Lesion segmentation method, device and electronic equipment

Abstract

The present disclosure relates to a lesion segmentation method, device and electronic equipment, the method comprising determining at least one preset configuration corresponding to a target object, and determining at least one target magnetic resonance signal based on the overlap degree of at least one preset magnetic resonance image of each preset object in at least one preset magnetic resonance signal and the corresponding annotation image of at least one preset magnetic resonance image of each preset object in at least one preset magnetic resonance signal; performing configuration recognition on the at least one target magnetic resonance image to determine at least one target configuration; determining at least one target segmentation model based on the at least one target magnetic resonance signal and the at least one target configuration; inputting the at least one target magnetic resonance image into the at least one target segmentation model for segmentation, and performing weighted processing on the obtained at least one lesion segmentation image to obtain a target lesion segmentation image. The embodiments of the present disclosure can improve the efficiency and accuracy of lesion segmentation.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, and in particular, to a lesion segmentation method and device and electronic equipment. BACKGROUND

[0002] With the development of artificial intelligence, the application scenarios of artificial intelligence technology are also becoming more and more extensive. For example, artificial intelligence can be combined to perform lesion segmentation. In the related art, a traditional full convolution segmentation network mainly based on supervised learning and relying on manual annotation of a training data set is often used to implement lesion segmentation of an MRI image. However, different MRI signals correspond to images that display lesions differently, and the training data set lacks data of all standard forms of lesions, resulting in inaccurate lesion recognition and low efficiency and accuracy of lesion segmentation. SUMMARY

[0003] The present disclosure provides a lesion segmentation method, device, electronic equipment and storage medium to at least solve the problem that in the related art, different MRI signals correspond to images that display lesions differently, and the training data set lacks data of all standard forms of lesions, resulting in inaccurate lesion recognition and low efficiency and accuracy of lesion segmentation. The technical solutions of the present disclosure are as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, a lesion segmentation method is provided, comprising:

[0005] determining at least one target magnetic resonance signal corresponding to a target object and at least one preset configuration, the at least one target magnetic resonance signal being determined based on an overlap degree of at least one preset magnetic resonance image of each preset object in at least one preset magnetic resonance signal and at least one preset magnetic resonance image corresponding label image of the each preset object in the at least one preset magnetic resonance signal;

[0006] obtaining at least one target magnetic resonance image of the target object under the at least one target magnetic resonance signal;

[0007] performing configuration recognition on the at least one target magnetic resonance image based on the at least one preset configuration to determine at least one target configuration corresponding to the target object;

[0008] determining at least one target segmentation model from at least one preset segmentation model based on the at least one target magnetic resonance signal and the at least one target configuration, the at least one preset segmentation model being a segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal;

[0009] inputting the at least one target magnetic resonance image into the at least one target segmentation model for segmentation to obtain at least one lesion segmentation image.

[0010] performing weighted processing on the at least one lesion segmentation image to obtain a target lesion segmentation image corresponding to the target object.

[0011] In an optional embodiment, the configuration recognition based on the at least one preset configuration on the at least one target magnetic resonance image to determine at least one target configuration corresponding to the target object comprises:

[0012] performing feature extraction processing on the at least one target magnetic resonance image to obtain at least one target feature information;

[0013] obtaining at least one preset sample data; the at least one preset sample data is obtained by fusing the at least one target magnetic resonance image and the at least one preset configuration;

[0014] performing feature extraction processing on the at least one preset sample data to obtain at least one preset feature information;

[0015] performing clustering processing on the at least one preset feature information and the at least one target feature information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image;

[0016] performing configuration recognition on the at least one target magnetic resonance image based on first association information between each preset feature information in each clustering feature information set and each target feature information in the clustering feature information set to obtain at least one target configuration corresponding to the target object.

[0017] In an optional embodiment, the clustering processing on the at least one preset feature information and the at least one target feature information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image comprises:

[0018] determining second association information between each target feature information in the at least one target feature information and the at least one preset feature information;

[0019] performing clustering processing on the at least one preset feature information and the at least one target feature information based on the second association information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image.

[0020] In an optional embodiment, the method further comprises:

[0021] obtaining a target annotation image corresponding to the at least one target magnetic resonance image;

[0022] determine an overlap degree of the at least one target magnetic resonance image and the target annotation image based on the at least one target magnetic resonance image and the target annotation image;

[0023] the weighting processing of the at least one lesion segmentation image includes:

[0024] the at least one lesion segmentation image under the same magnetic resonance signal is weighted and merged to obtain at least one lesion segmentation image under at least one target magnetic resonance signal;

[0025] the overlap degree of the at least one target magnetic resonance image and the target annotation image is normalized to obtain a weight corresponding to each of the at least one lesion segmentation image under at least one target magnetic resonance signal;

[0026] based on the weight corresponding to each of the at least one lesion segmentation image under at least one magnetic resonance signal, the at least one lesion segmentation image under at least one target magnetic resonance signal is weighted to obtain a target lesion segmentation image corresponding to the target object.

[0027] In an optional embodiment, the determination of the overlap degree of the at least one target magnetic resonance image and the target annotation image based on the at least one target magnetic resonance image and the target annotation image includes:

[0028] the at least one target magnetic resonance image is normalized to obtain at least one normalized target magnetic resonance image;

[0029] the at least one normalized target magnetic resonance image is filtered to obtain at least one feature image;

[0030] the overlap degree of the at least one target magnetic resonance image and the target annotation image is determined based on the overlap degree of the at least one feature image and the target annotation image.

[0031] In an optional embodiment, the method further includes:

[0032] based on the overlap degree of at least one preset magnetic resonance image of each preset object under at least one preset magnetic resonance signal and a corresponding annotation image of the at least one preset magnetic resonance image of the each preset object under at least one preset magnetic resonance signal, at least one target magnetic resonance signal corresponding to the target object is determined.

[0033] based on the at least one target magnetic resonance signal corresponding to the each preset object, at least one target magnetic resonance signal corresponding to the target object is determined.

[0034] In an optional embodiment, the method further comprises:

[0035] obtaining a training image set and a preset training model, the training image set comprising at least one to-be-trained magnetic resonance image of each preset object in the at least one preset object under the at least one target magnetic resonance signal and a label image corresponding to the at least one to-be-trained magnetic resonance image of the each preset object under the at least one target magnetic resonance signal;

[0036] training the preset training model based on the training image set to obtain the at least one preset segmentation model.

[0037] In an optional embodiment, the method further comprises:

[0038] obtaining a label image corresponding to the at least one to-be-trained magnetic resonance image of the each preset object under the at least one target magnetic resonance signal and a target training magnetic resonance image, the target training magnetic resonance image being any one of the at least one to-be-trained magnetic resonance image;

[0039] after spatially aligning the at least one to-be-trained magnetic resonance image of the each preset object under the at least one target magnetic resonance signal, taking the label image corresponding to the target training magnetic resonance image as the label image corresponding to the at least one to-be-trained magnetic resonance image of the each preset object under the at least one preset magnetic resonance signal.

[0040] According to a second aspect of the embodiments of the present disclosure, a lesion segmentation device is provided, comprising:

[0041] a first determination module configured to determine at least one target magnetic resonance signal corresponding to a target object and at least one preset configuration, the at least one target magnetic resonance signal being determined based on an overlap degree of at least one preset magnetic resonance image of each preset object in at least one preset object under at least one preset magnetic resonance signal and a label image corresponding to the at least one preset magnetic resonance image of the each preset object under the at least one preset magnetic resonance signal;

[0042] a target magnetic resonance image obtaining module configured to obtain at least one target magnetic resonance image of the target object under the at least one target magnetic resonance signal;

[0043] a second determination module configured to perform configuration identification on the at least one target magnetic resonance image based on the at least one preset configuration to determine at least one target configuration corresponding to the target object.

[0044] The third determining module is configured to determine at least one target segmentation model from at least one preset segmentation model based on the at least one target magnetic resonance signal and the at least one target configuration; the at least one preset segmentation model is a segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal.

[0045] The lesion segmentation image acquisition module is configured to segment the at least one target magnetic resonance image by using the at least one target segmentation model to obtain at least one lesion segmentation image.

[0046] The target lesion segmentation image acquisition module is configured to perform weighting processing on the at least one lesion segmentation image to obtain a target lesion segmentation image corresponding to the target object.

[0047] In an optional embodiment, the second determining module comprises:

[0048] The target feature information acquisition unit is configured to perform feature extraction processing on the at least one target magnetic resonance image to obtain at least one target feature information.

[0049] The preset sample data acquisition unit is configured to acquire at least one preset sample data; the at least one preset sample data is obtained by performing fusion processing on the at least one target magnetic resonance image and the at least one preset configuration.

[0050] The preset feature information acquisition unit is configured to perform feature extraction processing on the at least one preset sample data to obtain at least one preset feature information.

[0051] The clustering feature information set acquisition unit is configured to perform clustering processing on the at least one preset feature information and the at least one target feature information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image.

[0052] The target configuration acquisition unit is configured to perform configuration recognition on the at least one target magnetic resonance image based on first association information between each preset feature information in each clustering feature information set and each target feature information in the clustering feature information set to obtain at least one target configuration corresponding to the target object.

[0053] In an optional embodiment, the clustering feature information set acquisition unit comprises:

[0054] The second association information determining subunit is configured to determine second association information between each target feature information in the at least one target feature information and the at least one preset feature information.

[0055] The clustering feature information set acquisition subunit is configured to perform clustering processing on the at least one preset feature information and the at least one target feature information based on the second association information, to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image.

[0056] In an optional embodiment, the device further comprises:

[0057] The target annotation image acquisition module is configured to acquire a target annotation image corresponding to the at least one target magnetic resonance image.

[0058] The overlap degree determination module is configured to determine an overlap degree of the at least one target magnetic resonance image and the target annotation image based on the at least one target magnetic resonance image and the target annotation image.

[0059] The target lesion segmentation image acquisition module comprises:

[0060] The lesion segmentation image unit is configured to perform weighted merging processing on lesion segmentation images in the at least one lesion segmentation image under the same magnetic resonance signal, to obtain at least one lesion segmentation image under the at least one target magnetic resonance signal.

[0061] The weight acquisition unit is configured to perform normalization processing on the overlap degree of the at least one target magnetic resonance image and the target annotation image, to obtain a weight corresponding to each of the at least one lesion segmentation image under the at least one target magnetic resonance signal.

[0062] The target lesion segmentation image acquisition unit is configured to perform weighted processing on the at least one lesion segmentation image under the at least one target magnetic resonance signal based on the weight corresponding to each of the at least one lesion segmentation image under the at least one magnetic resonance signal, to obtain a target lesion segmentation image corresponding to the target object.

[0063] In an optional embodiment, the overlap degree determination module comprises:

[0064] The normalized target magnetic resonance image determination unit is configured to perform normalization processing on the at least one target magnetic resonance image, to obtain at least one normalized target magnetic resonance image.

[0065] The at least one feature image acquisition unit is configured to perform filtering processing on the at least one normalized target magnetic resonance image, to obtain at least one feature image.

[0066] The overlap degree determination unit is configured to determine the overlap degree of the at least one target magnetic resonance image and the target annotation image based on the overlap degree of the at least one feature image and the target annotation image.

[0067] In an optional embodiment, the apparatus further includes:

[0068] The fourth determining module is configured to determine, based on the at least one preset magnetic resonance image of each of the at least one preset object under at least one preset magnetic resonance signal and the overlap degree of the at least one preset magnetic resonance image of the each of the at least one preset object under the at least one preset magnetic resonance signal and the corresponding annotation image of the each of the at least one preset object under the at least one preset magnetic resonance signal, at least one to-be-selected magnetic resonance signal corresponding to the each of the at least one preset object.

[0069] The fifth determining module is configured to determine, based on the at least one to-be-selected magnetic resonance signal corresponding to the each of the at least one preset object, at least one target magnetic resonance signal corresponding to the target object.

[0070] In an optional embodiment, the apparatus further includes:

[0071] The first obtaining module is configured to obtain a training image set and a preset training model; the training image set includes at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal and an annotation image corresponding to the at least one to-be-trained magnetic resonance image of the each of the at least one preset object under the at least one target magnetic resonance signal.

[0072] The preset segmentation model obtaining module is configured to train the preset training model based on the training image set to obtain the at least one preset segmentation model.

[0073] In an optional embodiment, the apparatus further includes:

[0074] The second obtaining module is configured to obtain an annotation image corresponding to the at least one to-be-trained magnetic resonance image of the each of the at least one preset object under the at least one target magnetic resonance signal and a target training magnetic resonance image, the target training magnetic resonance image being any one of the at least one to-be-trained magnetic resonance image.

[0075] The annotation image obtaining module is configured to, after spatially aligning the at least one to-be-trained magnetic resonance image of the each of the at least one preset object under the at least one target magnetic resonance signal, take the annotation image corresponding to the target training magnetic resonance image as the annotation image corresponding to the at least one to-be-trained magnetic resonance image of the each of the at least one preset object under the at least one preset magnetic resonance signal.

[0076] According to a third aspect of the embodiments of the present disclosure, an electronic device is provided, including a processor, a memory for storing instructions executable by the processor, and the processor is configured to execute the instructions to implement the method according to any one of the first aspect.

[0077] The technical solutions provided by the embodiments of the present disclosure bring at least the following beneficial effects:

[0078] The at least one target magnetic resonance signal is determined based on at least one preset configuration of the target object and an overlap degree of at least one preset magnetic resonance image of each preset object in the at least one preset magnetic resonance signal and the at least one preset magnetic resonance image corresponding to the label image of each preset object in the at least one preset magnetic resonance signal; at least one target magnetic resonance image of the target object under the at least one target magnetic resonance signal is obtained; based on the at least one preset configuration, configuration recognition is performed on the at least one target magnetic resonance image to determine at least one target configuration corresponding to the target object; based on the at least one target magnetic resonance signal and the at least one target configuration, at least one target segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal is determined from the at least one preset segmentation model; the accuracy and pertinence of segmentation can be improved based on the high matching of the target object and the at least one target segmentation model; the at least one target magnetic resonance image is input into the at least one target segmentation model for segmentation to obtain at least one lesion segmentation image; the at least one lesion segmentation image is subjected to weighted processing to obtain a target lesion segmentation image corresponding to the target object, which can further optimize the segmentation effect and greatly improve the efficiency and accuracy of lesion segmentation, and thus improve the reliability of diagnosis.

[0079] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0080] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present disclosure and, together with the specification, serve to explain the principles of the present disclosure and do not limit the present disclosure.

[0081] Figure 1 is a schematic diagram of an application environment according to an exemplary embodiment;

[0082] Figure 2 is a flowchart of a lesion segmentation method according to an exemplary embodiment;

[0083] Figure 3 is a schematic diagram of a preset training model according to an exemplary embodiment;

[0084] Figure 4 is a block diagram of a lesion segmentation device according to an exemplary embodiment;

[0085] Figure 5 is a block diagram of an electronic device for lesion segmentation according to an exemplary embodiment. DETAILED DESCRIPTION

[0086] In order for those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings.

[0087] It should be noted that the terms "first", "second" and the like in the specification and claims of the present disclosure and the above-described drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The implementation described in the following exemplary embodiments does not represent all implementations consistent with the present disclosure. Rather, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0088] 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 for display, analyzed data, etc.) involved in the present disclosure are all information and data authorized by the user or authorized by all parties.

[0089] Please refer to Figure 1 , Figure 1 is a schematic diagram of an application environment according to an exemplary embodiment, as shown in Figure 1 , the application environment can include a terminal 100 and a server 200.

[0090] In an optional embodiment, the terminal 100 can be used to provide a lesion segmentation service for users. Specifically, the terminal 100 can include, but is not limited to, electronic devices such as smart phones, desktop computers, tablet computers, notebook computers, smart speakers, digital assistants, augmented reality (AR) / virtual reality (VR) devices, and smart wearable devices, and can also be software such as an application program running on the above-mentioned electronic devices. Optionally, the operating system running on the electronic device can include, but is not limited to, an Android system, an IOS system, Linux, Windows, etc.

[0091] In an optional embodiment, the server 200 can be configured to pre-train at least one preset segmentation model and provide the terminal 100 with lesion segmentation support based on the at least one preset segmentation model. Specifically, the server 200 can be a physical server, a server cluster or a distributed system composed of multiple physical servers, a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and basic cloud computing services such as big data and artificial intelligence platforms.

[0092] In addition, it should be noted that Figure 1 The above-mentioned only shows one application environment provided by the present disclosure, and other application environments can also be included in actual applications.

[0093] In the embodiments of the present disclosure, the terminal 100 and the server 200 can be directly or indirectly connected through wired or wireless communication, and the present disclosure does not limit this.

[0094] Figure 2 is a flowchart of a lesion segmentation method according to an exemplary embodiment, as Figure 2 The lesion segmentation method shown in the figure is used in a server and includes the following steps.

[0095] In step S201, at least one target magnetic resonance signal corresponding to a target object and at least one preset configuration are determined.

[0096] In a specific embodiment, the target object can be a provider of an image that needs to be segmented, and each preset object in the target object can be a patient, for example.

[0097] In a specific embodiment, the at least one target magnetic resonance signal corresponding to the target object can be determined based on the overlap of at least one preset magnetic resonance image of each preset object in the at least one preset object under at least one preset magnetic resonance signal and a corresponding labeled image of the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal.

[0098] In an optional embodiment, the above-mentioned method can further include:

[0099] Based on the overlap of at least one preset magnetic resonance image of each preset object in the at least one preset object under at least one preset magnetic resonance signal and a corresponding labeled image of the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal, at least one target magnetic resonance signal corresponding to each preset object is determined.

[0100] determine the at least one target magnetic resonance signal corresponding to the target object based on the at least one to-be-selected magnetic resonance signal corresponding to each preset object.

[0101] In one specific embodiment, the overlap degree of the at least one preset magnetic resonance image of each preset object in the at least one preset object under the at least one preset magnetic resonance signal and the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal corresponding to the labeled image can be the overlap degree of the at least one preset feature expression image corresponding to the at least one preset magnetic resonance image of each preset object in the at least one preset object under the at least one preset magnetic resonance signal and the high-value part of the labeled image corresponding to the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal. Specifically, the at least one preset feature expression image can be obtained by normalizing and filtering the at least one preset magnetic resonance image of each preset object in the at least one preset object under the at least one preset magnetic resonance signal.

[0102] In one specific embodiment, based on a preset threshold, a magnetic resonance signal with an overlap degree greater than the preset threshold can be selected from the at least one preset magnetic resonance signal as the at least one to-be-selected magnetic resonance signal corresponding to each preset object. The at least one to-be-selected magnetic resonance signal corresponding to the at least one preset object is statistically analyzed as a whole to obtain the number of objects corresponding to each to-be-selected magnetic resonance signal. Based on the number of objects corresponding to each to-be-selected magnetic resonance signal, the at least one target magnetic resonance signal is determined. Correspondingly, the target magnetic resonance signal can also be determined in combination with a preset object number threshold. When the number of objects corresponding to the to-be-selected magnetic resonance signal is greater than the object number threshold, the to-be-selected magnetic resonance signal is determined as the target magnetic resonance signal.

[0103] In the above embodiments, the at least one target magnetic resonance signal corresponding to the target object is determined based on the to-be-selected magnetic resonance signal corresponding to each preset object, which can filter out the most targeted magnetic resonance signal for the target object, avoiding the use of irrelevant or ineffective signals for analysis, thereby improving the analysis efficiency and accuracy. The at least one target magnetic resonance signal corresponding to the target object is determined based on the at least one to-be-selected magnetic resonance signal corresponding to each preset object of the at least one preset object, which comprehensively considers the conditions of multiple objects and provides more comprehensive information for the target object, thereby improving the accuracy and comprehensiveness of lesion segmentation.

[0104] In one specific embodiment, the at least one preset object can be at least one object for providing the at least one preset magnetic resonance image of the at least one preset object under the at least one preset magnetic resonance signal and the labeled image corresponding to the at least one preset magnetic resonance image under the at least one preset magnetic resonance signal. The at least one preset object can be at least one first preset object.

[0105] In one specific embodiment, the at least one preset magnetic resonance signal can be set in combination with actual needs, and specifically, the at least one preset magnetic resonance signal can include at least one imaging sequence commonly used in magnetic resonance imaging, such as a T1 WI sequence (T1 weighted imaging), a T2 WI sequence (T2 weighted imaging), a T2 FS sequence, and the like.

[0106] In one specific embodiment, the at least one preset magnetic resonance image of each of the at least one preset object under the at least one preset magnetic resonance signal can be at least one preset magnetic resonance image obtained by detecting a to-be-segmented lesion of each of the at least one preset object under the at least one preset magnetic resonance signal. The to-be-segmented lesion can be a region requiring lesion segmentation. The to-be-segmented lesion can correspond to different body lesion sites in combination with actual needs.

[0107] In one specific embodiment, the at least one preset magnetic resonance image of each of the at least one preset object under the at least one preset magnetic resonance signal can be at least one preset magnetic resonance image obtained by detecting a to-be-segmented lesion of each of the at least one preset object under the at least one preset magnetic resonance signal. The to-be-segmented lesion can be a region requiring lesion segmentation. The to-be-segmented lesion can correspond to different body lesion sites in combination with actual needs.

[0108] In one specific embodiment, the overlap degree of the at least one preset magnetic resonance image of each of the at least one preset object under the at least one preset magnetic resonance signal and the at least one preset magnetic resonance image corresponding to the at least one preset magnetic resonance signal of each of the at least one preset object can be an overlap degree of a high-value part. Specifically, the overlap degree can be obtained by first performing normalization and filtering processing on the at least one preset magnetic resonance image of each of the at least one preset object under the at least one preset magnetic resonance signal to obtain at least one preset feature expression image, and then calculating an intersection area of a high-value part of the at least one preset feature expression image and the at least one preset magnetic resonance image corresponding to the at least one preset magnetic resonance signal.

[0109] In one specific embodiment, the at least one preset configuration can be at least one shape of the to-be-segmented lesion. Specifically, in the case where the to-be-segmented lesion is an anal fistula, the at least one preset configuration can include four configurations of the anal fistula: low single type, low complex type, high single type, and high complex type.

[0110] In step S203, at least one target magnetic resonance image of the target object under at least one target magnetic resonance signal is acquired.

[0111] In one specific embodiment, the at least one target magnetic resonance image can be at least one target magnetic resonance image obtained by detecting a lesion to be segmented of the target object under the at least one target magnetic resonance signal.

[0112] In step S205, at least one target configuration of the target object is determined by performing configuration recognition on the at least one target magnetic resonance image based on at least one preset configuration.

[0113] In one specific embodiment, the at least one target configuration can be at least one morphology corresponding to the lesion to be segmented of the target object.

[0114] In an optional embodiment, the above configuration recognition on the at least one target magnetic resonance image based on the at least one preset configuration to determine the at least one target configuration of the target object comprises:

[0115] performing feature extraction processing on the at least one target magnetic resonance image to obtain at least one target feature information;

[0116] acquiring at least one preset sample data;

[0117] performing feature extraction processing on the at least one preset sample data to obtain at least one preset feature information;

[0118] performing clustering processing on the at least one preset feature information and the at least one target feature information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image;

[0119] performing configuration recognition on the at least one target magnetic resonance image based on first association information between each preset feature information in each clustering feature information set and each target feature information in the clustering feature information set to obtain the at least one target configuration of the target object.

[0120] In one specific embodiment, the at least one preset feature information can be at least one feature vector corresponding to the at least one target magnetic resonance image.

[0121] In a specific embodiment, the at least one preset sample data can be used for configuration recognition of the at least one target magnetic resonance image, the at least one preset sample data is obtained by fusing the at least one target magnetic resonance image and at least one preset configuration, and the at least one preset feature information can be at least one feature vector corresponding to the at least one preset sample data. Specifically, at least one standard three-dimensional model corresponding to the at least one preset configuration can be drawn, the at least one standard three-dimensional model is fused with the at least one target magnetic resonance image through a data generation method (through a Cycle-GAN model, a Diffusion model, etc.), and the at least one preset sample data is obtained.

[0122] In a specific embodiment, the clustering feature information set corresponding to each of the at least one target magnetic resonance image can be a set of at least one feature vector. Specifically, in the case where the feature information is in the form of a vector, the clustering feature information set corresponding to each of the at least one target magnetic resonance image can be obtained by performing hierarchical clustering on the at least one preset feature vector and the at least one target feature vector.

[0123] In an optional embodiment, the clustering of the at least one preset feature information and the at least one target feature information to obtain the clustering feature information set corresponding to each of the at least one target magnetic resonance image comprises:

[0124] determining second association information between each of the at least one target feature information and the at least one preset feature information;

[0125] performing clustering on the at least one preset feature information and the at least one target feature information based on the second association information to obtain the clustering feature information set corresponding to each of the at least one target magnetic resonance image.

[0126] In a specific embodiment, the second association information can be distance information between each of the at least one target feature information and the at least one preset feature information. Specifically, in the case where the feature information is in the form of a vector, the distance between each of the at least one target feature vector and the at least one preset feature vector can be used. Optionally, the distance between the feature vectors can use Chebyshev distance as a measure.

[0127] In a specific embodiment, in the case where the feature information is in the form of a vector, the second association information can be distance between each of the at least one target feature vector and the at least one preset feature vector. Based on the second association information, hierarchical clustering can be performed on the at least one preset feature vector and the at least one target feature vector to obtain the feature vector set corresponding to each of the at least one target magnetic resonance image.

[0128] In the above embodiments, the at least one preset feature information and the at least one target feature information are clustered to obtain the clustering feature information set corresponding to each of the at least one target magnetic resonance image, the similar target feature information and preset feature information can be classified into a category, the similarity between the target object and the at least one preset sample data is accurately understood, and thus a diagnosis basis is provided for doctors and the reliability of diagnosis is improved.

[0129] In a specific embodiment, the first association relationship can be distance information between each preset feature information in each clustering feature information set and each target feature information in each clustering feature information set. Specifically, in the case of vector form of feature information, the preset feature information in each clustering feature information set can be divided according to the configuration, and then the Chebyshev distance between each preset feature information in each clustering feature information set and each target feature information in each clustering feature information set under the same configuration is calculated. The Chebyshev distances obtained under the same configuration are added to obtain distance information d1, d2, d3, d4 between the at least one target magnetic resonance image and each configuration. The similarity between the at least one target magnetic resonance image and the at least one preset configuration can be calculated based on the distance information. Thus, the configuration of the at least one target magnetic resonance image is recognized, and the preset configuration with a similarity greater than a preset threshold is determined as the at least one target configuration corresponding to the target object based on the preset threshold.

[0130] In the above embodiments, the at least one preset feature information and the at least one target feature information are clustered to obtain the clustering feature information set corresponding to each of the at least one target magnetic resonance image, the similar target feature information and preset feature information can be classified into a category, the similarity between the target object and the at least one preset sample data is accurately understood, and thus a diagnosis basis is provided for doctors and the reliability of diagnosis is improved.

[0131] In step S207, based on the at least one target magnetic resonance signal and the at least one target configuration, at least one target segmentation model is determined from the at least one preset segmentation model.

[0132] In a specific embodiment, the at least one preset segmentation model can be a segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal, and correspondingly, the at least one target segmentation model can be a segmentation model corresponding to the at least one target configuration under the at least one target magnetic resonance signal.

[0133] In an optional embodiment, the method can further include:

[0134] obtaining a training image set and a preset training model;

[0135] training the preset training model based on the training image set to obtain the at least one preset segmentation model.

[0136] In a specific embodiment, the training image set can be a data set for training the preset training model, and the training image set can include at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal and a labeled image corresponding to the at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal, specifically, the at least one preset object can be the at least one second preset object, the at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal can be at least one to-be-segmented magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal, and the labeled image corresponding to the at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal can be at least one magnetic resonance image corresponding to each of the at least one preset object under the at least one target magnetic resonance signal and containing lesion annotation information.

[0137] In a specific embodiment, the at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal can be subjected to grayscale value normalization processing, and the image can be subjected to filtering processing by using a density-related filter, a contour-related filter, an image entropy-related filter, etc., to obtain at least one feature expression image corresponding to the at least one to-be-trained magnetic resonance image of each of the at least one preset object under the at least one target magnetic resonance signal.

[0138] In a specific embodiment, the preset training model can be a deep learning model to be trained, and a specific model structure can be set according to actual needs. Optionally, the preset training model can include an SC Block module and a Decode Block module, and a branch formed by the connection of the SC Block module and the Decode Block module can be used to enhance the feature extraction capability.

[0139] In a specific embodiment, as shown in Figure 3 Figure 3 ​is a schematic diagram of a preset training model according to an exemplary embodiment, in particular, in combination with Figure 3 It can be seen that at least one feature expression image is input into the preset training model as guide data and a training image set, where the SC Block mainly includes a visual Transformer layer and a convolution layer, the Decode Block mainly includes a deep supervision layer and a SoftMax function layer, and the branch connected with the SC Block and the Decode Block includes two SC Blocks and Decode Blocks, wherein the branch connected with the SC Block and the Decode Block is used to enhance the feature extraction capability, and different magnetic resonance images under different magnetic resonance signals are fused to enable the model to learn the lesion features.

[0140] In one specific embodiment, the training of the preset training model based on the training image set to obtain at least one preset segmentation model includes: determining a current preset object from the at least one preset object, taking at least one current feature expression image corresponding to at least one current training magnetic resonance image of the current preset object under at least one target magnetic resonance signal as guide data, inputting at least one training magnetic resonance image of the current preset object under at least one target magnetic resonance signal into the preset training model for lesion segmentation to obtain at least one current segmentation magnetic resonance image of the current preset object under at least one target magnetic resonance signal, determining a lesion segmentation loss according to the at least one current segmentation magnetic resonance image of the current preset object under at least one target magnetic resonance signal and a current annotation image corresponding to the at least one current training magnetic resonance image of the current preset object under at least one target magnetic resonance signal, and training the preset training model based on the lesion segmentation loss to obtain the preset segmentation model.

[0141] In one specific embodiment, the current preset object can be a training object in a current training cycle, and specifically, the current preset object can be randomly determined from the at least one preset object, or a part of the preset objects not participating in the model training can be randomly selected from the at least one preset object as the current preset object.

[0142] In one specific embodiment, the determination of the lesion segmentation loss according to the at least one current segmentation magnetic resonance image of the current preset object under at least one target magnetic resonance signal and the current annotation image corresponding to the at least one current training magnetic resonance image of the current preset object under at least one target magnetic resonance signal can be combined with a preset loss function, and specifically, the lesion segmentation loss can represent the lesion segmentation performance of the current preset training model. Specifically, the preset loss function can be set in combination with actual applications.

[0143] In a specific embodiment, the preset convergence condition can be set in combination with actual application, for example, the number of execution of the loop iteration operation reaches a preset number, the change detection loss is less than a specified threshold, etc., which can be set in combination with training speed and model accuracy requirement.

[0144] In the above embodiment, the preset training model is trained based on the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal and the annotation image corresponding to the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal, to obtain the at least one preset segmentation model. The at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal can be input during the training process, so that the model learns diversified data, reduces overfitting of the model to specific data, and further improves the accuracy of lesion segmentation.

[0145] In an optional embodiment, the above method can further include:

[0146] obtaining the annotation image corresponding to the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal and the target training magnetic resonance image;

[0147] After spatial alignment of the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal, the annotation image corresponding to the target training magnetic resonance image is taken as the annotation image corresponding to the at least one to-be-trained magnetic resonance image of each preset object under the at least one preset magnetic resonance signal.

[0148] In a specific embodiment, the target training magnetic resonance image can be any one of the at least one to-be-trained magnetic resonance image, and correspondingly, the target training magnetic resonance image can correspond to any one of the at least one target magnetic resonance signal, which can be selected in combination with actual requirement.

[0149] In a specific embodiment, the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal is spatially aligned. Specifically, the magnetic resonance images other than the target training magnetic resonance image in the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal can be coarsely registered to the target training magnetic resonance image based on the spatial pose relationship in the DICOM (Digital Imaging and Communications in Medicine) header information, and then fine registration is performed by using the skin contour as a feature point and adopting ICP registration (Iterative Closest Point), so as to realize the consistency of the image pose of the at least one to-be-trained magnetic resonance image and the target training magnetic resonance image.

[0150] In the above embodiment, the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal is spatially aligned, so that the same anatomical structure and lesion in the at least one to-be-trained magnetic resonance image of each preset object under the at least one target magnetic resonance signal correspond in space, which helps the model to more accurately learn the characteristics of the lesion. The annotation image corresponding to the target training magnetic resonance image is used as the annotation image corresponding to the at least one to-be-trained magnetic resonance image of each preset object under the at least one preset magnetic resonance signal, which greatly simplifies the annotation work and improves the annotation efficiency.

[0151] In step S209, the at least one target magnetic resonance image is input into the at least one target segmentation model for segmentation, to obtain at least one lesion segmentation image.

[0152] In a specific embodiment, the at least one target lesion segmentation image can be at least one segmentation image output by the at least one target segmentation model. Specifically, the at least one target lesion segmentation image corresponds to at least one image obtained by inputting the at least one target magnetic resonance image into the segmentation model corresponding to the at least one target configuration under the at least one target magnetic resonance signal.

[0153] In step S2011, the at least one lesion segmentation image is subjected to weighted processing, to obtain a target lesion segmentation image corresponding to the target object.

[0154] In a specific embodiment, the target lesion segmentation image can be a lesion segmentation image corresponding to the target object.

[0155] In an optional embodiment, the above method can further include:

[0156] obtaining a target annotation image corresponding to the at least one target magnetic resonance image;

[0157] determine the overlap degree of the at least one target magnetic resonance image and the target annotation image based on the at least one target magnetic resonance image and the target annotation image;

[0158] In an optional embodiment, the weighting processing of the at least one lesion segmentation image to obtain the target lesion segmentation image corresponding to the target object comprises:

[0159] performing weighted merging processing on the lesion segmentation images under the same magnetic resonance signal in the at least one lesion segmentation image to obtain at least one lesion segmentation image under at least one target magnetic resonance signal;

[0160] performing normalization processing on the overlap degree of the at least one target magnetic resonance image and the target annotation image to obtain the weight corresponding to each of the at least one lesion segmentation image under at least one target magnetic resonance signal;

[0161] performing weighting processing on the at least one lesion segmentation image under at least one target magnetic resonance signal based on the weight corresponding to each of the at least one lesion segmentation image under at least one magnetic resonance signal to obtain the target lesion segmentation image corresponding to the target object.

[0162] In a specific embodiment, the target annotation image can be an annotation image corresponding to the at least one target magnetic resonance image, which can be an image obtained by performing lesion annotation on any target magnetic resonance image of the at least one target magnetic resonance image. Specifically, the annotation manner of the target annotation image can refer to the above content.

[0163] In a specific embodiment, the overlap degree of the at least one target magnetic resonance image and the target annotation image can be the overlap degree of the high value part of the at least one feature image and the target annotation image.

[0164] In an optional embodiment, the determination of the overlap degree of the at least one target magnetic resonance image and the target annotation image based on the at least one target magnetic resonance image and the target annotation image comprises:

[0165] performing normalization processing on the at least one target magnetic resonance image to obtain normalized at least one target magnetic resonance image;

[0166] performing filtering processing on the normalized at least one target magnetic resonance image to obtain at least one feature image;

[0167] determine the overlap degree of the at least one target magnetic resonance image and the target annotation image based on the overlap degree of the at least one feature image and the target annotation image.

[0168] In a specific embodiment, the normalized at least one target magnetic resonance image can be the at least one magnetic resonance image corresponding to the at least one target magnetic resonance image after normalization processing, specifically, the normalization processing can be normalization processing on the gray value of the magnetic resonance image, so that the value range is between 0 and 1.

[0169] In a specific embodiment, the at least one feature image can be an image obtained by filtering processing on the normalized at least one target magnetic resonance image, and the filtering processing mode can refer to the above content.

[0170] In a specific embodiment, the overlap degree of the at least one feature image and the target annotation image can be used as the overlap degree of the at least one target magnetic resonance image and the target annotation image.

[0171] In the above embodiment, the normalization processing on the at least one target magnetic resonance image can unify the gray value of different images to the same range, eliminate the image feature difference, improve the stability and efficiency of model training, and thus more accurately learn the feature information of the lesion. The filtering processing on the normalized target magnetic resonance image can enhance the image features and reduce the influence of noise on the overlap degree calculation and subsequent analysis. After the normalization and filtering processing, the at least one feature image highlights the key features of the lesion, and improves the accuracy and reliability of the target lesion segmentation image.

[0172] In a specific embodiment, the at least one lesion segmentation image can correspond to the at least one magnetic resonance signal. The at least one lesion segmentation image is divided according to the type of magnetic resonance signal to obtain at least one lesion segmentation image corresponding to each type of magnetic resonance signal. The lesion segmentation images under the same magnetic resonance signal are weighted and combined to obtain the lesion segmentation image corresponding to each type of magnetic resonance signal.

[0173] In a specific embodiment, the weight corresponding to each of the at least one lesion segmentation image can be used for weighting the at least one lesion segmentation image, and specifically, the weight corresponding to each of the at least one lesion segmentation image can be obtained by normalization processing on the overlap degree of the at least one target magnetic resonance image and the target annotation image.

[0174] In the above embodiment, the lesion segmentation images under the same magnetic resonance signal are first weighted and combined, the information of at least one magnetic resonance image under the same magnetic resonance signal can be comprehensively utilized, the error caused by a single segmentation image is reduced, the stability and reliability of the segmentation result are improved, the overlap degree is normalized to obtain the weight corresponding to each of the at least one lesion segmentation image, and the weight of each segmentation image can be assigned based on the similarity between the at least one target magnetic resonance image and the target annotation image. The segmentation accuracy of the target lesion is greatly improved.

[0175] According to the technical solutions provided by the embodiments of the present specification, the at least one target magnetic resonance signal corresponding to the target object is determined based on the at least one preset configuration and the overlap degree between the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal and the corresponding annotation image of the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal; the at least one target magnetic resonance image of the target object under the at least one target magnetic resonance signal is obtained; the at least one target configuration corresponding to the target object is determined based on the at least one preset configuration and the configuration recognition of the at least one target magnetic resonance image; the at least one target segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal is determined from the at least one preset segmentation model based on the at least one target magnetic resonance signal and the at least one target configuration; the accuracy and the pertinence of the segmentation can be improved based on the high matching between the target object and the at least one target segmentation model; the at least one target segmentation model is segmented by inputting the at least one target magnetic resonance image into the at least one target segmentation model, and the at least one lesion segmentation image is obtained; the target lesion segmentation image corresponding to the target object is obtained by weighting the at least one lesion segmentation image, which can further optimize the segmentation effect and greatly improve the efficiency and the accuracy of the lesion segmentation, and further improve the reliability of the diagnosis.

[0176] Figure 4 is a block diagram of a lesion segmentation device according to an exemplary embodiment. Referring to Figure 4 The device comprises:

[0177] The first determination module 410 is configured to determine the at least one target magnetic resonance signal corresponding to the target object and the at least one preset configuration, wherein the at least one target magnetic resonance signal is determined based on the overlap degree between the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal and the corresponding annotation image of the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal.

[0178] The target magnetic resonance image acquisition module 430 is configured to acquire at least one target magnetic resonance image of the target object under at least one target magnetic resonance signal;

[0179] The second determination module 450 is configured to perform configuration identification on the at least one target magnetic resonance image based on at least one preset configuration, and determine at least one target configuration corresponding to the target object;

[0180] The third determination module 470 is configured to determine at least one target segmentation model from at least one preset segmentation model based on the at least one target magnetic resonance signal and the at least one target configuration; the at least one preset segmentation model is a segmentation model corresponding to at least one preset configuration under at least one preset magnetic resonance signal;

[0181] The lesion segmentation image acquisition module 490 is configured to input the at least one target magnetic resonance image into the at least one target segmentation model for segmentation to obtain at least one lesion segmentation image;

[0182] The target lesion segmentation image acquisition module 4110 is configured to perform weighting processing on the at least one lesion segmentation image to obtain a target lesion segmentation image corresponding to the target object.

[0183] In an optional embodiment, the second determination module 450 includes:

[0184] The target feature information acquisition unit is configured to perform feature extraction processing on the at least one target magnetic resonance image to obtain at least one target feature information;

[0185] The preset sample data acquisition unit is configured to acquire at least one preset sample data; the at least one preset sample data is obtained by performing fusion processing on the at least one target magnetic resonance image and the at least one preset configuration;

[0186] The preset feature information acquisition unit is configured to perform feature extraction processing on the at least one preset sample data to obtain at least one preset feature information;

[0187] The clustered feature information set acquisition unit is configured to perform clustering processing on the at least one preset feature information and the at least one target feature information to obtain at least one clustered feature information set corresponding to each of the at least one target magnetic resonance image;

[0188] The target configuration acquisition unit is configured to perform configuration identification on the at least one target magnetic resonance image based on first association information between each preset feature information in each clustered feature information set and each target feature information in the clustered feature information set, to obtain at least one target configuration corresponding to the target object.

[0189] In an optional embodiment, the clustered feature information set acquisition unit includes:

[0190] The second association information determination subunit is configured to determine second association information between each target feature information in the at least one target feature information and the at least one preset feature information.

[0191] The clustering feature information set acquisition subunit is configured to perform clustering processing on the at least one preset feature information and the at least one target feature information based on the second association information, to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image.

[0192] In an optional embodiment, the apparatus further includes:

[0193] The target annotation image acquisition module is configured to acquire a target annotation image corresponding to the at least one target magnetic resonance image.

[0194] The overlap degree determination module is configured to determine an overlap degree of the at least one target magnetic resonance image and the target annotation image based on the at least one target magnetic resonance image and the target annotation image.

[0195] The target lesion segmentation image acquisition module 4110 includes:

[0196] The lesion segmentation image unit is configured to perform weighted merging processing on the lesion segmentation images under the same magnetic resonance signal in the at least one lesion segmentation image, to obtain at least one lesion segmentation image under the at least one target magnetic resonance signal.

[0197] The weight acquisition unit is configured to perform normalization processing on the overlap degree of the at least one target magnetic resonance image and the target annotation image, to obtain a weight corresponding to each of the at least one lesion segmentation image under the at least one target magnetic resonance signal.

[0198] The target lesion segmentation image acquisition unit is configured to perform weighted processing on the at least one lesion segmentation image under the at least one target magnetic resonance signal based on the weight corresponding to each of the at least one lesion segmentation image under the at least one magnetic resonance signal, to obtain a target lesion segmentation image corresponding to the target object.

[0199] In an optional embodiment, the overlap degree determination module includes:

[0200] The normalized target magnetic resonance image determination unit is configured to perform normalization processing on the at least one target magnetic resonance image, to obtain at least one normalized target magnetic resonance image.

[0201] The at least one feature image acquisition unit is configured to perform filtering processing on the at least one normalized target magnetic resonance image, to obtain at least one feature image.

[0202] The overlap determination unit is configured to determine the overlap between the at least one target magnetic resonance image and the target annotation image based on the overlap between the at least one feature image and the target annotation image.

[0203] In an optional embodiment, the apparatus described above further includes:

[0204] The fourth determination module is configured to determine the at least one target magnetic resonance signal corresponding to the target object based on the overlap between the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal and the annotation image corresponding to the at least one preset magnetic resonance image of each preset object under the at least one preset magnetic resonance signal.

[0205] The fifth determination module is configured to determine the at least one target magnetic resonance signal corresponding to the target object based on the at least one target magnetic resonance signal corresponding to each preset object.

[0206] In an optional embodiment, the apparatus described above further includes:

[0207] The first acquisition module is configured to acquire a training image set and a preset training model; the training image set includes the at least one training magnetic resonance image of each preset object under the at least one target magnetic resonance signal and the annotation image corresponding to the at least one training magnetic resonance image of each preset object under the at least one target magnetic resonance signal.

[0208] The preset segmentation model acquisition module is configured to train the preset training model based on the training image set to obtain at least one preset segmentation model.

[0209] In an optional embodiment, the apparatus described above further includes:

[0210] The second acquisition module is configured to acquire the annotation image corresponding to the at least one training magnetic resonance image of each preset object under the at least one target magnetic resonance signal and the target training magnetic resonance image; the target training magnetic resonance image is any one of the at least one training magnetic resonance image.

[0211] The annotation image acquisition module is configured to, after spatially aligning the at least one training magnetic resonance image of each preset object under the at least one target magnetic resonance signal, take the annotation image corresponding to the target training magnetic resonance image as the annotation image corresponding to the at least one training magnetic resonance image of each preset object under the at least one target magnetic resonance signal.

[0212] As to the apparatus in the above-described embodiments, the specific manners in which various modules perform operations have been described in detail in the embodiments of the method, and thus will not be described herein.

[0213] Figure 5 is a block diagram of an electronic device for lesion segmentation according to an exemplary embodiment. The electronic device can be a server, and its internal structure diagram can be as shown in Figure 5 The electronic device includes a processor, a memory and a network interface connected through a system bus. The processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic 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 operating system and the computer program in the non-volatile storage medium to run. The network interface of the electronic device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement a lesion segmentation method.

[0214] Those skilled in the art can understand that Figure 5 the structure shown in the above is only a block diagram of part of the structure related to the present disclosure, and does not constitute a limitation on the electronic device to which the present disclosure is applied. The specific electronic device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0215] In an exemplary embodiment, an electronic device is also provided, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement a lesion segmentation method as in the embodiments of the present disclosure.

[0216] Those skilled in the art will readily understand that other embodiments of the present disclosure can be made in light of the description and practice of the invention disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known or customary practice in the art of the present disclosure. The specification and examples are to be regarded as illustrative only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0217] It should be understood that the present disclosure is not limited to the precise structures described and shown in the above and that various modifications and changes can be made without departing from its scope. The scope of the present disclosure is limited only by the claims that follow.

Claims

1. A method for segmenting lesions, characterized in that, The method comprises the following steps: determining at least one target magnetic resonance signal corresponding to a target object and at least one preset configuration, wherein the at least one target magnetic resonance signal is determined based on at least one preset magnetic resonance image of each preset object in at least one preset object under at least one preset magnetic resonance signal and the overlap degree of the at least one preset magnetic resonance image and a corresponding annotation image of the at least one preset magnetic resonance image, wherein the at least one preset magnetic resonance image is obtained by detecting a lesion to be segmented of the each preset object under the at least one preset magnetic resonance signal; the corresponding annotation image of the at least one preset magnetic resonance image is an image obtained by lesion annotation on any preset magnetic resonance image in the at least one preset magnetic resonance image; obtaining at least one target magnetic resonance image of the target object under the at least one target magnetic resonance signal; based on the at least one preset configuration, performing configuration recognition on the at least one target magnetic resonance image to determine at least one target configuration corresponding to the target object; based on the at least one target magnetic resonance signal and the at least one target configuration, determining at least one target segmentation model from at least one preset segmentation model, wherein the at least one preset segmentation model is a segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal; inputting the at least one target magnetic resonance image into the at least one target segmentation model for segmentation to obtain at least one lesion segmentation image; performing weighted processing on the at least one lesion segmentation image to obtain a target lesion segmentation image corresponding to the target object.

2. The method of claim 1, wherein, The configuration recognition on the at least one target magnetic resonance image based on the at least one preset configuration to determine at least one target configuration corresponding to the target object comprises: performing feature extraction processing on the at least one target magnetic resonance image to obtain at least one target feature information; obtaining at least one preset sample data, wherein the at least one preset sample data is obtained by fusing the at least one target magnetic resonance image and the at least one preset configuration; performing feature extraction processing on the at least one preset sample data to obtain at least one preset feature information; performing clustering processing on the at least one preset feature information and the at least one target feature information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image; based on first association information between each preset feature information in each clustering feature information set and each target feature information in each clustering feature information set, performing configuration recognition on the at least one target magnetic resonance image to obtain at least one target configuration corresponding to the target object.

3. The method of claim 2, wherein, The clustering processing on the at least one preset feature information and the at least one target feature information to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image comprises: determining second association information between each target feature information in the at least one target feature information and the at least one preset feature information; Based on the second association information, the at least one preset feature information and the at least one target feature information are clustered to obtain a clustering feature information set corresponding to each of the at least one target magnetic resonance image.

4. The method of claim 1, wherein, The method further comprises: obtaining a target labeled image corresponding to the at least one target magnetic resonance image; determining an overlap degree of the at least one target magnetic resonance image and the target labeled image based on the at least one target magnetic resonance image and the target labeled image; the weighting processing of the at least one lesion segmentation image to obtain the target lesion segmentation image corresponding to the target object comprises: weighting and merging the lesion segmentation images under the same magnetic resonance signal in the at least one lesion segmentation image to obtain at least one lesion segmentation image under the at least one target magnetic resonance signal; normalizing the overlap degree of the at least one target magnetic resonance image and the target labeled image to obtain a weight corresponding to each of the at least one lesion segmentation image under the at least one target magnetic resonance signal; based on the weight corresponding to each of the at least one lesion segmentation image under the at least one target magnetic resonance signal, weighting processing is performed on the at least one lesion segmentation image under the at least one target magnetic resonance signal to obtain the target lesion segmentation image corresponding to the target object.

5. The method of claim 4, wherein, The determination of the overlap degree of the at least one target magnetic resonance image and the target labeled image based on the at least one target magnetic resonance image and the target labeled image comprises: normalizing the at least one target magnetic resonance image to obtain a normalized at least one target magnetic resonance image; filtering the normalized at least one target magnetic resonance image to obtain at least one feature image; determining the overlap degree of the at least one target magnetic resonance image and the target labeled image based on the overlap degree of the at least one feature image and the target labeled image.

6. The method of claim 1, wherein, The method further comprises: based on the overlap degree of the at least one preset magnetic resonance image of each preset object in the at least one preset magnetic resonance signal and the corresponding labeled image of the at least one preset magnetic resonance image of each preset object in the at least one preset magnetic resonance signal, determining at least one to-be-selected magnetic resonance signal corresponding to each preset object; based on the at least one to-be-selected magnetic resonance signal corresponding to each preset object, determining at least one target magnetic resonance signal corresponding to the target object.

7. The lesion segmentation method of any one of claims 1 to 6, wherein, The method further comprises: obtaining a training image set and a preset training model; the training image set comprises at least one to-be-trained magnetic resonance image of each preset object in the at least one target magnetic resonance signal and a labeled image corresponding to the at least one to-be-trained magnetic resonance image of each preset object in the at least one target magnetic resonance signal; based on the training image set, training the preset training model to obtain the at least one preset segmentation model.

8. The method of claim 7, wherein, The method further comprises: obtaining at least one training magnetic resonance image and a corresponding label image of a target training magnetic resonance image of each preset object under at least one target magnetic resonance signal, the target training magnetic resonance image being any one of the at least one training magnetic resonance image; after spatially aligning the at least one training magnetic resonance image of each preset object under the at least one target magnetic resonance signal, taking the corresponding label image of the target training magnetic resonance image as the corresponding label image of the at least one training magnetic resonance image of each preset object under at least one preset magnetic resonance signal.

9. A lesion segmentation apparatus characterized by comprising: comprising: a first determination module configured to determine at least one target magnetic resonance signal corresponding to a target object and at least one preset configuration, the at least one target magnetic resonance signal being determined based on an overlap degree of at least one preset magnetic resonance image of each preset object in at least one preset magnetic resonance signal and a corresponding label image of the at least one preset magnetic resonance image of each preset object under at least one preset magnetic resonance signal; the at least one preset magnetic resonance image being obtained by detecting a to-be-segmented lesion of each preset object through the at least one preset magnetic resonance signal; the corresponding label image of the at least one preset magnetic resonance image being an image obtained by lesion labeling on any preset magnetic resonance image in the at least one preset magnetic resonance image; a target magnetic resonance image acquisition module configured to obtain at least one target magnetic resonance image of the target object under the at least one target magnetic resonance signal; a second determination module configured to perform configuration recognition on the at least one target magnetic resonance image based on the at least one preset configuration, and determine at least one target configuration corresponding to the target object; a third determination module configured to determine at least one target segmentation model from at least one preset segmentation model based on the at least one target magnetic resonance signal and the at least one target configuration; the at least one preset segmentation model being a segmentation model corresponding to the at least one preset configuration under the at least one preset magnetic resonance signal; a lesion segmentation image acquisition module configured to input the at least one target magnetic resonance image into the at least one target segmentation model for segmentation to obtain at least one lesion segmentation image; a target lesion segmentation image acquisition module configured to perform weighted processing on the at least one lesion segmentation image to obtain a target lesion segmentation image corresponding to the target object.

10. An electronic device, comprising: comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the lesion segmentation method according to any one of claims 1 to 8.

Images