correction instructions, adjusts the segmentation results of the region of interest in the initial medical image, and outputs the adjusted segmentation results.
[0201] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.
[0202] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0203] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0204] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for segmenting a region of interest, characterized in that, The method includes: Acquire an initial medical image of the first modality, wherein the initial medical image comprises multiple image slices; The multiple image slices are sampled at intervals to obtain an intermediate medical image; The intermediate medical image is input into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model; Based on the first segmentation result, interpolation processing is performed on other image slices in the initial medical image, excluding the intermediate medical image, to obtain a second segmentation result of the region of interest within the other image slices; Based on the first segmentation result and the second segmentation result, the segmentation result of the region of interest in the initial medical image is determined.
2. The method according to claim 1, characterized in that, The step of performing interval sampling on the multiple image slices to obtain an intermediate medical image includes: The intermediate medical image is obtained by sampling the plurality of image slices at intervals of N slices, where N is a positive integer.
3. The method according to claim 1, characterized in that, The method further includes: Obtain the original medical image sample and the labels of the true regions of interest within multiple image slice samples in the original medical image sample; The training sample set is obtained by performing interval sampling on the plurality of image slice samples every M image slice samples; wherein M is a positive integer; The initial segmentation network is trained based on the training sample set to obtain the segmentation network.
4. The method according to claim 3, characterized in that, The process of training a preset initial segmentation network based on the training sample set to obtain the segmentation network includes: Using image slice samples from the training sample set as input to the initial segmentation network, the predicted segmentation result of the region of interest within the image slice samples from the training sample set is obtained; Based on the predicted segmentation results and the labels of the true regions of interest within the image slice samples in the training sample set, the initial segmentation network is trained to obtain the segmentation network.
5. The method according to claim 1 or 2, characterized in that, After determining the segmentation result of the region of interest in the initial medical image, the method further includes: The initial medical image, with the segmentation results of the region of interest identified, is registered with the scanned image of the second modality to obtain the target image; Based on the segmentation results of the region of interest in the initial medical image, the segmentation results of the region of interest in the target image are obtained.
6. The method according to claim 1 or 2, characterized in that, After determining the segmentation result of the region of interest in the initial medical image, the method further includes: The system receives correction instructions from the user, and in response to the correction instructions, adjusts the segmentation results of the region of interest in the initial medical image, and outputs the adjusted segmentation results.
7. A region of interest segmentation device, characterized in that, The device includes: An acquisition module is used to acquire an initial medical image of a first modality, wherein the initial medical image includes multiple image slices; The sampling processing module is used to perform interval sampling processing on the multiple image slices to obtain an intermediate medical image; A segmentation module is used to input the intermediate medical image into a segmentation network to obtain a first segmentation result of the region of interest in the intermediate medical image; wherein, the segmentation network is obtained by training a neural network model; An interpolation processing module is used to perform interpolation processing on other image slices in the initial medical image, excluding the intermediate medical image, based on the first segmentation result, to obtain a second segmentation result of the region of interest within the other image slices; The determination module is used to determine the segmentation result of the region of interest in the initial medical image based on the first segmentation result and the second segmentation result.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, 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 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.