Image sketching method and device, computer device, and storage medium

Abstract

The application relates to an image sketching method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining a to-be-sketched image, performing first sketching processing on a target part included in the to-be-sketched image to obtain a first image, and completing primary sketching; identifying an abnormal type corresponding to the to-be-sketched image, and obtaining target target area sketching criteria based on the abnormal type; determining a target target area boundary in the first image based on the target target area sketching criteria, and performing cutting on the first image based on the target target area boundary to obtain a second image, so that redundant parts in the first image can be removed, and only a core part image related to the abnormal type is reserved; performing second sketching processing on the target part included in the second image to obtain a target image in which the target part is sketched, and completing secondary sketching which is more accurate. The method can effectively solve problems such as boundary positioning deviation, shrinkage and small broken points in a target area segmentation task, and improve target area sketching accuracy.

Description

Technical Field

[0001] This application relates to the field of medical imaging technology, and in particular to an image delineation method, apparatus, computer equipment, storage medium, and computer program product. Background Technology

[0002] Currently, medical image segmentation can be divided into manual segmentation and automatic segmentation. Manual segmentation has high accuracy and is often considered the gold standard. However, the accuracy of manual segmentation is largely dependent on the operator's prior knowledge, and the process is time-consuming, requiring significant effort and time for delineation. Therefore, automatic segmentation of medical images is particularly important and urgent, especially since deep learning has made a significant comeback in the image processing field, and the accuracy of deep learning-based automatic segmentation algorithms has been improving year by year.

[0003] In current automatic target segmentation tasks, there are often positioning deviations, shrinkage, or small fragments at the target boundary, resulting in inaccurate target delineation. Summary of the Invention

[0004] Based on this, it is necessary to provide an image delineation method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of target area delineation in response to the above-mentioned technical problems.

[0005] Firstly, this application provides an image delineation method. The method includes:

[0006] Obtain the image to be outlined, and perform a first outlining process on the target parts included in the image to be outlined to obtain the first image;

[0007] Identify the anomaly type corresponding to the image to be delineated, and obtain the target area delineation criteria based on the anomaly type;

[0008] The target region boundary in the first image is determined based on the target region delineation criteria, and the first image is cropped based on the target region boundary to obtain the second image;

[0009] The target area included in the second image is subjected to a second delineation process to obtain a target image in which the target area is delineated.

[0010] In one embodiment, the anomaly type corresponding to the image to be delineated is identified, and the target area delineation criteria are obtained based on the anomaly type, including:

[0011] Identify abnormal regions corresponding to target parts included in the image to be delineated, and determine the abnormality type of the image to be delineated based on the abnormal regions;

[0012] Based on the correspondence between anomaly types and target delineation criteria, target delineation criteria corresponding to anomaly types are obtained from multiple target delineation criteria and used as target target delineation criteria.

[0013] In one embodiment, determining the boundary of the target region in the first image based on the target region delineation criterion includes:

[0014] The target area boundary and target center in the first image are determined according to the target area delineation criteria.

[0015] A fixed-size image patch is obtained from the first image based on the center of the target area;

[0016] Based on the target area delineation criteria, retain the portion of the image patch within the target area boundary and remove the portion outside the target area boundary.

[0017] Secondly, this application also provides an image outlining device. The device includes:

[0018] The first outlining module is used to acquire the image to be outlined and perform a first outlining process on the target parts included in the image to be outlined to obtain the first image.

[0019] The criteria matching module is used to identify the anomaly type of the image to be delineated and to obtain the target area delineation criteria based on the anomaly type.

[0020] The image processing module is used to determine the boundary of the target region in the first image based on the target region delineation criteria, and to crop the first image based on the target region boundary to obtain the second image;

[0021] The second outlining module is used to perform a second outlining process on the target parts included in the second image to obtain a target image with the target parts outlined.

[0022] In one embodiment, the criterion matching module is further configured to identify abnormal regions corresponding to target parts included in the image to be delineated, and determine the abnormal type of the image to be delineated based on the abnormal regions; based on the correspondence between the abnormal type and the target delineation criteria, the target delineation criteria corresponding to the abnormal type are obtained from multiple target delineation criteria and used as the target target delineation criteria.

[0023] In one embodiment, the image processing module is further configured to determine the target region boundary and target region center in the first image according to the target region delineation criteria; obtain a fixed-size image block in the first image according to the target region center; retain the portion of the image block within the target region boundary according to the target region delineation criteria, and remove the portion outside the target region boundary.

[0024] 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:

[0025] Obtain the image to be outlined, and perform a first outlining process on the target parts included in the image to be outlined to obtain the first image;

[0026] Identify the anomaly type corresponding to the image to be delineated, and obtain the target area delineation criteria based on the anomaly type;

[0027] The target region boundary in the first image is determined based on the target region delineation criteria, and the first image is cropped based on the target region boundary to obtain the second image;

[0028] The target area included in the second image is subjected to a second delineation process to obtain a target image in which the target area is delineated.

[0029] 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:

[0030] Obtain the image to be outlined, and perform a first outlining process on the target parts included in the image to be outlined to obtain the first image;

[0031] Identify the anomaly type corresponding to the image to be delineated, and obtain the target area delineation criteria based on the anomaly type;

[0032] The target region boundary in the first image is determined based on the target region delineation criteria, and the first image is cropped based on the target region boundary to obtain the second image;

[0033] The target area included in the second image is subjected to a second delineation process to obtain a target image in which the target area is delineated.

[0034] 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:

[0035] Obtain the image to be outlined, and perform a first outlining process on the target parts included in the image to be outlined to obtain the first image;

[0036] Identify the anomaly type corresponding to the image to be delineated, and obtain the target area delineation criteria based on the anomaly type;

[0037] The target region boundary in the first image is determined based on the target region delineation criteria, and the first image is cropped based on the target region boundary to obtain the second image;

[0038] The target area included in the second image is subjected to a second delineation process to obtain a target image in which the target area is delineated.

[0039] The aforementioned image delineation method, apparatus, computer equipment, storage medium, and computer program product acquire an image to be delineated, perform a first delineation process on the target regions included in the image to be delineated to obtain a first image, completing the initial delineation; identify the anomaly type corresponding to the image to be delineated, and obtain target region delineation criteria based on the anomaly type; determine the target region boundary in the first image based on the target region delineation criteria, and crop the first image based on the target region boundary to obtain a second image. This removes redundant parts from the first image, retaining only the core parts related to the anomaly type; and perform a second delineation process on the target regions included in the second image to obtain a target image with the target regions delineated, completing a more refined secondary delineation. This effectively solves problems such as boundary positioning deviation, shrinkage, and small fragments in target region segmentation tasks, improving the accuracy of target region delineation.

[0040] Sixthly, this application also provides a method for training a segmentation model. The method includes:

[0041] Obtain the first training set, which includes multiple sample medical images and the first target area delineation image corresponding to each sample medical image;

[0042] The first initial neural network is trained based on the first training set, and the trained first initial neural network is used as the first segmentation model; the first segmentation model is used to perform the first delineation process on the target parts included in the image to be delineated, so as to obtain the first image.

[0043] The second training set is obtained based on the first training set. The second training set includes multiple cropped sample medical images and the second target area delineation image corresponding to each cropped sample medical image.

[0044] The second initial neural network is trained based on the second training set, and the trained second initial neural network is used as the second segmentation model. The second segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0045] In one embodiment, obtaining a second training set based on a first training set includes:

[0046] Based on the first target region delineation image corresponding to each sample medical image in the first training set, the target region boundary of each sample medical image is determined.

[0047] The medical images of each sample are cropped based on the target region boundary to obtain multiple sample segmentation images;

[0048] Obtain the second target region delineation image corresponding to the segmentation image of each sample;

[0049] A second training set is constructed based on the segmented images of each sample and the corresponding second target region delineation images.

[0050] In one embodiment, the target region boundary of each medical image sample is determined based on the first target region delineation image corresponding to each sample medical image in the first training set, including:

[0051] Identify the abnormality types corresponding to each sample medical image, and based on the abnormality types corresponding to each sample medical image, obtain the target area delineation criteria corresponding to each sample medical image.

[0052] The target area boundaries of each medical image sample are determined based on the target area delineation criteria corresponding to each sample medical image.

[0053] In one embodiment, the abnormality type corresponding to each sample medical image is identified, and based on the abnormality type corresponding to each sample medical image, the target area delineation criteria corresponding to each sample medical image are obtained, including:

[0054] Identify the abnormal regions in each sample medical image and determine the abnormality type in each sample medical image based on the abnormal regions;

[0055] Based on the correspondence between anomaly type and target delineation criteria, the target delineation criteria corresponding to each sample medical image are obtained from multiple target delineation criteria and used as the target delineation criteria for each sample medical image.

[0056] Seventhly, this application also provides an image outlining device. The device includes:

[0057] The first construction module is used to obtain the first training set, which includes multiple sample medical images and the first target area delineation image corresponding to each sample medical image.

[0058] The first training module is used to train the first initial neural network based on the first training set, and use the trained first initial neural network as the first segmentation model; the first segmentation model is used to perform a first delineation process on the target parts included in the image to be delineated, and obtain the first image.

[0059] The second construction module is used to obtain a second training set based on the first training set. The second training set includes multiple cropped sample medical images and the second target area delineation image corresponding to each cropped sample medical image.

[0060] The second training module is used to train the second initial neural network based on the second training set, and the trained second initial neural network is used as the second segmentation model; the second segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0061] In one embodiment, the second construction module is further configured to determine the target region boundary of each sample medical image based on the first target region delineation image corresponding to each sample medical image in the first training set; crop each sample medical image based on the target region boundary to obtain multiple sample segmentation images; obtain the second target region delineation image corresponding to each sample segmentation image; and construct a second training set based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image.

[0062] In one embodiment, the second construction module is further configured to identify the abnormality type corresponding to each sample medical image, and obtain the target area delineation criteria corresponding to each sample medical image based on the abnormality type corresponding to each sample medical image; and determine the target area boundary of each sample medical image based on the target area delineation criteria corresponding to each sample medical image.

[0063] In one embodiment, the second construction module is further configured to identify the abnormal regions corresponding to each sample medical image, and determine the abnormal types corresponding to each sample medical image based on the abnormal regions; based on the correspondence between the abnormal types and the target delineation criteria, obtain the target delineation criteria corresponding to each sample medical image from multiple target delineation criteria, and use them as the target delineation criteria corresponding to each sample medical image.

[0064] Eighthly, 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:

[0065] Obtain the first training set, which includes multiple sample medical images and the first target area delineation image corresponding to each sample medical image;

[0066] The first initial neural network is trained based on the first training set, and the trained first initial neural network is used as the first segmentation model; the first segmentation model is used to perform the first delineation process on the target parts included in the image to be delineated, so as to obtain the first image.

[0067] The second training set is obtained based on the first training set. The second training set includes multiple cropped sample medical images and the second target area delineation image corresponding to each cropped sample medical image.

[0068] The second initial neural network is trained based on the second training set, and the trained second initial neural network is used as the second segmentation model. The second segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0069] Ninthly, 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:

[0070] Obtain the first training set, which includes multiple sample medical images and the first target area delineation image corresponding to each sample medical image;

[0071] The first initial neural network is trained based on the first training set, and the trained first initial neural network is used as the first segmentation model; the first segmentation model is used to perform the first delineation process on the target parts included in the image to be delineated, so as to obtain the first image.

[0072] The second training set is obtained based on the first training set. The second training set includes multiple cropped sample medical images and the second target area delineation image corresponding to each cropped sample medical image.

[0073] The second initial neural network is trained based on the second training set, and the trained second initial neural network is used as the second segmentation model. The second segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0074] Tenthly, 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:

[0075] Obtain the first training set, which includes multiple sample medical images and the first target area delineation image corresponding to each sample medical image;

[0076] The first initial neural network is trained based on the first training set, and the trained first initial neural network is used as the first segmentation model; the first segmentation model is used to perform the first delineation process on the target parts included in the image to be delineated, so as to obtain the first image.

[0077] The second training set is obtained based on the first training set. The second training set includes multiple cropped sample medical images and the second target area delineation image corresponding to each cropped sample medical image.

[0078] The second initial neural network is trained based on the second training set, and the trained second initial neural network is used as the second segmentation model. The second segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0079] The training method, apparatus, computer equipment, storage medium, and computer program product of the above-mentioned segmentation model involve obtaining a first training set, training a first initial neural network based on the first training set, and using the trained first initial neural network as the first segmentation model; then obtaining a second training set based on the first training set, training a second initial neural network based on the second training set, and using the trained second initial neural network as the second segmentation model. By performing target region delineation twice on the image to be delineated using the first segmentation model and the second segmentation model, problems such as boundary positioning deviation, shrinkage, and small fragments in target region segmentation tasks can be effectively solved, thereby improving the accuracy of target region delineation. Attached Figure Description

[0080] Figure 1 This is a flowchart illustrating an image delineation method in one embodiment;

[0081] Figure 2 This is a schematic diagram illustrating the effect of an image delineation method in one embodiment;

[0082] Figure 3 This is a flowchart illustrating the training method of a segmentation model in one embodiment;

[0083] Figure 4 This is a flowchart illustrating the overall training process of a segmentation model in one embodiment.

[0084] Figure 5 This is a structural block diagram of an image outlining device in one embodiment;

[0085] Figure 6 This is a structural block diagram of a training device for a segmentation model in one embodiment;

[0086] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0087] 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.

[0088] In one embodiment, such as Figure 1As shown, an image delineation method is provided. Based on traditional medical image segmentation, it combines two image delineation steps and target area delineation criteria to jointly solve problems such as boundary positioning deviation, shrinkage, and small fragments. This embodiment illustrates the method by applying it to a computer device. It is understood that the computer device can specifically be a terminal or a server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart medical devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. The server can be a standalone server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0089] Step 102: Obtain the image to be outlined, and perform a first outlining process on the target parts included in the image to be outlined to obtain a first image.

[0090] The image to be delineated can be a medical image, such as a single-modal image, including functional and structural images, such as CT (Computed Tomography), MRI (Magnetic Resonance Imaging), PET (Positron Emission Computed Tomography), and ultrasound images. The target site refers to the body part or region where the patient needs medical examination, including but not limited to human or animal organs, tissues, and target areas; for example, when a patient needs radiotherapy for tumors, the target site could be the patient's lesion target area and / or high-risk target area. The image to be delineated can also be a multimodal image, which refers to a collection of medical image data from multiple modalities. The image to be delineated can also be a non-medical image, including but not limited to optical images and infrared images.

[0091] Optionally, the computer device may perform the first delineation processing on the image to be delineated in a manner that includes, but is not limited to, using an image segmentation model. The computer device inputs the image to be delineated into the image segmentation model to obtain a first image with delineation marks, thus completing the first delineation processing. The first delineation processing is equivalent to performing a coarse target region segmentation on the target area in the image to be delineated.

[0092] Step 104: Identify the anomaly type corresponding to the image to be delineated, and obtain the target area delineation criteria based on the anomaly type.

[0093] The abnormality type can be the lesion type of the target site, such as a tumor, muscle tear, or fracture; it can also be the type or size of the target site, for example, different delineation criteria are selected based on different organs; it can also be the image type, for example, the delineation criteria are selected based on the image modality; or it can be the patient information corresponding to the image, such as the patient's age, gender, and other medical information, which are used to select delineation criteria. The target delineation criterion refers to the method of cropping the target area boundary in the image to be delineated, determined based on the abnormality type. Generally, different abnormality types require different delineation criteria depending on different medical needs. For example, a target delineation criterion based on fracture abnormality types may only ensure that the outline of the damaged bone is within the cropping boundary; another target delineation criterion based on tumor abnormality types requires that the tumor and the entire treatment radiation area be within the cropping boundary; more specifically, the target delineation criterion based on breast cancer abnormality types uses the lower boundary of the clavicle head in the image to be delineated (breast cancer CT image) as the upper boundary of the cropping boundary, and the horizontal position of the positioning point (or lead point) as the lower boundary of the cropping boundary.

[0094] Optionally, the computer device can directly perform image recognition on the image to be delineated to obtain the anomaly type; it can also identify the anomaly type based on the text information (label, image name, image description text) corresponding to the image to be delineated; or it can directly input the anomaly type corresponding to the current image to be delineated from an external source. After determining the anomaly type corresponding to the image to be delineated, the computer device selects the target area delineation criterion that matches the current anomaly type from multiple preset target area delineation criteria.

[0095] Step 106: Determine the target region boundary in the first image based on the target region delineation criteria, and crop the first image based on the target region boundary to obtain the second image.

[0096] The target area boundary contains the complete target part.

[0097] Optionally, the computer device determines the target area boundary in the first image according to the target area delineation criteria. The target area boundary can be a regular boundary including but not limited to the upper boundary, lower boundary, left boundary and right boundary, or it can be an irregular boundary formed by irregular curves surrounding the target area. The first image is cropped based on the target area boundary, retaining the part within the target area boundary range as the second image, and removing the part outside the target area boundary range.

[0098] In one feasible implementation, the computer device determines the boundary and center of the target region in the first image according to the target region delineation criteria. The target region boundary includes an upper boundary, a lower boundary, a left boundary, a right boundary, a front boundary, and a rear boundary. First, a fixed-size image block (or at a specific distance from the boundary) is extracted from the first image based on the target region center. Then, according to the target region delineation criteria, the portion of this image block within the specific boundary range is retained, while the portion outside the specific boundary range of the target region is removed. The fixed size can be determined according to the target region delineation criteria, or it can be predetermined or determined manually.

[0099] Step 108: Perform a second delineation process on the target area included in the second image to obtain a target image with the target area delineated.

[0100] Optionally, the computer device may perform a second delineation process on the second image in a manner that includes, but is not limited to, using an image segmentation model. The computer device inputs the second image into the image segmentation model to obtain a second image with delineation marks. These marks delineate the target region, and the second image with the delineation marks is used as the target image. This second delineation process is equivalent to performing fine target region segmentation on the target region in the image to be delineated.

[0101] In one feasible implementation, before performing the second delineation process on the target part included in the second image, the pixel value of a specified area in the second image can be set to a preset fixed value. The specified area refers to the area that does not contain the target part, and the preset fixed value can be set to 0.

[0102] In one feasible implementation, after the second delineation process, the target image can be delineated more times. For example, a first delineation process can be performed first, followed by a second delineation, and then a third delineation. Alternatively, the image after the first delineation process can be divided into two regions: a first region and a second region. The first region can be delineated second, and the second region can be delineated third. The image segmentation model used for each delineation can be different.

[0103] In the above image delineation method, the first image is obtained by acquiring the image to be delineated and performing a first delineation process on the target areas included in the image to be delineated, thus completing the initial delineation. The anomaly type corresponding to the image to be delineated is identified, and a target area delineation criterion is obtained based on the anomaly type. The target area boundary in the first image is determined based on the target area delineation criterion, and the first image is cropped based on the target area boundary to obtain a second image. This removes redundant parts from the first image, retaining only the core parts related to the anomaly type. A second delineation process is then performed on the target areas included in the second image to obtain a target image with the target areas delineated, completing a more refined secondary delineation. This method effectively solves problems such as boundary positioning deviation, shrinkage, and small fragments in target area segmentation tasks, improving the accuracy of target area delineation.

[0104] In one embodiment, the first delineation processing of the target area included in the image to be delineated is performed by a first segmentation model. The method of obtaining the first segmentation model includes: obtaining a first training set, which includes multiple sample medical images and a first target area delineation image corresponding to each sample medical image; training a first initial neural network based on the first training set, and using the trained first initial neural network as the first segmentation model.

[0105] Optionally, multiple cascaded neural networks (equivalent to the first initial neural network) are constructed using computer equipment. The network structure is 3D U-Net. Multiple images with delimited labels are used as the first training set, and each cascaded neural network is trained using this first training set. The Adam optimizer can be used to optimize the neural network parameters, and cross-entropy can be used as the loss function for image segmentation and high heterogeneity directional delimited information. Then, multiple images with delimited labels are used as the test set, and each trained cascaded neural network is tested using the test set to obtain an evaluation parameter for each trained cascaded neural network. The model with the best evaluation parameter is selected as the first segmentation model. The DICE coefficient (set similarity measure function) can be used to evaluate the performance of each model, and the model with the highest DICE coefficient in the test set is selected as the first segmentation model.

[0106] In this embodiment, a first training set is acquired, which includes multiple sample medical images and a first target region delineation image corresponding to each sample medical image. A first initial neural network is trained based on the first training set, and the trained first initial neural network is used as a first segmentation model. A coarse segmentation model can be obtained, which can be used to achieve the first delineation of the target region in the image to be delineated.

[0107] Furthermore, based on this embodiment, the same method can be used to draw multiple times to obtain a more accurate drawing result.

[0108] In one embodiment, obtaining a first training set includes: extracting images based on medical images and preprocessing the extracted images to obtain sample medical images, wherein the preprocessing includes at least one of image enhancement, flipping, translation, and rotation; obtaining a first target region delineation image corresponding to the sample medical image; obtaining a training instance based on the sample medical image and the first target region delineation image; and constructing a first training set based on multiple training instances.

[0109] Optionally, the computer device preprocesses the medical images containing the target areas to be delineated. This preprocessing includes: image data format conversion, image data normalization, random partitioning of the data into training and testing sets according to a set ratio, and data augmentation of the training set data, including flipping, translation, and rotation. The preprocessed medical images are used as sample medical images, and a delineation label for the target region is added to each sample medical image to obtain the delineated images of each first target region corresponding to each sample medical image. Each set of sample medical images and the delineated images of the first target regions are used as a training instance (training sample), and a first training set is constructed based on multiple training instances.

[0110] In this embodiment, image extraction is performed based on medical images, and the extracted images are preprocessed to obtain sample medical images. The preprocessing includes at least one of image enhancement, flipping, translation, and rotation. A first target region delineation image corresponding to the sample medical image is obtained. A training instance is obtained based on the sample medical image and the first target region delineation image, and a first training set is constructed based on multiple training instances. By training the image segmentation model using the first training set, a coarse segmentation model can be obtained.

[0111] In one embodiment, the second delineation processing of the target area included in the second image is performed by a second segmentation model. The method of obtaining the second segmentation model includes: obtaining a second training set based on a first training set, the second training set including multiple cropped sample medical images and second target area delineation images corresponding to each cropped sample medical image; training a second initial neural network based on the second training set, and using the trained second initial neural network as the second segmentation model.

[0112] Optionally, the computer equipment can be used to build multiple cascaded neural networks (equivalent to a second initial neural network) with a 3D U-Net structure. Multiple images with delimited labels from the first training set are cropped to retain only the target areas as much as possible. These cropped images serve as the second training set, and each cascaded neural network is trained using this second training set. The Adam optimizer can be used to optimize the neural network parameters, with cross-entropy as the loss function for image segmentation and high heterogeneity directional delimited information. Then, multiple cropped images with delimited labels are used as the test set to test each trained cascaded neural network, obtaining an evaluation parameter for each. The model with the best evaluation parameter is selected as the second segmentation model. The DICE coefficient can be used to evaluate the performance of each model, and the model with the highest DICE coefficient on the test set is selected as the second segmentation model.

[0113] In this embodiment, a second training set is obtained based on a first training set. The second training set includes multiple cropped sample medical images and corresponding second target region delineation images for each cropped sample medical image. A second initial neural network is trained based on the second training set, and the trained second initial neural network is used as a second segmentation model. A fine segmentation model can be obtained, which can be used to achieve a second delineation of the target region in the image to be delineated.

[0114] In one embodiment, obtaining a second training set based on a first training set includes: determining the target region boundary of each sample medical image based on the first target region delineation image corresponding to each sample medical image in the first training set; cropping each sample medical image based on the target region boundary to obtain multiple sample segmentation images; obtaining the second target region delineation image corresponding to each sample segmentation image; and constructing a second training set based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image.

[0115] Optionally, the computer device acquires a training instance from the first training set, namely a pair of sample medical images and a corresponding first target region delineation image. Based on the abnormality type corresponding to this pair of images, the target region boundary of this pair of images is determined, and the sample medical image and the first target region delineation image are cropped according to the target region boundary to obtain a training instance (training sample) from the second training set. The computer device processes each training instance in the first training set using the same method to obtain the second training set.

[0116] In this embodiment, the target region boundaries of each medical image sample are determined based on the first target region delineation image corresponding to each medical image sample in the first training set; the medical images sample are cropped based on the target region boundaries to obtain multiple sample segmentation images; the second target region delineation image corresponding to each sample segmentation image is obtained; and a second training set is constructed based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image. A second training set associated with the first training set can be obtained. By training the image segmentation model using the second training set, a fine segmentation model associated with the coarse segmentation model can be obtained.

[0117] In one embodiment, identifying the anomaly type corresponding to the image to be delineated and obtaining the target delineation criteria based on the anomaly type includes: identifying the abnormal region corresponding to the target part included in the image to be delineated, and determining the anomaly type of the image to be delineated based on the abnormal region; and obtaining the target delineation criteria corresponding to the anomaly type from multiple target delineation criteria based on the preset correspondence between the anomaly type and the target delineation criteria, and using it as the target target delineation criteria.

[0118] Abnormal areas refer to organs, small body parts or areas that require medical examination in a case, including but not limited to the brain, heart, bones, blood vessels, liver, kidneys, gallbladder, pancreas, thyroid, urinary system, uterus and adnexa, teeth, etc. in humans or animals; for example, when a case requires tumor examination, the abnormal area can be the tumor area of ​​the case.

[0119] Optionally, multiple target delineation criteria are pre-set in the computer device, and each pre-set target delineation criterion corresponds to a different type of abnormality. For example, one target delineation criterion for a fracture abnormality may only ensure that the outline of the damaged bone is within the clipping boundary, while another target delineation criterion for a tumor abnormality may require that the tumor and the entire treatment radiation area be within the clipping boundary.

[0120] In this embodiment, abnormal regions corresponding to target parts included in the image to be delineated are identified, and the abnormality type of the image to be delineated is determined based on the abnormal regions. Based on the preset correspondence between the abnormality type and the target delineation criteria, the target delineation criteria corresponding to the abnormality type are obtained from multiple target delineation criteria and used as the target target delineation criteria. This allows for the automatic selection of suitable target target delineation criteria for the target parts included in the image to be delineated, thereby determining suitable target boundaries. Subsequent image cropping based on these target boundaries facilitates a second delineation of the target parts on the cropped image, preventing issues such as delineation boundary positioning deviation, shrinkage, and small fragments, thus improving the accuracy of target delineation.

[0121] In one embodiment, an image delineation method includes:

[0122] The computer device extracts images from medical images and preprocesses the extracted images to obtain sample medical images. The preprocessing includes at least one of image enhancement, flipping, translation, and rotation. It then obtains a first target region delineation image corresponding to the sample medical image. Based on the sample medical image and the first target region delineation image, a training instance is obtained. A first training set is constructed based on multiple training instances. The first training set includes multiple sample medical images and the first target region delineation image corresponding to each sample medical image.

[0123] The computer device trains the first initial neural network based on the first training set, and uses the trained first initial neural network as the first segmentation model.

[0124] The computer device determines the target region boundary of each medical image sample based on the first target region delineation image corresponding to each sample medical image in the first training set; it then crops each medical image sample based on the target region boundary to obtain multiple sample segmentation images; it acquires the second target region delineation image corresponding to each sample segmentation image; and it constructs a second training set based on each sample segmentation image and the corresponding second target region delineation image. The second training set includes multiple cropped medical images and the corresponding second target region delineation images for each cropped medical image.

[0125] The computer device trains the second initial neural network based on the second training set, and uses the trained second initial neural network as the second segmentation model.

[0126] The computer device acquires the image to be delineated and performs a first delineation process on the target parts included in the image to be delineated using a first segmentation model to obtain a first image.

[0127] The computer device identifies abnormal regions corresponding to target parts included in the image to be delineated, and determines the abnormal type of the image to be delineated based on the abnormal regions; based on the preset correspondence between the abnormal type and the target delineation criteria, the target delineation criteria corresponding to the abnormal type are obtained from multiple target delineation criteria and used as the target target delineation criteria.

[0128] The computer device determines the boundary of the target region in the first image based on the target region delineation criteria, and crops the first image based on the target region boundary to obtain the second image.

[0129] The computer device performs a second delineation process on the target parts included in the second image using a second segmentation model, and obtains a target image with the target parts delineated.

[0130] In one feasible implementation example, an image delineation method, taken as an example for delineating breast cancer target regions, includes:

[0131] Computer equipment acquires CT images of the patient's chest region and performs coarse segmentation of the target areas included in the CT images using a first segmentation model, resulting in the following... Figure 2 The first image shown on the left.

[0132] The computer device identifies abnormal regions corresponding to target parts included in the image to be delineated, and determines the abnormal type of the image to be delineated based on the abnormal regions; based on the preset correspondence between the abnormal type and the target delineation criteria, the target delineation criteria corresponding to the abnormal type are obtained from multiple target delineation criteria and used as the target target delineation criteria.

[0133] The computer device determines the boundary of the target region in the first image based on the target region delineation criteria, and crops the first image based on the target region boundary to obtain the following result: Figure 2 The second image shown in the middle.

[0134] In the second image, a clavicle head truncating line is defined. The computer device sets the pixel values ​​of the portion of the image above the clavicle head truncating line to 0, resulting in the image shown below. Figure 2 The third image shown on the right.

[0135] The computer device performs fine segmentation on the target area included in the third image using a second segmentation model to obtain a target image that outlines the target area.

[0136] In one embodiment, such as Figure 3 As shown, a training method for a segmentation model is provided. This embodiment illustrates the method by applying it to a computer device; it is understood that the computer device can specifically be a terminal or a server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart medical devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. The server can be a standalone server or a server cluster consisting of multiple servers. In this embodiment, the method includes the following steps:

[0137] Step 302: Obtain the first training set, which includes multiple sample medical images and the first target area delineation image corresponding to each sample medical image.

[0138] Optionally, the computer device may use multiple images with delimited labels as the first training set.

[0139] Step 304: Train the first initial neural network based on the first training set, and use the trained first initial neural network as the first segmentation model; the first segmentation model is used to perform the first delineation processing on the target parts included in the image to be delineated, and obtain the first image.

[0140] Optionally, multiple cascaded neural networks (equivalent to the first initial neural network) are constructed using computer equipment. The network structure is 3D U-Net, and each cascaded neural network is trained using the first training set. The Adam optimizer can be used to optimize the neural network parameters, and cross-entropy can be used as the loss function for image segmentation and high heterogeneity directional delineation information. Then, multiple images with delineation labels are used as a test set to test each trained cascaded neural network, obtaining an evaluation parameter for each trained cascaded neural network. The model with the best evaluation parameter is selected as the first segmentation model. The DICE coefficient can be used to evaluate the performance of each model, and the model with the highest DICE coefficient in the test set is selected as the first segmentation model.

[0141] Step 306: Obtain a second training set based on the first training set. The second training set includes multiple cropped sample medical images and the second target region delineation image corresponding to each cropped sample medical image.

[0142] Optionally, the computer device performs boundary cropping on multiple images with delineated labels in the first training set, retaining only the target parts as much as possible, and uses the cropped images as the second training set.

[0143] Step 308: Train the second initial neural network based on the second training set, and use the trained second initial neural network as the second segmentation model; the second segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0144] Optionally, multiple cascaded neural networks (equivalent to a second initial neural network) are constructed on the computer equipment. The network structure is 3D U-Net, and each cascaded neural network is trained using a second training set. The Adam optimizer can be used to optimize the neural network parameters, and cross-entropy can be used as the loss function for image segmentation and high heterogeneity directional delineation information. Then, multiple cropped images with delineation labels are used as a test set. Each trained cascaded neural network is tested using the test set to obtain an evaluation parameter for each trained cascaded neural network. The model with the best evaluation parameter is selected as the second segmentation model. The DICE coefficient can be used to evaluate the performance of each model, and the model with the highest DICE coefficient in the test set is selected as the second segmentation model.

[0145] In the training method of the above segmentation model, a first training set is obtained, and a first initial neural network is trained based on the first training set, with the trained first initial neural network serving as the first segmentation model; a second training set is obtained based on the first training set, and a second initial neural network is trained based on the second training set, with the trained second initial neural network serving as the second segmentation model. By performing target region delineation twice on the image to be delineated using the first segmentation model and the second segmentation model, problems such as boundary positioning deviation, shrinkage, and small fragments in target region segmentation tasks can be effectively solved, thereby improving the accuracy of target region delineation.

[0146] In one embodiment, obtaining a first training set includes: extracting images based on medical images and preprocessing the extracted images to obtain sample medical images, wherein the preprocessing includes at least one of image enhancement, flipping, translation, and rotation; obtaining a first target region delineation image corresponding to the sample medical image; obtaining a training instance based on the sample medical image and the first target region delineation image; and constructing a first training set based on multiple training instances.

[0147] Optionally, the computer device preprocesses the medical images containing the target areas to be delineated. This preprocessing includes: image data format conversion, image data normalization, random partitioning of the data into training and testing sets according to a set ratio, and data augmentation of the training set data, including flipping, translation, and rotation. The preprocessed medical images are used as sample medical images, and a delineation label for the target region is added to each sample medical image to obtain the delineated images of each first target region corresponding to each sample medical image. Each set of sample medical images and the delineated images of the first target regions are used as a training instance (training sample), and a first training set is constructed based on multiple training instances.

[0148] In this embodiment, image extraction is performed based on medical images, and the extracted images are preprocessed to obtain sample medical images. The preprocessing includes at least one of image enhancement, flipping, translation, and rotation. A first target region delineation image corresponding to the sample medical image is obtained. A training instance is obtained based on the sample medical image and the first target region delineation image, and a first training set is constructed based on multiple training instances. By training the image segmentation model using the first training set, a coarse segmentation model can be obtained.

[0149] In one embodiment, obtaining a second training set based on a first training set includes: determining the target region boundary of each sample medical image based on the first target region delineation image corresponding to each sample medical image in the first training set; cropping each sample medical image based on the target region boundary to obtain multiple sample segmentation images; obtaining the second target region delineation image corresponding to each sample segmentation image; and constructing a second training set based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image.

[0150] Optionally, the computer device acquires a training instance from the first training set, namely a pair of sample medical images and a corresponding first target region delineation image. Based on the abnormality type corresponding to this pair of images, the target region boundary of this pair of images is determined, and the sample medical image and the first target region delineation image are cropped according to the target region boundary to obtain a training instance (training sample) from the second training set. The computer device processes each training instance in the first training set using the same method to obtain the second training set.

[0151] In this embodiment, the target region boundaries of each medical image sample are determined based on the first target region delineation image corresponding to each medical image sample in the first training set; the medical images sample are cropped based on the target region boundaries to obtain multiple sample segmentation images; the second target region delineation image corresponding to each sample segmentation image is obtained; and a second training set is constructed based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image. A second training set associated with the first training set can be obtained. By training the image segmentation model using the second training set, a fine segmentation model associated with the coarse segmentation model can be obtained.

[0152] In one feasible implementation example, such as Figure 4 As shown, a training method for a segmentation model includes:

[0153] The computer device extracts images from medical images and preprocesses the extracted images to obtain sample medical images. The preprocessing includes at least one of image enhancement, flipping, translation, and rotation. It then obtains a first target region delineation image corresponding to the sample medical image. Based on the sample medical image and the first target region delineation image, a training instance is obtained. Multiple training instances are divided into a first training set and a first test set. The first training set and the first test set each include multiple sample medical images and the corresponding first target region delineation image.

[0154] The computer device determines the target region boundary of each sample medical image based on the first target region delineation image corresponding to each sample medical image in the first training set; it then crops each sample medical image based on the target region boundary to obtain multiple sample segmentation images; it acquires the second target region delineation image corresponding to each sample segmentation image; and it divides each sample segmentation image and its corresponding second target region delineation image into a second training set and a second test set. The second training set and the second test set each include multiple cropped sample medical images and their corresponding second target region delineation images.

[0155] The computer device trains multiple initial neural networks based on a first training set, and tests the multiple trained initial neural networks using a first test set. The neural network with the best test result is used as the coarse segmentation model. The coarse segmentation model is used to perform the first delineation process on the target parts included in the image to be delineated, thereby obtaining the first image.

[0156] The computer device trains the second initial neural network based on the second training set, and tests multiple trained second initial neural networks using the second test set. The neural network with the best test result is used as the fine segmentation model. The fine segmentation model is used to perform a second delineation process on the target parts included in the cropped first image to obtain a target image with the target parts delineated.

[0157] It should be understood that although the steps in the flowcharts of the embodiments described above 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 embodiments described above 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.

[0158] Based on the same inventive concept, this application also provides an image outlining apparatus for implementing the image outlining method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more image outlining apparatus embodiments provided below can be found in the limitations of the image outlining method described above, and will not be repeated here.

[0159] In one embodiment, such as Figure 5 As shown, an image outlining device 500 is provided, including: a first outlining module 501, a criterion matching module 502, an image processing module 503, and a second outlining module 504, wherein:

[0160] The first outlining module 501 is used to acquire the image to be outlined and perform a first outlining process on the target parts included in the image to be outlined to obtain a first image.

[0161] The criterion matching module 502 is used to identify the anomaly type corresponding to the image to be delineated, and to obtain the target area delineation criteria based on the anomaly type.

[0162] Image processing module 503 is used to determine the boundary of the target region in the first image based on the target region delineation criteria, and to crop the first image based on the target region boundary to obtain the second image;

[0163] The second outlining module 504 is used to perform a second outlining process on the target parts included in the second image to obtain a target image with the target parts outlined.

[0164] In one embodiment, the first delineation module 501 is further configured to acquire a first training set, which includes multiple sample medical images and a first target area delineation image corresponding to each sample medical image; and to train a first initial neural network based on the first training set, using the trained first initial neural network as a first segmentation model.

[0165] In one embodiment, the first delineation module 501 is further configured to perform image extraction based on medical images, and preprocess the extracted images to obtain sample medical images. The preprocessing includes at least one of image enhancement, flipping, translation, and rotation. The module also obtains a first target area delineation image corresponding to the sample medical image, and obtains a training instance based on the sample medical image and the first target area delineation image. Finally, the module constructs a first training set based on multiple training instances.

[0166] In one embodiment, the second delineation module 504 is further configured to obtain a second training set based on the first training set, the second training set including multiple cropped sample medical images and second target area delineation images corresponding to each cropped sample medical image; and to train a second initial neural network based on the second training set, using the trained second initial neural network as a second segmentation model.

[0167] In one embodiment, the second delineation module 504 is further configured to determine the target region boundary of each sample medical image based on the first target region delineation image corresponding to each sample medical image in the first training set; crop each sample medical image based on the target region boundary to obtain multiple sample segmentation images; obtain the second target region delineation image corresponding to each sample segmentation image; and construct a second training set based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image.

[0168] In one embodiment, the criterion matching module 502 is used to identify abnormal regions corresponding to target parts included in the image to be delineated, and determine the abnormal type of the image to be delineated based on the abnormal regions; based on the preset correspondence between the abnormal type and the target delineation criteria, the target delineation criteria corresponding to the abnormal type are obtained from multiple target delineation criteria and used as the target target delineation criteria.

[0169] Each module in the aforementioned image drawing 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 call and execute the operations corresponding to each module.

[0170] Based on the same inventive concept, this application also provides a segmentation model training apparatus for implementing the segmentation model training method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations of one or more segmentation model training apparatus embodiments provided below can be found in the limitations of the segmentation model training method described above, and will not be repeated here.

[0171] In one embodiment, such as Figure 6 As shown, a training device 600 for a segmentation model is provided, comprising: a first building module 601, a second building module 602, a first training module 603, and a second training module 604, wherein:

[0172] The first construction module 601 is used to extract images based on medical images and preprocess the extracted images to obtain sample medical images. The preprocessing includes at least one of image enhancement, flipping, translation and rotation. It also obtains a first target area delineation image corresponding to the sample medical image, obtains a training instance based on the sample medical image and the first target area delineation image, and constructs a first training set based on multiple training instances.

[0173] The second construction module 602 is used to determine the target region boundary of each sample medical image based on the first target region delineation image corresponding to each sample medical image in the first training set; to crop each sample medical image based on the target region boundary to obtain multiple sample segmentation images; to obtain the second target region delineation image corresponding to each sample segmentation image; and to construct a second training set based on each sample segmentation image and the second target region delineation image corresponding to each sample segmentation image.

[0174] The first training module 603 is used to acquire a first training set, which includes multiple sample medical images and first target area delineation images corresponding to each sample medical image; to train a first initial neural network based on the first training set, and to use the trained first initial neural network as a first segmentation model; the first segmentation model is used to perform first delineation processing on the target parts included in the image to be delineated, and to obtain a first image.

[0175] The second training module 604 is used to acquire a second training set, which includes multiple cropped sample medical images and second target region delineation images corresponding to each cropped sample medical image; the second initial neural network is trained based on the second training set, and the trained second initial neural network is used as a second segmentation model; the second segmentation model is used to perform second delineation processing on the target regions included in the cropped first image to obtain a target image with the target regions delineated.

[0176] Each module in the training device of the above segmentation model can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0177] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements an image delineation method or a segmentation model training method. The display unit of the computer device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0178] Those skilled in the art will understand that Figure 7The 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.

[0179] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described image outlining method embodiment.

[0180] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described image outlining method embodiment.

[0181] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described image outlining method embodiment.

[0182] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described training method embodiment for the segmentation model.

[0183] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described training method embodiment for the segmentation model.

[0184] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described training method embodiment for the segmentation model.

[0185] 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, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0186] 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.

[0187] 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.

[0188] 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. An image delineation method, characterized in that, The method includes: The image to be delineated is obtained, and the target parts included in the image to be delineated are processed by the first segmentation model to obtain the first image. Identify the anomaly type corresponding to the image to be delineated, and obtain the target area delineation criteria based on the anomaly type; The target region boundary and target region center in the first image are determined based on the target region delineation criteria. A fixed-size image patch is obtained in the first image based on the center of the target area; According to the target area delineation criteria, the portion of the image block within the target area boundary is retained, the portion outside the target area boundary is removed, and the first image is cropped based on the target area boundary to obtain the second image; The target region included in the second image is delineated using the second segmentation model to obtain a target image with the target region delineated.

2. The method according to claim 1, characterized in that, The step of identifying the anomaly type corresponding to the image to be delineated and obtaining the target area delineation criteria based on the anomaly type includes: Identify the abnormal regions corresponding to the target parts included in the image to be delineated, and determine the abnormal type of the image to be delineated based on the abnormal regions; Based on the correspondence between anomaly types and target delineation criteria, the target delineation criteria corresponding to the anomaly type are obtained from multiple target delineation criteria and used as the target target delineation criteria.

3. The method according to claim 1, characterized in that, The method for obtaining the first segmentation model includes: Obtain a first training set, which includes multiple sample medical images and a first target region delineation image corresponding to each sample medical image; The first initial neural network is trained based on the first training set, and the trained first initial neural network is used as the first segmentation model.

4. The method according to claim 3, characterized in that, The second segmentation model is obtained through the following methods: A second training set is obtained based on the first training set. The second training set includes multiple cropped sample medical images and second target area delineation images corresponding to each cropped sample medical image. The second initial neural network is trained based on the second training set, and the trained second initial neural network is used as the second segmentation model.

5. The method according to claim 4, characterized in that, The step of obtaining the second training set based on the first training set includes: Based on the first target region delineation image corresponding to each sample medical image in the first training set, the target region boundary of each sample medical image is determined. Based on the target region boundary, each sample medical image is cropped to obtain multiple sample segmentation images; Obtain the second target region delineation image corresponding to the segmentation image of each sample; The second training set is constructed based on the segmented images of each sample and the corresponding second target region delineation images.

6. The method according to claim 5, characterized in that, The step of determining the target region boundary of each medical image sample based on the first target region delineation image corresponding to each sample medical image in the first training set includes: Identify the abnormality type corresponding to each sample medical image, and obtain the target area delineation criteria corresponding to each sample medical image based on the abnormality type corresponding to each sample medical image. The target area boundaries of each medical image sample are determined based on the target area delineation criteria corresponding to each sample medical image.

7. The method according to claim 6, characterized in that, The process of identifying the abnormality type corresponding to each of the medical images and obtaining the target delineation criteria corresponding to each of the medical images based on the abnormality type includes: Identify the abnormal regions in each sample medical image and determine the abnormal type in each sample medical image based on the abnormal regions; Based on the correspondence between anomaly type and target delineation criteria, target delineation criteria corresponding to each sample medical image are obtained from multiple target delineation criteria and used as target delineation criteria for each sample medical image.

8. An image outlining device, characterized in that, The device includes: The first delineation module is used to acquire the image to be delineated and to perform a first delineation process on the target parts included in the image to be delineated using a first segmentation model to obtain a first image. The criteria matching module is used to identify the anomaly type corresponding to the image to be delineated, and to obtain the target area delineation criteria based on the anomaly type. An image processing module is used to determine the target region boundary and target region center in the first image based on the target region delineation criteria; obtain a fixed-size image block in the first image according to the target region center; retain the portion of the image block within the target region boundary according to the target region delineation criteria, remove the portion outside the target region boundary, and crop the first image based on the target region boundary to obtain a second image. The second delineation module is used to perform a second delineation process on the target parts included in the second image through a second segmentation model to obtain a target image with the target parts delineated.

9. 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 7.

10. A computer program product, said computer program product comprising a computer program, 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 7.

Images