Medical image segmentation method and computer readable storage medium

Abstract

The present application relates to a medical image segmentation method and a computer-readable storage medium, the method comprising: obtaining each organ image in the image to be segmented; inputting each organ image into the corresponding organ abnormal point segmentation model respectively, and obtaining the abnormal point segmentation result corresponding to each organ image; wherein the organ abnormal point segmentation model comprises a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network comprises at least one output result of the previous segmentation network; according to the abnormal point segmentation result corresponding to each organ image, the whole body abnormal point segmentation result corresponding to the image to be segmented is obtained. The accuracy of the whole body abnormal point segmentation result obtained in this method can be greatly improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a medical image segmentation method and a computer-readable storage medium. Background Art

[0002] As a nuclear medicine imaging technology, PET-CT plays a vital role in the detection, discovery, staging, treatment and follow-up of early lesions or abnormalities by increasing the accuracy of anatomical positioning in functional imaging. When detecting abnormalities in PET-CT images, image segmentation methods are usually used to segment the abnormal point areas in the PET-CT images to assist doctors in the subsequent diagnosis process.

[0003] In the early days, doctors generally only focused on the local abnormal points of patients, so they could only segment the local abnormal point areas in the PET-CT images; currently, in order to enable doctors to have a more comprehensive understanding of the health status of patients, the need for segmentation of the whole body abnormal point areas in PET-CT images has been proposed. In traditional technology, the whole body abnormal point area segmentation model is used to segment the abnormal point areas of the PET-CT whole body image of the subject.

[0004] However, the accuracy of the segmentation results of the abnormal point area segmented by the whole body abnormal point area segmentation model based on traditional technology is low. Summary of the invention

[0005] Based on this, it is necessary to provide a medical image segmentation method and a computer-readable storage medium to address the problem of low accuracy of segmentation results when performing whole-body abnormal point region segmentation in traditional technologies.

[0006] A medical image segmentation method, the method comprising:

[0007] Acquire images of various organs in the image to be segmented;

[0008] Inputting each organ image into the corresponding organ abnormality segmentation model respectively to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network;

[0009] According to the abnormal point segmentation results corresponding to each organ image, the whole body abnormal point segmentation results corresponding to the image to be segmented are obtained.

[0010] In one of the embodiments, the organ abnormality segmentation model includes a first segmentation network and a second segmentation network, each segmentation network includes a plurality of decoders;

[0011] Input each organ image into the corresponding organ abnormality segmentation model to obtain the abnormality segmentation results corresponding to each organ image, including:

[0012] For each organ image, the organ image is input into a first segmentation network of a corresponding organ abnormality segmentation model, and multiple decoders of the first segmentation network output multiple reference segmentation results; wherein one decoder outputs one reference segmentation result;

[0013] The organ image and a target segmentation result from a plurality of reference segmentation results are input into a second segmentation network of the organ abnormality segmentation model, and the abnormality segmentation result corresponding to the organ image is output through the last decoder from among the plurality of decoders of the second segmentation network; wherein the target segmentation result is the reference segmentation result output by the last N decoders from among the plurality of decoders of the first segmentation network, and N≥1.

[0014] In one embodiment, the segmentation network further includes encoders having the same number as the decoders; and the second segmentation network inputting the organ image and the target segmentation result from the plurality of reference segmentation results into the organ abnormality segmentation model comprises:

[0015] The organ image and the last target segmentation result are input into the first encoder of the second segmentation network, and the nth to last target segmentation result is input into the nth encoder of the second segmentation network; wherein n≤N.

[0016] In one embodiment, each organ image is input into the corresponding organ abnormality segmentation model to obtain the abnormality segmentation result corresponding to each organ image, including:

[0017] For each organ image, performing a first conversion operation on the organ image to obtain a conversion image corresponding to the organ image;

[0018] Inputting the organ image into the organ abnormality point segmentation model to obtain a first segmentation result corresponding to the organ image; inputting the transformed image into the organ abnormality point segmentation model to obtain a second segmentation result corresponding to the transformed image;

[0019] A second conversion operation is performed on the second segmentation result to obtain a third segmentation result, and the first segmentation result and the third segmentation result are used as abnormal point segmentation results corresponding to the organ image; wherein the second conversion operation is an inverse operation of the first conversion operation.

[0020] In one embodiment, according to the abnormal point segmentation results corresponding to each organ image, the whole body abnormal point segmentation results corresponding to the image to be segmented are obtained, including:

[0021] Merging the first segmentation results corresponding to the organ images to obtain a first merged result;

[0022] Merging the third segmentation results corresponding to the organ images to obtain a second merged result;

[0023] According to the first merging result and the second merging result, a whole body abnormal point segmentation result corresponding to the image to be segmented is obtained.

[0024] In one embodiment, obtaining a whole body abnormal point segmentation result corresponding to the image to be segmented according to the first merging result and the second merging result includes:

[0025] The first merging result and the second merging result are channel-connected and input into the whole body abnormal point segmentation model to obtain the whole body abnormal point segmentation result.

[0026] In one embodiment, obtaining each organ image in the image to be segmented includes:

[0027] Input the images to be segmented into the organ segmentation models of the organs respectively to obtain the organ segmentation results of the organs;

[0028] For each organ, an organ image corresponding to the organ is obtained from the image to be segmented according to the organ segmentation result of the organ.

[0029] In one embodiment, the method further comprises:

[0030] According to the results of the segmentation of abnormal points throughout the body, the maximum standard uptake value, major and minor axes, and volume of each abnormal point area are determined;

[0031] The whole body abnormal point segmentation result, at least one of the maximum standard uptake value, the major and minor axes and the volume of the abnormal point area are displayed to the user terminal.

[0032] In one embodiment, the method further comprises:

[0033] Receiving an outlier point deletion instruction and / or an outlier point addition instruction input by a user through a user terminal;

[0034] According to the abnormal point deletion instruction and / or abnormal point addition instruction, the whole body abnormal point segmentation result is updated.

[0035] In one embodiment, the method further comprises:

[0036] Obtaining multiple whole body abnormal point segmentation results corresponding to the same subject within a preset time period;

[0037] The segmentation results of multiple abnormal points of the whole body are compared to determine the degree of change of the abnormal point area of ​​the subject within the time period.

[0038] In one embodiment, each organ image is input into the corresponding organ abnormality segmentation model to obtain the abnormality segmentation result corresponding to each organ image, including:

[0039] If the organ type corresponding to each organ image belongs to an organ category set, each organ image is input into the corresponding organ abnormality point segmentation model to obtain the abnormality point segmentation result corresponding to each organ image; wherein the organ category set includes other organ types except the brain, kidney, bladder and heart.

[0040] A medical image segmentation device, the device comprising:

[0041] An acquisition module, used for acquiring images of various organs in the image to be segmented;

[0042] A segmentation module, used to input each organ image into a corresponding organ abnormality segmentation model to obtain an abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of a subsequent segmentation network includes at least one output result of a previous segmentation network;

[0043] The determination module is used to obtain the whole body abnormal point segmentation result corresponding to the image to be segmented according to the abnormal point segmentation result corresponding to each organ image.

[0044] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0045] Acquire images of various organs in the image to be segmented;

[0046] Inputting each organ image into the corresponding organ abnormality segmentation model respectively to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network;

[0047] According to the abnormal point segmentation results corresponding to each organ image, the whole body abnormal point segmentation results corresponding to the image to be segmented are obtained.

[0048] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0049] Acquire images of various organs in the image to be segmented;

[0050] Inputting each organ image into the corresponding organ abnormality segmentation model respectively to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network;

[0051] According to the abnormal point segmentation results corresponding to each organ image, the whole body abnormal point segmentation results corresponding to the image to be segmented are obtained.

[0052] The above-mentioned medical image segmentation method, device, computer equipment and storage medium can input the acquired organ images into the corresponding organ abnormality segmentation model respectively, obtain the abnormality segmentation results corresponding to each organ image, and then obtain the whole body abnormality segmentation results corresponding to the image to be segmented according to the abnormality segmentation results corresponding to each organ image; wherein, the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network. In this method, by decomposing the whole body abnormality segmentation process into the segmentation process of each organ abnormality, so as to convert the whole body area segmentation into the segmentation of each small area, the calculation amount of the organ abnormality segmentation model is reduced, and the accuracy of the abnormality segmentation results is improved, and the automatic segmentation of the whole body abnormalities is realized, and the statistical efficiency of the clinical imaging work is improved; and the organ abnormality segmentation model adopted includes a cascade structure of multiple segmentation networks, and the former segmentation network provides multi-scale segmentation semantic information for the latter segmentation network, further improving the accuracy of the organ abnormality segmentation results, and thus improving the accuracy of the obtained whole body abnormality segmentation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is an internal structure diagram of a computer device in one embodiment;

[0054] Figure 2 is a schematic flow chart of a medical image segmentation method in one embodiment;

[0055] Figure 3 is a flow chart of a medical image segmentation method in another embodiment;

[0056] Figure 3a A schematic diagram of the structure of an organ abnormality segmentation model in one embodiment;

[0057] Figure 4 A flowchart of a training method for an organ abnormality segmentation model in one embodiment;

[0058] Figure 5 is a flowchart of a medical image segmentation method in yet another embodiment;

[0059] Figure 6 is a flowchart of a medical image segmentation method in yet another embodiment;

[0060] Figure 7 is a flowchart of a medical image segmentation method in yet another embodiment;

[0061] Figure 8 FIG. 4 is a structural block diagram of a medical image segmentation device in one embodiment. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] The medical image segmentation method provided in the embodiment of the present application can be applied to Figure 1 The computer device shown. The computer device includes a processor and a memory connected by a system bus, and a computer program is stored in the memory. When the processor executes the computer program, the steps of the following method embodiment can be executed. Optionally, the computer device may also include a communication interface, a display screen and an input device. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory, and the non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal through a network connection. Optionally, the computer device can be a personal computer (personal computer, referred to as PC), a personal digital assistant, or other terminal devices, such as a tablet computer (portable android device, referred to as PAD), a mobile phone, etc., or a cloud or remote server. The embodiment of the present application does not limit the specific form of the computer device.

[0064] In one embodiment, Figure 2 As shown, a medical image segmentation method is provided, and the method is applied to Figure 1 The computer device in the example is used to illustrate that the present embodiment relates to a specific process of the computer device performing whole body abnormal point segmentation on the image to be segmented. The method comprises the following steps:

[0065] S101, acquiring images of various organs in the image to be segmented.

[0066] The image to be segmented may be a PCT image and a CT image obtained by a PCT-CT device performing a full-body scan on the subject, and the image to be segmented includes various organ regions. The computer device may distinguish the image to be segmented according to the pixel values ​​of the image to be segmented to divide the various organ regions, and intercept the various organ regions to obtain various organ images. Optionally, the computer device may also receive the organ regions marked by the user (such as a doctor) on the image to be segmented to obtain the organ image. In addition, for each organ image, its organ type may also be marked, such as heart, kidney, pancreas, and other organ types.

[0067] S102, inputting each organ image into the corresponding organ abnormality segmentation model to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network.

[0068] Specifically, the computer device can input the above-mentioned organ images into the corresponding organ abnormality point segmentation model respectively, such as inputting the pancreatic image into the pancreatic abnormality point segmentation model, the lung image into the lung abnormality point segmentation model, etc., to obtain the abnormality point segmentation results corresponding to each organ image; wherein the abnormal point can be a lesion point (suspected lesion point) or an abnormally high metabolic area on the organ, etc., and the abnormal point segmentation result can be an abnormal point mask or an abnormal point probability map.

[0069] Among them, the above-mentioned organ abnormal point segmentation model may include a cascade structure of multiple segmentation networks, and the segmentation network may be a VNet network, a UNet network, etc., and the input data of the latter segmentation network includes at least one output result of the previous segmentation network, that is, the output result of the previous segmentation network is used as the input of the latter segmentation network. Exemplarily, the organ abnormal point segmentation model is obtained by cascading two or more UNet networks, and the previous segmentation network can obtain multi-scale feature semantic information of the abnormal point area through the processing of each network layer, and input the multi-scale feature semantic information into the latter segmentation network to guide the latter segmentation network to learn from the analysis, that is, the high-level multi-scale semantic information of the previous segmentation network supervises and guides the latter segmentation network to perform semantic segmentation, thereby improving the segmentation accuracy of the organ abnormal point segmentation model.

[0070] Optionally, the computer device may first resample the organ image to the same resolution space as the sample image in the training phase of the organ abnormality segmentation model, and standardize the resampled organ image, and then input the standardized organ image into the organ abnormality segmentation model. The resampled organ image is standardized by the relationship of , where I is the resampled organ image, μ is the mean of the sample images in the training phase of the organ abnormality segmentation model, and σ is the standard deviation of the sample images in the training phase of the organ abnormality segmentation model. In addition, for the PET image and CT image of the same organ, the computer device can connect the standardized PET image and CT image along the channel dimension and input them into the organ abnormality segmentation model.

[0071] Optionally, since organs such as the brain, kidneys, bladder, and heart are high-uptake areas in PET images and the abnormal points are not very obvious, in order to avoid the influence of such organs on the segmentation results of the high-uptake abnormal points, the computer device can remove the organ images corresponding to such organs. If the organ type corresponding to each organ image belongs to the organ category set, then each organ image is input into the corresponding organ abnormal point segmentation model; if the organ type corresponding to each organ image does not belong to the organ category set, then the organ image is not segmented; wherein, the organ category set includes other organ types except the brain, kidneys, bladder, and heart.

[0072] S103, obtaining a whole body abnormal point segmentation result corresponding to the image to be segmented according to the abnormal point segmentation results corresponding to each organ image.

[0073] Specifically, the computer device can combine (or merge) the abnormal point segmentation results corresponding to each organ image to obtain the whole body abnormal point segmentation result corresponding to the image to be segmented. Optionally, because the organ image is obtained from the image to be segmented, the abnormal point segmentation results can be spliced ​​and merged into the whole body abnormal point segmentation result.

[0074] Optionally, after obtaining the whole body abnormal point segmentation results, the whole body abnormal point regions can also be displayed in a highlighted form on the original image to be segmented to assist doctors in viewing and diagnosing each abnormal point region.

[0075] In the medical image segmentation method provided in this embodiment, the computer device inputs the acquired organ images into the corresponding organ abnormality segmentation model respectively, obtains the abnormality segmentation results corresponding to the organ images, and then obtains the whole body abnormality segmentation results corresponding to the image to be segmented according to the abnormality segmentation results corresponding to the organ images; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network. In this method, by decomposing the whole body abnormality segmentation process into the segmentation process of each organ abnormality, the whole body region segmentation is converted into the segmentation of each small region, which reduces the calculation amount of the organ abnormality segmentation model, improves the accuracy of the abnormality segmentation results, realizes the automatic segmentation of the whole body abnormalities, and improves the statistical efficiency of the clinical imaging work; and the organ abnormality segmentation model used includes a cascade structure of multiple segmentation networks, and the former segmentation network provides multi-scale segmentation semantic information for the latter segmentation network, further improves the accuracy of the organ abnormality segmentation results, and thus improves the accuracy of the obtained whole body abnormality segmentation results.

[0076] In one embodiment, the organ abnormality segmentation model includes a first segmentation network and a second segmentation network, each of which includes multiple decoders; it should be noted that the first segmentation network and the second segmentation network here are only used to illustrate different segmentation networks, and the number of segmentation networks is not limited. Figure 3 As shown, the above S102 may include:

[0077] S201, for each organ image, input the organ image into the first segmentation network of the corresponding organ abnormality segmentation model, and output multiple reference segmentation results through multiple decoders of the first segmentation network; wherein one decoder outputs one reference segmentation result.

[0078] Specifically, for multiple decoders, a subsequent decoder is used to perform an upsampling operation on the output result of a previous decoder, and each decoder outputs an upsampling result as a reference segmentation result. After the computer device inputs the organ image into the first segmentation network of the organ abnormality segmentation model, multiple decoders of the first segmentation network can output multiple reference segmentation results, and the resolution scales of the multiple reference segmentation results are different.

[0079] S202, input the organ image and the target segmentation result among the multiple reference segmentation results into the second segmentation network of the organ abnormality segmentation model, and output the abnormality segmentation result corresponding to the organ image through the last decoder among the multiple decoders of the second segmentation network; wherein the target segmentation result is the reference segmentation result output by the last N decoders among the multiple decoders of the first segmentation network, N≥1.

[0080] Specifically, the computer device can select a target segmentation result from the above-mentioned multiple reference segmentation results, and the target segmentation result is the reference segmentation result output by the last N decoders among the multiple decoders of the above-mentioned first segmentation network, that is, the last N reference segmentation results. Then, the original organ image and the target segmentation result are input into the second segmentation network together, so that the second segmentation network refers to the output result of the first segmentation network and outputs the final abnormal point segmentation result through the last decoder among the multiple decoders of the second segmentation network; the first segmentation network provides high-level semantic information to the second segmentation network, and guides the second segmentation network to perform feature extraction and semantic analysis, so as to improve the accuracy of the abnormal point segmentation result output by the second segmentation network.

[0081] Optionally, the segmentation network further includes encoders having the same number as the decoders, and the latter encoder is used to downsample the output result of the former encoder. The computer device can input the organ image and the target segmentation result into the first encoder of the second segmentation network, and the downsampled result of the first encoder into the second encoder, and so on. Optionally, the computer device can also input the organ image and the target segmentation result into the encoders at the corresponding positions; Figure 3a The figure shows a structural diagram of the organ abnormal point segmentation model. The computer device inputs the organ image and the last target segmentation result into the first decoder of the second segmentation network, and inputs the nth (n≤N) target segmentation result from the end into the nth encoder of the second segmentation network, that is, each target segmentation result is input into an encoder adapted to its scale to improve the accuracy of the output results of each encoder. Optionally, the first n target segmentation results can also be input into the nth encoder together, so that the nth encoder can perform segmentation processing based on the target segmentation results of different sizes.

[0082] Optionally, continue as Figure 3a As shown in the figure, between the encoder and decoder at the corresponding position of each segmentation network, a dilated block with different expansion coefficients is set, which can greatly ensure the receptive field of the deep network while extracting multi-scale information, and to a certain extent avoid the loss of tiny information caused by downsampling operation, thereby improving the accuracy of the outlier segmentation results.

[0083] In the medical image segmentation method provided in this embodiment, for each organ image, a computer device inputs the organ image into a first segmentation network of a corresponding organ abnormality segmentation model, and outputs multiple reference segmentation results through multiple decoders of the first segmentation network, wherein one decoder outputs a reference segmentation result; then the organ image and a target segmentation result among the multiple reference segmentation results are input into a second segmentation network of the organ abnormality segmentation model, and the last decoder among the multiple decoders of the second segmentation network outputs an abnormality segmentation result corresponding to the organ image; thereby, the first segmentation network provides supervision and guidance information to the second segmentation network, thereby improving the accuracy of the abnormality segmentation results output by the second segmentation network, thereby improving the accuracy of subsequent whole body abnormality segmentation results.

[0084] In one embodiment, Figure 3a Based on the model structure shown in , the training process of the model is also involved. Figure 4 As shown, the training method of the above-mentioned organ abnormality segmentation model includes:

[0085] S301, obtaining a sample image and a gold standard for outlier segmentation corresponding to the sample image.

[0086] Specifically, a plurality of PET images and CT images containing abnormal points can be scanned by the PET-CT device, and a plurality of organ images can be obtained as sample images from the PCT image and the CT image, and the abnormal point segmentation gold standard (mask) of each sample image can be annotated by an experienced doctor. Optionally, the computer device can also resample all sample images to the same resolution space, and set appropriate mean and standard deviation to standardize the resampled sample images. The standardization process can refer to the description of the above embodiment, which will not be repeated here.

[0087] S302, input the sample image into the first segmentation network of the initial organ abnormality segmentation network, and output multiple first predicted segmentation results through multiple decoders of the first segmentation network; input the sample image and the first target segmentation result among the multiple first predicted segmentation results into the second segmentation network of the initial organ abnormality segmentation network, and output multiple second predicted segmentation results through multiple decoders of the second segmentation network.

[0088] The processing process of the first segmentation network and the second segmentation network in the initial organ abnormality segmentation network in this step can refer to the description of the above embodiment, and the implementation principle is similar; it should be noted that the initial organ abnormality segmentation network in this step is a network in the training process.

[0089] S303, calculating a first loss between a first target segmentation result among multiple first predicted segmentation results and a gold standard for outlier segmentation, and a second loss between a second target segmentation result among multiple second predicted segmentation results and a gold standard for outlier segmentation; training an initial organ outlier segmentation network according to the first loss and the second loss to obtain an organ outlier segmentation model.

[0090] Among them, the first target segmentation result is the predicted segmentation result output by the last N decoders among the multiple decoders of the first segmentation network, and the second target segmentation result is the predicted segmentation result output by the last N decoders among the multiple decoders of the second segmentation network. The computer device can respectively calculate the first loss between the first target segmentation result and the outlier segmentation gold standard, and the second loss between the second target segmentation result and the outlier segmentation gold standard. Exemplarily, the losses Loss1, Loss2 and Loss3 between the last three predicted segmentation results of the first segmentation network (i.e., the first target segmentation result) and the outlier segmentation gold standard, and the losses Loss4, Loss5 and Loss6 between the last three predicted segmentation results of the second segmentation network (i.e., the second target segmentation result) and the outlier segmentation gold standard, can be calculated. Optionally, each loss can also be composed of two parts: BCE loss and Dice loss, wherein, y is the gold standard for outlier segmentation, To predict the segmentation results; X is the predicted segmentation result, and Y is the gold standard for abnormal point segmentation. Then, the initial organ abnormal point segmentation network is trained according to the first loss and the second loss to obtain the organ abnormal point segmentation model; optionally, the initial organ abnormal point segmentation network can be trained according to the direct summation or weighted summation result of the first loss and the second loss.

[0091] Optionally, for multiple sample images, the computer device can randomly select image patches from them, and the size of the patches can be [64, 128, 128]. Since the network can be connected along the channel dimension to input the PET image and the CT image, the input data size can be [2, 64, 128, 128]. In addition, the optimizer during the initial organ abnormality segmentation network training can be Adam.

[0092] In the medical image segmentation method provided in this embodiment, a computer device trains an initial organ abnormality segmentation network according to a sample image and a gold standard for abnormality segmentation corresponding to the sample image, so as to continuously improve the accuracy of the initial organ abnormality segmentation network, and finally obtain an organ abnormality segmentation model with higher accuracy; and then the organ abnormality segmentation model is used to segment abnormalities, thereby greatly improving the accuracy of abnormality segmentation results.

[0093] In one embodiment, the computer device may also perform data enhancement on the organ image input to the organ abnormality segmentation model, so that the organ abnormality segmentation model considers more data information when performing segmentation, thereby improving segmentation accuracy. Figure 5 As shown, the above S102 may include:

[0094] S401 : For each organ image, perform a first conversion operation on the organ image to obtain a conversion image corresponding to the organ image.

[0095] The first conversion operation may be a flipping operation on the organ image, for example, the organ image may be flipped along the Z direction, along the Y direction, and along the X direction to obtain a conversion image I corresponding to the organ image I. Zflip ,I Yflip and I Xflip It can also be a scaling operation on the organ image, such as shrinking the organ image toward the center of the image, or enlarging it in the opposite direction of the image center, so as to enhance the data of the organ image.

[0096] S402, inputting the organ image into the organ abnormality segmentation model to obtain a first segmentation result corresponding to the organ image; inputting the transformed image into the organ abnormality segmentation model to obtain a second segmentation result corresponding to the transformed image.

[0097] Specifically, the computer device inputs the organ image into the organ abnormality segmentation model to obtain a first segmentation result I corresponding to the organ image I. mask , and the converted image (such as I Zflip ,I Yflip ,I Xflip ) Input the organ abnormality segmentation model to obtain the second segmentation result (I mask-Zflip ,I mask-Yflip ,I mask-Xflip ).

[0098] S403, performing a second conversion operation on the second segmentation result to obtain a third segmentation result, and using the first segmentation result and the third segmentation result as abnormal point segmentation results corresponding to the organ image; wherein the second conversion operation is an inverse operation of the first conversion operation.

[0099] Specifically, after obtaining the second segmentation result, the computer device may perform a second conversion operation on the second segmentation result. The second conversion operation is an inverse operation of the first conversion operation. For example, if the first conversion operation is to flip the organ image along the Z direction, the second conversion operation is to flip the organ image along the Z direction. mask-Zflip If the first conversion operation is to flip the organ image along the Y direction, the second conversion operation is to flip I mask-YflipIf the first conversion operation is to flip the organ image along the X direction, the second conversion operation is to flip I mask-Xflip Flip along the X direction to get the third segmentation result (I A-mask-Zflip ,I A-mask-Yflip ,I A-mask-Xflip ), that is, the third segmentation result is converted back to the same image direction as the first segmentation result, and the first segmentation result and the third segmentation result are used as the abnormal point segmentation results corresponding to the organ image. Alternatively, if the first conversion operation is a reduction operation on the organ image, the second conversion operation is an enlargement operation on the second segmentation result; if the first conversion operation is an enlargement operation on the organ image, the second conversion operation is a reduction operation on the second segmentation result. By converting the input organ image, the organ abnormal point segmentation model can integrate more data information when performing segmentation, thereby improving the segmentation accuracy. It should be noted that the method of converting the organ image for data enhancement in this embodiment can also be used in the training process of the above-mentioned organ abnormal point segmentation model to perform data enhancement on the sample images used in training, thereby improving the model accuracy of the organ abnormal point segmentation model.

[0100] After obtaining the first segmentation result and the third segmentation result, the computer device can obtain the final whole body abnormal point segmentation result based on the first segmentation result and the third segmentation result. Optionally, the computer device can merge the first segmentation results corresponding to each organ image to obtain a first merged result; merge the third segmentation results corresponding to each organ image to obtain a second merged result; that is, merge the first segmentation results of the unflipped organ images and merge the third segmentation results corresponding to the flipped images to obtain two types of merged results; then obtain the whole body abnormal point segmentation result based on the first merged result and the second merged result, for example, the first merged result and the second merged result can be fused, such as weighted summing the pixel values ​​at the same position of the first merged result and the second merged result, etc., to obtain the whole body abnormal point segmentation result.

[0101] In one embodiment, the computer device may also process the first merged result and the second merged result with the aid of a whole body abnormal point segmentation model to obtain a whole body abnormal point segmentation result. Optionally, the computer device may connect the first merged result and the second merged result through a channel and input them into a preset whole body abnormal point segmentation model to obtain a whole body abnormal point segmentation result.

[0102] Among them, the whole body abnormal point segmentation model includes but is not limited to a fully convolutional network model, such as a VNet model, a UNet model, etc., and its training method may include: obtaining multiple PET images and CT images as sample images through PET-CT equipment scanning, and having a senior doctor annotate the whole body abnormal point segmentation gold standard (mask) of each sample image; then obtaining an organ image from the sample image, inputting the above-mentioned organ abnormal point segmentation model to obtain the organ abnormal point segmentation result; then, after the segmentation results of each organ abnormal point are merged and the channels are connected, they are input into the initial whole body abnormal point segmentation network, and the loss between the predicted segmentation result and the whole body abnormal point segmentation gold standard is calculated, and the initial whole body abnormal point segmentation network is trained according to the loss to obtain the whole body abnormal point segmentation model. Optionally, the computer device can also randomly select a patch from the sample image as the network input, such as the patch size can be [64,128,128]; the optimizer for network training can be Adam; the calculated loss can also include BCE loss and Dice loss, and the calculation method of the two types of losses can refer to the description of the above embodiment, which will not be repeated here.

[0103] In one embodiment, the computer device may also use the organ segmentation model to obtain the organ image from the image to be segmented. Optionally, Figure 6 As shown, the above S101 may include:

[0104] S501, inputting the image to be segmented into the organ segmentation model of each organ respectively to obtain the organ segmentation result of each organ.

[0105] Optionally, the computer device may first resample the image to be segmented to the same resolution space as the sample image in the training phase of the organ segmentation model, and normalize the resampled image to be segmented, and then input the normalized image to be segmented into the organ segmentation model. The resampled image to be segmented is standardized by the relationship of , I is the resampled image to be segmented, μ is the mean of the sample images in the training stage of the organ segmentation model, and σ is the standard deviation of the sample images in the training stage of the organ segmentation model. In addition, for PET images and CT images, the computer device can connect the standardized PET images and CT images along the channel dimension and then input them into the organ segmentation model. Among them, the computer device needs to input the image to be segmented into the organ segmentation model of each organ respectively, such as inputting the image to be segmented into the pancreas segmentation model, the lung segmentation model and other models respectively, to obtain the organ segmentation results of the pancreas, lungs and other organs, such as organ masks.

[0106] Optionally, the above-mentioned organ segmentation model includes but is not limited to a fully convolutional network model, such as a VNet, UNet model, and its training method may include: a plurality of PET images and CT images can be scanned by a PET-CT device as sample images, and the organ segmentation gold standard (mask) of each sample image is annotated by a senior doctor. Optionally, the computer device can also resample the sample image to the same resolution space, and set a suitable mean and standard deviation to standardize the resampled sample image. The standardization process can refer to the description of the above-mentioned embodiment, which will not be repeated here. Then the standardized sample image is connected along the channel dimension and input into the initial organ segmentation network for training; the parameters such as the loss function and optimizer calculated during the training of the organ segmentation model can refer to the training process of the above-mentioned organ abnormality segmentation model, which will not be repeated here.

[0107] S502 : For each organ, according to the organ segmentation result of the organ, an organ image corresponding to the organ is obtained from the image to be segmented.

[0108] Specifically, the coordinate position of the organ in the image to be segmented can be known according to the organ segmentation result, and the computer device can then obtain (such as by intercepting, cropping, etc.) the organ image corresponding to the organ from the image to be segmented according to the coordinate position.

[0109] In the medical image segmentation method provided in this embodiment, the computer device can input the image to be segmented into the organ segmentation model of each organ to obtain the organ segmentation result of each organ; and obtain the corresponding organ image from the image to be segmented according to the organ segmentation result for each organ. Compared with the segmentation method based on the pixel value of the image to be segmented or artificial marking, the accuracy of the obtained organ image can be improved through the segmentation processing of the organ segmentation model, thereby providing an accurate data basis for the subsequent abnormal point segmentation.

[0110] In one embodiment, the computer device can also determine the maximum standard uptake value (SUVmax), the major and minor axes, and the volume of each abnormal point area based on the obtained whole body abnormal point segmentation results. Optionally, the above data information can be calculated based on the number of pixels in the abnormal point area in the whole body abnormal point segmentation results; in addition, only the data information that the user wants to view can be calculated according to the viewing instructions input by the user to reduce the amount of calculation of the computer device. Then, at least one of the whole body abnormal point segmentation results, the maximum standard uptake value, the major and minor axes, and the volume of the abnormal point area is displayed to the user terminal (such as a computer device used by a doctor). Optionally, the computer device can also display at least one of the whole body abnormal point segmentation results, the maximum standard uptake value, the major and minor axes, and the volume of the abnormal point area according to the selection instructions input by the user through the user terminal.

[0111] In addition, in some clinical applications, doctors can view and switch the above-mentioned whole-body abnormal point segmentation results at any time with one click through the user terminal. For example, if the abnormal point area of ​​the pancreas is being viewed, the abnormal point area of ​​the lung can be directly switched. In addition, there may be errors in the output whole-body abnormal point segmentation results. For example, some abnormal points are not segmented or the segmented ones are not abnormal points. The computer device can also receive abnormal point deletion instructions and / or abnormal point addition instructions input by the user through the user terminal, and update the whole-body abnormal point segmentation results according to the abnormal point deletion instructions and / or abnormal point addition instructions. Optionally, when the user views the whole-body abnormal point segmentation results, it can also switch from one abnormal point display interface to another abnormal point display interface to improve interactivity with the user.

[0112] In one embodiment, the user can also be assisted in completing the follow-up and tracking treatment of the subject. The computer device can obtain multiple whole-body abnormal point segmentation results corresponding to the same subject within a preset time period (such as half a year, a year, etc.), and compare the multiple whole-body abnormal point segmentation results to determine the degree of change in the abnormal point area of ​​the subject within the time period. For example, whether the number of abnormal points has increased, whether the area of ​​the same abnormal point has expanded, whether the treatment is effective, etc., thereby assisting the user to understand the abnormal change trend of the subject and provide accurate efficacy evaluation and post-healing evaluation.

[0113] In order to better understand the entire process of the above-mentioned medical image segmentation method, the method is introduced in the form of an overall embodiment. Figure 7 As shown, the method includes:

[0114] S601, inputting the image to be segmented into the organ segmentation model of each organ respectively to obtain the organ segmentation result of each organ;

[0115] S602, for each organ, according to the organ segmentation result of the organ, obtaining the organ image corresponding to the organ from the image to be segmented, and eliminating the organ images of organs such as the brain, kidney, bladder and heart;

[0116] S603, for each organ image, performing a first conversion operation on the organ image to obtain a flipped image corresponding to the organ image;

[0117] S604, inputting the organ image into the organ abnormality segmentation model to obtain a first segmentation result corresponding to the organ image; inputting the flipped image into the organ abnormality segmentation model to obtain a second segmentation result corresponding to the flipped image;

[0118] S605, performing a second conversion operation on the second segmentation result to obtain a third segmentation result;

[0119] S606, merging the first segmentation results corresponding to the organ images to obtain a first merged result; merging the third segmentation results corresponding to the organ images to obtain a second merged result;

[0120] S607, performing channel connection on the first merging result and the second merging result, and inputting the result into a whole body abnormal point segmentation model to obtain a whole body abnormal point segmentation result.

[0121] The implementation process of each step can refer to the description of the above embodiment. The implementation principle and technical effect are similar and will not be repeated here.

[0122] It should be understood that although Figure 2-Figure 7 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-Figure 7 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0123] In one embodiment, Figure 8 As shown, a medical image segmentation device is provided, including: an acquisition module 11, a segmentation module 12 and a determination module 13.

[0124] Specifically, the acquisition module 11 is used to acquire each organ image in the image to be segmented;

[0125] The segmentation module 12 is used to input each organ image into the corresponding organ abnormality segmentation model to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network;

[0126] The determination module 13 is used to obtain the whole body abnormal point segmentation result corresponding to the image to be segmented according to the abnormal point segmentation result corresponding to each organ image.

[0127] The medical image segmentation device provided in this embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0128] In one embodiment, the organ abnormality segmentation model includes a first segmentation network and a second segmentation network, each segmentation network includes multiple decoders; the segmentation module 12 is specifically used to input the organ image into the first segmentation network of the corresponding organ abnormality segmentation model for each organ image, and output multiple reference segmentation results through multiple decoders of the first segmentation network; wherein one decoder outputs one reference segmentation result; input the organ image and the target segmentation result among the multiple reference segmentation results into the second segmentation network of the organ abnormality segmentation model, and output the abnormality segmentation result corresponding to the organ image through the last decoder among the multiple decoders of the second segmentation network; wherein the target segmentation result is the reference segmentation result output by the last N decoders among the multiple decoders of the first segmentation network, and N≥1.

[0129] In one embodiment, the segmentation network also includes encoders with the same number as decoders; the segmentation module 12 is specifically used to input the organ image and the last target segmentation result into the first encoder of the second segmentation network, and input the nth to last target segmentation result into the nth encoder of the second segmentation network; wherein n≤N.

[0130] In one embodiment, the segmentation module 12 is specifically used to perform a first conversion operation on each organ image to obtain a conversion image corresponding to the organ image; input the organ image into an organ abnormality segmentation model to obtain a first segmentation result corresponding to the organ image; input the conversion image into the organ abnormality segmentation model to obtain a second segmentation result corresponding to the conversion image; perform a second conversion operation on the second segmentation result to obtain a third segmentation result, and use the first segmentation result and the third segmentation result as the abnormality segmentation results corresponding to the organ image; wherein the second conversion operation is an inverse operation of the first conversion operation.

[0131] In one embodiment, the determination module 13 is specifically used to merge the first segmentation results corresponding to each organ image to obtain a first merged result; merge the third segmentation results corresponding to each organ image to obtain a second merged result; and obtain a whole body abnormality point segmentation result corresponding to the image to be segmented based on the first merged result and the second merged result.

[0132] In one embodiment, the determination module 13 is specifically used to perform channel connection on the first merging result and the second merging result, and input the result into the whole body abnormal point segmentation model to obtain the whole body abnormal point segmentation result.

[0133] In one embodiment, the acquisition module 11 is specifically used to input the image to be segmented into the organ segmentation model of each organ respectively to obtain the organ segmentation result of each organ; for each organ, according to the organ segmentation result of the organ, the organ image corresponding to the organ is obtained from the image to be segmented.

[0134] In one embodiment, the above-mentioned determination module 13 is also used to determine the maximum standard uptake value, major and minor axes and volume of each abnormal point area based on the whole body abnormal point segmentation results; and display at least one of the whole body abnormal point segmentation results, the maximum standard uptake value, major and minor axes and volume of the abnormal point area to the user terminal.

[0135] In one embodiment, the above-mentioned device also includes a receiving module for receiving an abnormal point deletion instruction and / or an abnormal point addition instruction input by a user through a user terminal; and updating the whole body abnormal point segmentation result according to the abnormal point deletion instruction and / or the abnormal point addition instruction.

[0136] The specific definition of the medical image segmentation device can be found in the definition of the medical image segmentation method above, which will not be repeated here. Each module in the above-mentioned medical image segmentation device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0137] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 1 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a medical image segmentation method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0138] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0139] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0140] Acquire images of various organs in the image to be segmented;

[0141] Inputting each organ image into the corresponding organ abnormality segmentation model respectively to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network;

[0142] According to the abnormal point segmentation results corresponding to each organ image, the whole body abnormal point segmentation results corresponding to the image to be segmented are obtained.

[0143] The implementation principle and technical effects of the computer device provided in this embodiment are similar to those of the above method embodiments, and will not be repeated here.

[0144] In one embodiment, the organ abnormality segmentation model includes a first segmentation network and a second segmentation network, each segmentation network includes a plurality of decoders;

[0145] When the processor executes the computer program, it also performs the following steps:

[0146] For each organ image, the organ image is input into a first segmentation network of a corresponding organ abnormality segmentation model, and multiple decoders of the first segmentation network output multiple reference segmentation results; wherein one decoder outputs one reference segmentation result;

[0147] The organ image and a target segmentation result from a plurality of reference segmentation results are input into a second segmentation network of the organ abnormality segmentation model, and the abnormality segmentation result corresponding to the organ image is output through the last decoder from among the plurality of decoders of the second segmentation network; wherein the target segmentation result is the reference segmentation result output by the last N decoders from among the plurality of decoders of the first segmentation network, and N≥1.

[0148] In one embodiment, the segmentation network further includes encoders having the same number as the decoders; and the processor further implements the following steps when executing the computer program:

[0149] The organ image and the last target segmentation result are input into the first encoder of the second segmentation network, and the nth to last target segmentation result is input into the nth encoder of the second segmentation network; wherein n≤N.

[0150] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0151] For each organ image, performing a first conversion operation on the organ image to obtain a conversion image corresponding to the organ image;

[0152] Inputting the organ image into the organ abnormality point segmentation model to obtain a first segmentation result corresponding to the organ image; inputting the transformed image into the organ abnormality point segmentation model to obtain a second segmentation result corresponding to the transformed image;

[0153] A second conversion operation is performed on the second segmentation result to obtain a third segmentation result, and the first segmentation result and the third segmentation result are used as abnormal point segmentation results corresponding to the organ image; wherein the second conversion operation is an inverse operation of the first conversion operation.

[0154] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0155] Merging the first segmentation results corresponding to the organ images to obtain a first merged result;

[0156] Merging the third segmentation results corresponding to the organ images to obtain a second merged result;

[0157] According to the first merging result and the second merging result, a whole body abnormal point segmentation result corresponding to the image to be segmented is obtained.

[0158] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0159] The first merging result and the second merging result are channel-connected and input into the whole body abnormal point segmentation model to obtain the whole body abnormal point segmentation result.

[0160] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0161] Input the images to be segmented into the organ segmentation models of the organs respectively to obtain the organ segmentation results of the organs;

[0162] For each organ, an organ image corresponding to the organ is obtained from the image to be segmented according to the organ segmentation result of the organ.

[0163] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0164] According to the results of the segmentation of abnormal points throughout the body, the maximum standard uptake value, major and minor axes, and volume of each abnormal point area are determined;

[0165] The whole body abnormal point segmentation result, at least one of the maximum standard uptake value, the major and minor axes and the volume of the abnormal point area are displayed to the user terminal.

[0166] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0167] Receiving an outlier point deletion instruction and / or an outlier point addition instruction input by a user through a user terminal;

[0168] According to the abnormal point deletion instruction and / or abnormal point addition instruction, the whole body abnormal point segmentation result is updated.

[0169] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0170] Acquire images of various organs in the image to be segmented;

[0171] Inputting each organ image into the corresponding organ abnormality segmentation model respectively to obtain the abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of the latter segmentation network includes at least one output result of the previous segmentation network;

[0172] According to the abnormal point segmentation results corresponding to each organ image, the whole body abnormal point segmentation results corresponding to the image to be segmented are obtained.

[0173] The computer-readable storage medium provided in this embodiment has similar implementation principles and technical effects to those of the above method embodiments, and will not be described in detail here.

[0174] In one embodiment, the organ abnormality segmentation model includes a first segmentation network and a second segmentation network, each segmentation network includes a plurality of decoders;

[0175] When the computer program is executed by the processor, the following steps are also performed:

[0176] For each organ image, the organ image is input into a first segmentation network of a corresponding organ abnormality segmentation model, and multiple decoders of the first segmentation network output multiple reference segmentation results; wherein one decoder outputs one reference segmentation result;

[0177] The organ image and a target segmentation result from a plurality of reference segmentation results are input into a second segmentation network of the organ abnormality segmentation model, and the abnormality segmentation result corresponding to the organ image is output through the last decoder from among the plurality of decoders of the second segmentation network; wherein the target segmentation result is the reference segmentation result output by the last N decoders from among the plurality of decoders of the first segmentation network, and N≥1.

[0178] In one embodiment, the segmentation network further includes encoders having the same number as the decoders; and when the computer program is executed by the processor, the following steps are further implemented:

[0179] The organ image and the last target segmentation result are input into the first encoder of the second segmentation network, and the nth to last target segmentation result is input into the nth encoder of the second segmentation network; wherein n≤N.

[0180] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0181] For each organ image, performing a first conversion operation on the organ image to obtain a conversion image corresponding to the organ image;

[0182] Inputting the organ image into the organ abnormality point segmentation model to obtain a first segmentation result corresponding to the organ image; inputting the transformed image into the organ abnormality point segmentation model to obtain a second segmentation result corresponding to the transformed image;

[0183] A second conversion operation is performed on the second segmentation result to obtain a third segmentation result, and the first segmentation result and the third segmentation result are used as abnormal point segmentation results corresponding to the organ image; wherein the second conversion operation is an inverse operation of the first conversion operation.

[0184] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0185] Merging the first segmentation results corresponding to the organ images to obtain a first merged result;

[0186] Merging the third segmentation results corresponding to the organ images to obtain a second merged result;

[0187] According to the first merging result and the second merging result, a whole body abnormal point segmentation result corresponding to the image to be segmented is obtained.

[0188] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0189] The first merging result and the second merging result are channel-connected and input into the whole body abnormal point segmentation model to obtain the whole body abnormal point segmentation result.

[0190] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0191] Input the images to be segmented into the organ segmentation models of the organs respectively to obtain the organ segmentation results of the organs;

[0192] For each organ, an organ image corresponding to the organ is obtained from the image to be segmented according to the organ segmentation result of the organ.

[0193] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0194] According to the results of the segmentation of abnormal points throughout the body, the maximum standard uptake value, major and minor axes, and volume of each abnormal point area are determined;

[0195] The whole body abnormal point segmentation result, at least one of the maximum standard uptake value, the major and minor axes and the volume of the abnormal point area are displayed to the user terminal.

[0196] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0197] Receiving an outlier point deletion instruction and / or an outlier point addition instruction input by a user through a user terminal;

[0198] According to the abnormal point deletion instruction and / or abnormal point addition instruction, the whole body abnormal point segmentation result is updated.

[0199] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database 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 or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0200] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0201] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A medical image segmentation method, characterized in that: The method comprises: Acquire images of various organs in the image to be segmented; Inputting each organ image into a corresponding organ abnormality segmentation model to obtain an abnormality segmentation result corresponding to each organ image; wherein the organ abnormality segmentation model includes a cascade structure of multiple segmentation networks, and the input data of a subsequent segmentation network includes at least one output result of a previous segmentation network; According to the abnormal point segmentation results corresponding to the images of each organ, a whole body abnormal point segmentation result corresponding to the image to be segmented is obtained; The step of inputting the organ images into the corresponding organ abnormality segmentation model comprises: If the organ type corresponding to each organ image belongs to the organ category set, then each organ image is input into the corresponding organ abnormality point segmentation model; If the organ type corresponding to each organ image does not belong to the organ category set, the organ image is not segmented; the organ category set includes other organ types except the high uptake area.

2. The method according to claim 1, characterized in that The organ abnormality segmentation model includes a first segmentation network and a second segmentation network, each segmentation network includes a plurality of decoders; The step of inputting the organ images into the corresponding organ abnormality segmentation model to obtain abnormality segmentation results corresponding to the organ images comprises: For each organ image, the organ image is input into a first segmentation network of a corresponding organ abnormality segmentation model, and multiple decoders of the first segmentation network output multiple reference segmentation results; wherein one decoder outputs one reference segmentation result; The organ image and a target segmentation result among the multiple reference segmentation results are input into a second segmentation network of the organ abnormality segmentation model, and the abnormality segmentation result corresponding to the organ image is output through the last decoder among the multiple decoders of the second segmentation network; wherein the target segmentation result is the reference segmentation result output by the last N decoders among the multiple decoders of the first segmentation network, and N≥1.

3. The method according to claim 2, characterized in that The segmentation network also includes encoders having the same number as the decoders; the second segmentation network that inputs the organ image and the target segmentation result among the multiple reference segmentation results into the organ abnormality segmentation model includes: The organ image and the last target segmentation result are input into the first encoder of the second segmentation network, and the penultimate n-th target segmentation result is input into the n-th encoder of the second segmentation network; wherein n≤N.

4. The method according to claim 1, characterized in that: The step of inputting the organ images into the corresponding organ abnormality segmentation model to obtain abnormality segmentation results corresponding to the organ images comprises: For each organ image, performing a first conversion operation on the organ image to obtain a conversion image corresponding to the organ image; Inputting the organ image into the organ abnormality segmentation model to obtain a first segmentation result corresponding to the organ image; inputting the conversion image into the organ abnormality segmentation model to obtain a second segmentation result corresponding to the conversion image; A second conversion operation is performed on the second segmentation result to obtain a third segmentation result, and the first segmentation result and the third segmentation result are used as abnormal point segmentation results corresponding to the organ image; wherein the second conversion operation is an inverse operation of the first conversion operation.

5. The method according to claim 4, characterized in that The step of obtaining the whole body abnormal point segmentation result corresponding to the image to be segmented according to the abnormal point segmentation results corresponding to each organ image includes: Merging the first segmentation results corresponding to the respective organ images to obtain a first merged result; Merging the third segmentation results corresponding to the respective organ images to obtain a second merged result; A whole body abnormal point segmentation result corresponding to the image to be segmented is obtained according to the first merging result and the second merging result.

6. The method according to claim 5, characterized in that The step of obtaining a whole body abnormal point segmentation result corresponding to the image to be segmented according to the first merging result and the second merging result includes: The first merging result and the second merging result are channel-connected and input into a whole-body abnormal point segmentation model to obtain the whole-body abnormal point segmentation result.

7. The method according to claim 1, characterized in that The step of obtaining images of various organs in the image to be segmented comprises: Inputting the to-be-segmented images into organ segmentation models of respective organs respectively to obtain organ segmentation results of the respective organs; For each organ, an organ image corresponding to the organ is acquired from the image to be segmented according to the organ segmentation result of the organ.

8. The method according to claim 1, characterized in that The method further comprises: According to the whole body abnormal point segmentation results, determining the maximum standard uptake value, major and minor axes and volume of each abnormal point area; The whole body abnormal point segmentation result, at least one of the maximum standard uptake value, the major and minor axes and the volume of the abnormal point area are displayed to the user terminal.

9. The method according to claim 8, characterized in that The method further comprises: receiving an outlier point deletion instruction and / or an outlier point addition instruction input by a user through the user terminal; The whole body abnormal point segmentation result is updated according to the abnormal point deletion instruction and / or the abnormal point addition instruction.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images