CN117111696A - A medical image segmentation method and a training method for a medical image segmentation model - Google Patents

Abstract

The application provides a medical image segmentation method and a training method of a medical image segmentation model, wherein the medical image segmentation method comprises the following steps: acquiring a medical image to be processed; inputting the medical image to be processed into a medical image segmentation model trained in advance to obtain a segmentation result output by the medical image segmentation model; the medical image segmentation model comprises a natural image segmentation module, wherein the natural image segmentation module comprises an image encoder, an image decoder and a segmentation adapter, the segmentation adapter is embedded into the image encoder, and the image encoder is trained by updating parameters of the segmentation adapter. Therefore, the segmentation adapter can be introduced on the basis of the natural image segmentation module, so that the backbone network (namely the image encoder) can be frozen in the process of training the medical image segmentation model, and only a small amount of medical segmentation image data is utilized to train the segmentation adapter, so that the trained medical image segmentation model is obtained.

Description

A medical image segmentation method and a training method for a medical image segmentation model

Technical field

The present application relates to the field of image processing technology, specifically, to a medical image segmentation method and a training method of a medical image segmentation model.

Background technique

With the development of medical technology, the segmentation accuracy of medical image segmentation models is getting higher and higher. Take thyroid cancer as an example. In recent years, the incidence rate of thyroid cancer has been rising rapidly. The results of the 2020 Global Cancer Survey show that the incidence of thyroid cancer is more common in women than in men, and in urban areas than in rural areas.

Among them, ultrasound is the preferred examination method for thyroid lesions. It can detect lesions and make preliminary judgments on their biological behavior at the same time. It has the advantages of convenience and safety. The rapid development of science and technology has led to the widespread application of artificial intelligence (AI) technology in ultrasound images of medical big data, and its advantages are obvious. Under the pressure of daily overloaded workload and complex and high-risk examinations, the ultrasound AI system can optimize the examination process, standardize diagnostic standards, shorten examination and reporting time, and significantly improve the diagnostic confidence and work efficiency of ultrasound doctors. Ultrasound image segmentation of thyroid nodules based on artificial intelligence can assist doctors to locate the nodule location more quickly and confirm the nodule shape to facilitate the judgment of benign and malignant nodules. It is foreseeable that this technology will have broad innovation and development prospects in assisting ultrasound diagnosis and treatment technology, talent training and other aspects in the future.

Therefore, a more accurate method for segmenting ultrasound images of thyroid nodules is needed, which can assist doctors in making quick and accurate diagnoses during examinations. Extended to other medical images, similar to thyroid images and nodule ultrasound images, a more accurate segmentation method is also needed.

In the existing technology, the accuracy of segmentation is generally improved by improving the architecture of the medical image segmentation model or performing data enhancement. However, due to the privacy of medical data and the professionalism of data annotation, there is a lack of large-scale and high-quality medical segmentation image data, resulting in low accuracy of segmentation models.

Contents of the invention

The purpose of the embodiments of this application is to provide a medical image segmentation method and a medical image segmentation model training method to solve the problem of lack of large-scale and high-quality medical data in the existing technology due to the privacy of medical data and the professionalism of data annotation. Technical problems in segmenting image data, resulting in lower accuracy of segmentation models.

In a first aspect, embodiments of the present application provide a medical image segmentation method, which includes: acquiring a medical image to be processed; inputting the medical image to be processed into a pre-trained medical image segmentation model to obtain the output of the medical image segmentation model segmentation results; wherein, the medical image segmentation model includes a natural image segmentation module, the natural image segmentation module includes an image encoder, an image decoder, and a segmentation adapter, and the segmentation adapter is embedded in the image encoder. The parameters of the segmentation adapter are updated to implement training of the image encoder.

In the above solution, a pre-trained medical image segmentation model can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, the above-mentioned medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e., image encoder) can be frozen , only use a small amount of medical segmentation image data to train the segmentation adapter, and obtain a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In an optional embodiment, inputting the medical image to be processed into a pre-trained medical image segmentation model to obtain a segmentation result output by the medical image segmentation model includes: inputting the medical image to be processed In the image encoder, the encoding result output by the image encoder is obtained; the encoding result is input into the image decoder, and the segmentation result output by the image decoder is obtained. In the above solution, the natural image segmentation module can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, because the natural image segmentation module has powerful segmentation capabilities, the entire medical image segmentation model can achieve higher-precision segmentation effects in medical images.

In an optional embodiment, the medical image segmentation method further includes: training the medical image segmentation model using the following steps: obtaining medical segmentation image data and the medical image segmentation model to be trained; segmenting the natural image The module loads corresponding pre-training parameters; uses the medical segmentation image data to update the parameters of the segmentation adapter and the image decoder in the medical image segmentation model to obtain a trained medical image segmentation model. In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, only a small amount of medical segmentation image data is used to train the segmentation adapter, and the image decoder is trained at the same time, thereby obtaining the trained medical image segmentation Model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In an optional embodiment, the medical image segmentation model further includes: a spatial multi-scale information feature extraction module; the medical image to be processed is input into a pre-trained medical image segmentation model to obtain the medical image. The segmentation result output by the segmentation model includes: inputting the medical image to be processed into the spatial multi-scale information feature extraction module to obtain the feature data output by the spatial multi-scale information feature extraction module; inputting the medical image to be processed and The characteristic data is input into the image encoder to obtain the encoding result output by the image encoder; the encoding result is input into the image decoder to obtain the segmentation result output by the image decoder. In the above solution, a spatial multi-scale information feature extraction module can be introduced. Before image segmentation of the medical image to be processed, multi-scale spatial information is transferred to the backbone network based on the medical image to be processed, thereby improving the trained medical image segmentation model. segmentation accuracy.

In an optional embodiment, the medical image segmentation method further includes: training the medical image segmentation model using the following steps: obtaining medical segmentation image data and the medical image segmentation model to be trained; segmenting the natural image The module loads the corresponding pre-training parameters; uses the medical segmentation image data to update the parameters of the segmentation adapter, the image decoder and the spatial multi-scale information feature extraction module in the medical image segmentation model to obtain a well-trained Medical image segmentation model. In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, using only a small amount of medical segmentation image data to train the segmentation adapter, and training the image decoder and spatial multi-scale information feature extraction module at the same time. Thus, the trained medical image segmentation model is obtained. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In a second aspect, embodiments of the present application provide a training method for a medical image segmentation model, which includes: obtaining medical segmentation image data and a medical image segmentation model to be trained, wherein the medical image segmentation model includes a natural image segmentation module, so The natural image segmentation module includes an image encoder, an image decoder and a segmentation adapter, and the segmentation adapter is embedded in the image encoder; corresponding pre-training parameters are loaded for the natural image segmentation module; and the medical segmentation image data is used The parameters of the segmentation adapter and the image decoder in the medical image segmentation model are updated to obtain a trained medical image segmentation model.

In the above solution, the above medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e. image Encoder), only uses a small amount of medical segmentation image data to train the segmentation adapter, and obtains a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In an optional embodiment, the medical image segmentation model further includes: a spatial multi-scale information feature extraction module, which uses the medical segmentation image data to decode the segmentation adapter and the image in the medical image segmentation model. The parameters of the medical image segmentation model are updated to obtain a trained medical image segmentation model, which includes: using the medical segmentation image data to extract features of the segmentation adapter, the image decoder and the spatial multi-scale information in the medical image segmentation model. The parameters of the module are updated to obtain the trained medical image segmentation model.

In a third aspect, embodiments of the present application provide a medical image segmentation device, including: a first acquisition module for acquiring medical images to be processed; and an input module for inputting the medical images to be processed into pre-trained medical images. In the segmentation model, the segmentation result output by the medical image segmentation model is obtained; wherein the medical image segmentation model includes a natural image segmentation module, and the natural image segmentation module includes an image encoder, an image decoder, and a segmentation adapter, and the A segmentation adapter is embedded in the image encoder, and the image encoder is trained by updating parameters of the segmentation adapter.

In the above solution, a pre-trained medical image segmentation model can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, the above-mentioned medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e., image encoder) can be frozen , only use a small amount of medical segmentation image data to train the segmentation adapter, and obtain a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In an optional implementation, the input module is specifically configured to: input the medical image to be processed into the image encoder to obtain the encoding result output by the image encoder; input the encoding result into the In the image decoder, the segmentation result output by the image decoder is obtained. In the above solution, the natural image segmentation module can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, because the natural image segmentation module has powerful segmentation capabilities, the entire medical image segmentation model can achieve higher-precision segmentation effects in medical images.

In an optional embodiment, the medical image segmentation device further includes: a first training module for training the medical image segmentation model using the following steps: obtaining medical segmentation image data and the medical image segmentation model to be trained ; Load corresponding pre-training parameters for the natural image segmentation module; use the medical segmentation image data to update the parameters of the segmentation adapter and the image decoder in the medical image segmentation model to obtain trained medical images Segmentation model. In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, only a small amount of medical segmentation image data is used to train the segmentation adapter, and the image decoder is trained at the same time, thereby obtaining the trained medical image segmentation Model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In an optional implementation, the medical image segmentation model further includes: a spatial multi-scale information feature extraction module; the input module is specifically configured to: input the medical image to be processed into the spatial multi-scale information feature extraction module , obtain the feature data output by the spatial multi-scale information feature extraction module; input the medical image to be processed and the feature data into the image encoder to obtain the encoding result output by the image encoder; convert the The encoding result is input into the image decoder, and the segmentation result output by the image decoder is obtained. In the above solution, a spatial multi-scale information feature extraction module can be introduced. Before image segmentation of the medical image to be processed, multi-scale spatial information is transferred to the backbone network based on the medical image to be processed, thereby improving the trained medical image segmentation model. segmentation accuracy.

In an optional embodiment, the medical image segmentation device further includes: a second training module for training the medical image segmentation model using the following steps: obtaining medical segmentation image data and the medical image segmentation model to be trained ; Load corresponding pre-training parameters for the natural image segmentation module; use the medical segmentation image data to perform segmentation adapter, the image decoder and the spatial multi-scale information feature extraction module in the medical image segmentation model. The parameters are updated to obtain the trained medical image segmentation model. In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, using only a small amount of medical segmentation image data to train the segmentation adapter, and training the image decoder and spatial multi-scale information feature extraction module at the same time. Thus, the trained medical image segmentation model is obtained. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In a fourth aspect, embodiments of the present application provide a training device for a medical image segmentation model, including: a second acquisition module for acquiring medical segmentation image data and a medical image segmentation model to be trained, wherein the medical image segmentation model It includes a natural image segmentation module, which includes an image encoder, an image decoder, and a segmentation adapter. The segmentation adapter is embedded in the image encoder. A loading module is used to load the corresponding natural image segmentation module for the natural image segmentation module. Pre-training parameters; an update module, configured to use the medical segmentation image data to update the parameters of the segmentation adapter and the image decoder in the medical image segmentation model to obtain a trained medical image segmentation model.

In the above solution, the above medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e. image Encoder), only uses a small amount of medical segmentation image data to train the segmentation adapter, and obtains a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

In an optional embodiment, the medical image segmentation model further includes: a spatial multi-scale information feature extraction module, and the update module is specifically configured to: use the medical segmentation image data to perform segmentation in the medical image segmentation model. The parameters of the adapter, the image decoder and the spatial multi-scale information feature extraction module are updated to obtain a trained medical image segmentation model.

In a fifth aspect, embodiments of the present application provide an electronic device, including: a processor, a memory and a bus; the processor and the memory complete communication with each other through the bus; the memory stores information that can be used by the Computer program instructions executed by a processor, which can execute the medical image segmentation method as described in the first aspect or the training method of the medical image segmentation model as described in the second aspect by calling the computer program instructions.

In a sixth aspect, embodiments of the present application provide a computer-readable storage medium. The computer-readable storage medium stores computer program instructions. When the computer program instructions are run by a computer, they cause the computer to execute as described in the first aspect. The medical image segmentation method or the training method of the medical image segmentation model as described in the second aspect.

In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, embodiments of the present application are cited below and described in detail with reference to the attached drawings.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, therefore This should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of a medical image segmentation method provided by an embodiment of the present application;

Figure 2 is a schematic diagram of a ViT structure provided by an embodiment of the present application;

Figure 3 is a schematic diagram of a ViT+Adaper structure provided by an embodiment of the present application;

Figure 4 is a schematic diagram of an SMS structure provided by an embodiment of the present application;

Figure 5 is a flow chart of a training method for a medical image segmentation model provided by an embodiment of the present application;

Figure 6 is a structural block diagram of a medical image segmentation device according to an embodiment of the present application;

Figure 7 is a structural block diagram of a training device for a medical image segmentation model provided by an embodiment of the present application;

FIG. 8 is a structural block diagram of an electronic device provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

Please refer to Figure 1. Figure 1 is a flow chart of a medical image segmentation method provided by an embodiment of the present application. The medical image segmentation method may include the following steps:

Step S101: Obtain the medical image to be processed.

Step S102: Input the medical image to be processed into the pre-trained medical image segmentation model, and obtain the segmentation result output by the medical image segmentation model.

Specifically, in the above step S101, the medical image to be processed refers to the medical image that currently requires image segmentation. Medical images refer to images used in the medical industry. It should be noted that the embodiments of this application do not specifically limit the specific implementation of medical images. Those skilled in the art can make appropriate adjustments according to the actual situation, for example: medical images It can include thyroid ultrasound images, breast ultrasound images, carotid angiography images, etc.

In addition, the embodiments of the present application also do not specifically limit the specific implementation methods for obtaining the above-mentioned medical images to be processed, and those skilled in the art can also make appropriate adjustments according to the actual situation. For example, medical images to be processed from an external device can be received; or medical images to be processed stored locally or in the cloud can be read; or medical images to be processed can be collected in real time.

In the above-mentioned step S102, the medical image segmentation model refers to a neural network model used for image segmentation of the above-mentioned medical image to be processed, wherein the medical image segmentation model in the above-mentioned step S102 is a pre-trained medical image segmentation model.

The structure of a medical image segmentation model provided by the embodiment of the present application is introduced below. The medical image segmentation model may include a natural image segmentation module (Segment Anything Model, SAM); the natural image segmentation module may include an image encoder, an image decoder, and a segmentation adapter, where the segmentation adapter may be embedded in the image encoder.

In the above natural image segmentation module, the image encoder can be based on the standard Vision Transformer (ViT), and the attention mechanism can be composed of a 14*14 size window attention and four equally spaced global attentions; image The decoder is a Transformer-based decoder, including a dynamic prediction head that can handle mask information for different hints (text, points, boxes, etc.).

Please refer to Figure 2, which is a schematic diagram of a ViT structure provided by an embodiment of the present application. The ViT structure includes layer normalization (Layer Norm), multi-head attention (Muti-Head Attention) and multi-layer perceptron (MultilayerPerceptron, MLP). Among them, Layer Norm is used to perform Norm processing on each Token; MLP can include fully connected layers, GELU activation functions and Dropout.

On the ViT structure shown in Figure 2, a split adapter (Adaper) can be embedded. As an implementation manner, please refer to FIG. 3 , which is a schematic diagram of a ViT+Adaper structure provided by an embodiment of the present application. The ViT+Adaper structure is based on Figure 2 and introduces Adaper in the Muti-Head Attention layer and MLP layer respectively.

According to current research, partial parameter fine-tuning methods are more effective than full fine-tuning because they can avoid catastrophic forgetting and generalize better to out-of-domain scenarios, especially in the case of low data. Adapter serves as an effective tool not only in natural language processing but also in computer vision, and can be applied to fine-tune large basic vision models for downstream tasks. Considering that high-quality medical data is difficult to obtain and is not easy to label, Adapter is introduced in the embodiment of this application.

As another implementation, Scaling hyperparameters can also be introduced in the MLP layer to achieve better channel weight learning.

It should be noted that the embodiments of the present application do not specifically limit the specific structure of the Adapter, and those skilled in the art can make appropriate adjustments according to the actual situation. For example, as shown in Figure 3, the Adapter can include a Down layer, a ReLU layer, and an Up layer.

Based on the above ViT+Adaper structure, in the process of training the medical image segmentation model, the image encoder can be trained by updating the parameters of the segmentation adapter.

As an implementation manner, considering that the natural image segmentation module is a model trained by tens of millions of natural images, ultra-high-precision image segmentation can be achieved in natural image segmentation. Therefore, in the embodiment of the present application, natural image segmentation Modules can be pre-loaded with pre-trained parameters.

During the training process of the medical image segmentation model, the backbone network (i.e., the image encoder in the natural image segmentation module) can be frozen and only the Adapter part is trained. In other words, during the training process of the medical image segmentation model, the parameters of the image encoder will not change, and the parameters of the Adapter will be iteratively updated until the trained medical image segmentation model is obtained.

It can be understood that because the natural image segmentation module is pre-loaded with pre-training parameters with higher accuracy when segmenting natural images, the segmentation capabilities of the medical image segmentation model trained on this basis can be improved; in addition, since only The Adapter part is trained. Therefore, it is only necessary to use a small amount of medical segmentation image data to train the medical image segmentation model to obtain a medical image segmentation model with higher accuracy, thereby solving the problem of privacy and privacy of medical data in the existing technology. The professionalism of data annotation and the lack of large-scale and high-quality medical segmentation image data lead to technical problems such as low accuracy of segmentation models.

It should be noted that the natural image segmentation module in the prior art generally includes an image encoder, an image decoder and a mask decoder. However, in the embodiment of the present application, since only the image encoding and image decoding parts are sampled, there is no need to Dynamic mask input, therefore, the mask decoder can be omitted.

In addition, the embodiment of the present application does not specifically limit the specific implementation of the segmentation result in the above step S102, and those skilled in the art can make appropriate adjustments according to the actual situation. For example, when the medical image includes a thyroid ultrasound image, the segmentation result may include the location of the thyroid nodule; or when the medical image includes a carotid angiography image, the segmentation result may include the location of the carotid artery plaque, etc.

In the above solution, a pre-trained medical image segmentation model can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, the above-mentioned medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e., image encoder) can be frozen , only use a small amount of medical segmentation image data to train the segmentation adapter, and obtain a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Further, based on the above embodiment, the above step S102 may specifically include the following steps:

Step 1): Input the medical image to be processed into the image encoder to obtain the encoding result output by the image encoder.

Step 2): Input the encoding result into the image decoder to obtain the segmentation result output by the image decoder.

Specifically, in the above step 1), the image encoder can be based on the standard visual transformer (VisionTransformer, ViT), and the attention mechanism can be composed of a 14*14 size window attention and four equally spaced global attentions. . As shown in Figure 3, the medical image to be processed can be input into the above-mentioned image encoder. The medical image to be processed passes through Layer Norm, Muti-Head Attention, Adaper and MLP, and finally the encoding result output by the image encoder can be obtained.

As an implementation manner, the medical image to be processed can be down-sampled 16 times and used as the input of the image encoder.

In step 2) above, the image decoder is a Transformer-based decoder, including a dynamic prediction head that can handle mask information for different cues (text, points, boxes, etc.). The encoding result can be input into the above image decoder, and finally the segmentation result output by the image decoder can be obtained.

In the above solution, the natural image segmentation module can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, because the natural image segmentation module has powerful segmentation capabilities, the entire medical image segmentation model can achieve higher-precision segmentation effects in medical images.

Further, based on the above embodiments, the following steps can be used to train the medical image segmentation model in the above embodiments:

Step 1), obtain medical segmentation image data and the medical image segmentation model to be trained.

Step 2), load the corresponding pre-training parameters for the natural image segmentation module.

Step 3), use the medical segmentation image data to update the parameters of the segmentation adapter and image decoder in the medical image segmentation model to obtain a trained medical image segmentation model.

Specifically, in the above step 1), the medical segmentation image data refers to the medical images that have been segmented and labeled. It can be understood that since Adaper is introduced in the embodiment of the present application, the amount of the above medical segmentation image data does not need to be too large, that is, only a small amount of medical segmentation image data is needed to train the medical image segmentation model.

It should be noted that the embodiments of the present application also do not specifically limit the specific implementation manner for obtaining the above-mentioned medical segmentation image data, and those skilled in the art can also make appropriate adjustments according to the actual situation. For example, medical segmentation image data from an external device can be received; or medical segmentation image data stored locally or in the cloud can be read.

The medical image segmentation model refers to the neural network model used for image segmentation of the above-mentioned medical image to be processed, wherein the medical image segmentation model in the above step 1) refers to an untrained or untrained medical image segmentation model. It can be understood that the medical image segmentation model in the above step 1) has the same structure as the medical image segmentation model in the above step S102, and the only difference between them is that their model parameters are different.

Similar to obtaining the above-mentioned medical segmentation image data, the embodiments of the present application also do not specifically limit the specific implementation of obtaining the above-mentioned medical image segmentation model to be trained. Those skilled in the art can also make appropriate adjustments according to the actual situation. For example, a medical image segmentation model to be trained from an external device can be received; or a medical image segmentation model to be trained stored locally or in the cloud can be read.

In the above step 2), considering that the natural image segmentation module is a model trained by tens of millions of natural images, ultra-high-precision image segmentation can be achieved in natural image segmentation. Therefore, the corresponding module can be loaded for the natural image segmentation module. Pre-training parameters.

In the above step 3), during the training process of the medical image segmentation model, the backbone network (i.e., the image encoder in the natural image segmentation module) can be frozen, and the medical segmentation image data is used to train the segmentation adapter in the medical image segmentation model and The parameters of the image decoder are updated to obtain the trained medical image segmentation model.

In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, only a small amount of medical segmentation image data is used to train the segmentation adapter, and the image decoder is trained at the same time, thereby obtaining the trained medical image segmentation Model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Further, based on the above embodiments, the medical image segmentation model may also include: a spatial multi-scale information feature extraction module (Spatial Multi-Scale, SMS). Please refer to Figure 4, which is a schematic diagram of an SMS structure provided by an embodiment of the present application. The spatial multi-scale information feature extraction module can adopt a standard convolution system borrowed from ResNet.

As an implementation method, the spatial multi-scale information feature extraction module can include three convolutions and a maximum pooling layer; then a 3×3 convolution layer is used to double the number of channels and reduce the size of the feature map ; Then apply multiple 1×1 convolutional layers to project the feature map to D dimension; in this way, a feature pyramid {F1, F2, F3, F4} containing D-dimensional feature maps is obtained, with resolutions of the original image. 1/8, 1/16, 1/32, 1/64; then the above feature pyramid containing the D-dimensional feature map is mapped and flattened, and spliced into new features as feature interactions to obtain the spatial multi-scale information feature extraction module. output.

The spatial multi-scale information feature extraction module is used to transfer multi-scale spatial information to the natural image segmentation module. Among them, in order not to change the original structure of the natural image segmentation module, the spatial multi-scale information feature extraction module can be used in parallel with the patch embedding layer to model the local spatial context of the image. Therefore, the output of the spatial multi-scale information feature extraction module can be input to the subsequent segmentation adapter, which then performs operations together with the i-th image encoder module.

Further, based on the above embodiment, the above step S102 may specifically include the following steps:

Step 1): Input the medical image to be processed into the spatial multi-scale information feature extraction module to obtain the feature data output by the spatial multi-scale information feature extraction module.

Step 2): Input the medical image to be processed and the feature data into the image encoder to obtain the encoding result output by the image encoder.

Step 3): Input the encoding result into the image decoder to obtain the segmentation result output by the image decoder.

Specifically, in the above step 1), the spatial multi-scale information feature extraction module can sequentially include three convolutions, a maximum pooling layer, a 3×3 convolution layer, and multiple 1×1 convolution layers, so that A feature pyramid {F1, F2, F3, F4} containing D-dimensional feature maps is obtained, with resolutions of 1/8, 1/16, 1/32, and 1/64 of the original image respectively; then the above containing D The feature pyramid of the dimensional feature map is mapped and flattened, and spliced into new features as feature interactions to obtain the feature data output by the spatial multi-scale information feature extraction module.

In the above step 2), the image encoder can be based on the standard Vision Transformer (ViT), and the attention mechanism can be composed of a 14*14 size window attention and four equally spaced global attentions. As shown in Figure 3, the medical image and feature data to be processed can be input into the above-mentioned image encoder. The medical image and feature data to be processed go through Layer Norm, Muti-Head Attention, Adaper and MLP, and finally the output of the image encoder can be obtained. Encoding results.

Among them, the medical image to be processed can be input into the Layer Norm in the image encoder, and the feature data can be input into the Adaper in the image encoder.

In step 3) above, the image decoder is a Transformer-based decoder, including a dynamic prediction head that can handle mask information for different cues (text, points, boxes, etc.). The encoding result can be input into the above image decoder, and finally the segmentation result output by the image decoder can be obtained.

In the above solution, a spatial multi-scale information feature extraction module can be introduced. Before image segmentation of the medical image to be processed, multi-scale spatial information is transferred to the backbone network based on the medical image to be processed, thereby improving the trained medical image segmentation model. segmentation accuracy.

Further, based on the above embodiments, the following steps can be used to train the medical image segmentation model:

Step 1), obtain medical segmentation image data and the medical image segmentation model to be trained.

Step 2), load the corresponding pre-training parameters for the natural image segmentation module.

Step 3), use the medical segmentation image data to update the parameters of the segmentation adapter, image decoder and spatial multi-scale information feature extraction module in the medical image segmentation model to obtain a trained medical image segmentation model.

Specifically, in the above-mentioned step 1), similar to the above-mentioned embodiments, the embodiments of the present application also do not specifically limit the specific implementation methods of obtaining the above-mentioned medical segmentation image data and the medical image segmentation model to be trained. Those skilled in the art will also Appropriate adjustments can be made according to the actual situation. For example, the medical segmentation image data from an external device and the medical image segmentation model to be trained can be received; or the medical segmentation image data stored locally or in the cloud and the medical image segmentation model to be trained can be read.

In the above step 2), considering that the natural image segmentation module is a model trained by tens of millions of natural images, ultra-high-precision image segmentation can be achieved in natural image segmentation. Therefore, the corresponding module can be loaded for the natural image segmentation module. Pre-training parameters.

In the above step 3), during the training process of the medical image segmentation model, the backbone network (i.e., the image encoder in the natural image segmentation module) can be frozen, and the medical segmentation image data is used to train the segmentation adapter, The parameters of the image decoder and spatial multi-scale information feature extraction module are updated to obtain the trained medical image segmentation model.

In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, using only a small amount of medical segmentation image data to train the segmentation adapter, and training the image decoder and spatial multi-scale information feature extraction module at the same time. Thus, the trained medical image segmentation model is obtained. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Please refer to Figure 5. Figure 5 is a flow chart of a training method for a medical image segmentation model provided by an embodiment of the present application. The training method for the medical image segmentation model may include the following steps:

Step S501: Obtain medical segmentation image data and the medical image segmentation model to be trained.

Step S502: Load corresponding pre-training parameters for the natural image segmentation module.

Step S503: Use the medical segmentation image data to update the parameters of the segmentation adapter and the image decoder in the medical image segmentation model to obtain a trained medical image segmentation model.

Specifically, in the above step S501, the medical segmentation image data refers to the medical images that have been segmented and labeled. It can be understood that since Adaper is introduced in the embodiment of the present application, the amount of the above medical segmentation image data does not need to be too large, that is, only a small amount of medical segmentation image data is needed to train the medical image segmentation model.

It should be noted that the embodiments of the present application also do not specifically limit the specific implementation manner for obtaining the above-mentioned medical segmentation image data, and those skilled in the art can also make appropriate adjustments according to the actual situation. For example, medical segmentation image data from an external device can be received; or medical segmentation image data stored locally or in the cloud can be read.

The medical image segmentation model refers to a neural network model used for image segmentation of the above-mentioned medical image to be processed, wherein the medical image segmentation model in the above-mentioned step S501 refers to an untrained or untrained medical image segmentation model. As an implementation manner, the medical image segmentation model may include a natural image segmentation module. The natural image segmentation module includes an image encoder, an image decoder, and a segmentation adapter. The segmentation adapter may be embedded in the image encoder.

Similar to obtaining the above-mentioned medical segmentation image data, the embodiments of the present application also do not specifically limit the specific implementation of obtaining the above-mentioned medical image segmentation model to be trained. Those skilled in the art can also make appropriate adjustments according to the actual situation. For example, a medical image segmentation model to be trained from an external device can be received; or a medical image segmentation model to be trained stored locally or in the cloud can be read.

In the above step S502, considering that the natural image segmentation module is a model trained by tens of millions of natural images, ultra-high-precision image segmentation can be achieved in natural image segmentation. Therefore, the corresponding presets can be loaded for the natural image segmentation module. training parameters.

In the above step S503, during the training process of the medical image segmentation model, the backbone network (i.e., the image encoder in the natural image segmentation module) can be frozen, and the medical segmentation image data can be used to train the segmentation adapter and image in the medical image segmentation model. The parameters of the decoder are updated to obtain the trained medical image segmentation model.

In the above solution, the above medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e. image Encoder), only uses a small amount of medical segmentation image data to train the segmentation adapter, and obtains a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Furthermore, based on the above embodiments, the medical image segmentation model may also include: a spatial multi-scale information feature extraction module. The embodiments of this application introduce another training method for the medical image segmentation model:

Step 1), obtain medical segmentation image data and the medical image segmentation model to be trained.

Step 2), load the corresponding pre-training parameters for the natural image segmentation module.

Step 3), use the medical segmentation image data to update the parameters of the segmentation adapter, image decoder and spatial multi-scale information feature extraction module in the medical image segmentation model to obtain a trained medical image segmentation model.

Specifically, in the above-mentioned step 1), similar to the above-mentioned embodiments, the embodiments of the present application also do not specifically limit the specific implementation methods of obtaining the above-mentioned medical segmentation image data and the medical image segmentation model to be trained. Those skilled in the art will also Appropriate adjustments can be made according to the actual situation. For example, the medical segmentation image data from an external device and the medical image segmentation model to be trained can be received; or the medical segmentation image data stored locally or in the cloud and the medical image segmentation model to be trained can be read.

In the above step 2), considering that the natural image segmentation module is a model trained by tens of millions of natural images, ultra-high-precision image segmentation can be achieved in natural image segmentation. Therefore, the corresponding module can be loaded for the natural image segmentation module. Pre-training parameters.

In the above step 3), during the training process of the medical image segmentation model, the backbone network (i.e., the image encoder in the natural image segmentation module) can be frozen, and the medical segmentation image data is used to train the segmentation adapter, The parameters of the image decoder and spatial multi-scale information feature extraction module are updated to obtain the trained medical image segmentation model.

Please refer to Figure 6. Figure 6 is a structural block diagram of a medical image segmentation device according to an embodiment of the present application. The medical image segmentation device 600 includes: a first acquisition module 601 for acquiring medical images to be processed; an input module 602 for After inputting the medical image to be processed into a pre-trained medical image segmentation model, a segmentation result output by the medical image segmentation model is obtained; wherein the medical image segmentation model includes a natural image segmentation module, and the natural image segmentation The module includes an image encoder, an image decoder, and a segmentation adapter. The segmentation adapter is embedded in the image encoder. The image encoder is trained by updating parameters of the segmentation adapter.

In the above solution, a pre-trained medical image segmentation model can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, the above-mentioned medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e., image encoder) can be frozen , only use a small amount of medical segmentation image data to train the segmentation adapter, and obtain a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Further, on the basis of the above embodiments, the input module 602 is specifically configured to: input the medical image to be processed into the image encoder to obtain the encoding result output by the image encoder; The result is input into the image decoder, and the segmentation result output by the image decoder is obtained.

In the above solution, the natural image segmentation module can be used to segment the medical image to be processed, thereby obtaining the corresponding segmentation result. Among them, because the natural image segmentation module has powerful segmentation capabilities, the entire medical image segmentation model can achieve higher-precision segmentation effects in medical images.

Further, on the basis of the above embodiments, the medical image segmentation device 600 also includes: a first training module for training the medical image segmentation model using the following steps: obtaining medical segmentation image data and to-be-trained Medical image segmentation model; load corresponding pre-training parameters for the natural image segmentation module; use the medical segmentation image data to update the parameters of the segmentation adapter and the image decoder in the medical image segmentation model to obtain training Good medical image segmentation model.

In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, only a small amount of medical segmentation image data is used to train the segmentation adapter, and the image decoder is trained at the same time, thereby obtaining the trained medical image segmentation Model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Further, based on the above embodiments, the medical image segmentation model also includes: a spatial multi-scale information feature extraction module; the input module 602 is specifically used to: input the medical image to be processed into the spatial multi-scale The information feature extraction module obtains the feature data output by the spatial multi-scale information feature extraction module; inputs the medical image to be processed and the feature data into the image encoder to obtain the encoding result output by the image encoder ; Input the encoding result into the image decoder to obtain the segmentation result output by the image decoder.

In the above solution, a spatial multi-scale information feature extraction module can be introduced. Before image segmentation of the medical image to be processed, multi-scale spatial information is transferred to the backbone network based on the medical image to be processed, thereby improving the trained medical image segmentation model. segmentation accuracy.

Further, on the basis of the above embodiments, the medical image segmentation device 600 also includes: a second training module for training the medical image segmentation model using the following steps: obtaining medical segmentation image data and to-be-trained Medical image segmentation model; load corresponding pre-training parameters for the natural image segmentation module; use the medical segmentation image data to segment the segmentation adapter, the image decoder and the spatial multi-scale information in the medical image segmentation model The parameters of the feature extraction module are updated to obtain the trained medical image segmentation model.

In the above solution, during the training process of the medical image segmentation model, the image encoder can be frozen, using only a small amount of medical segmentation image data to train the segmentation adapter, and training the image decoder and spatial multi-scale information feature extraction module at the same time. Thus, the trained medical image segmentation model is obtained. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Please refer to Figure 7. Figure 7 is a structural block diagram of a training device 700 for a medical image segmentation model provided by an embodiment of the present application. The training device 700 for the medical image segmentation model includes: a second acquisition module 701 for acquiring medical segmentation images. data and a medical image segmentation model to be trained, wherein the medical image segmentation model includes a natural image segmentation module, the natural image segmentation module includes an image encoder, an image decoder and a segmentation adapter, the segmentation adapter is embedded in the image In the encoder; the loading module 702 is used to load the corresponding pre-training parameters for the natural image segmentation module; the update module 703 is used to use the medical segmentation image data to segment the segmentation adapter and the segmentation adapter in the medical image segmentation model. The parameters of the above image decoder are updated to obtain the trained medical image segmentation model.

In the above solution, the above medical image segmentation model includes a natural image segmentation module, and a segmentation adapter can be introduced based on the natural image segmentation module. In this way, during the training process of the medical image segmentation model, the backbone network (i.e. image Encoder), only uses a small amount of medical segmentation image data to train the segmentation adapter, and obtains a trained medical image segmentation model. Therefore, even if there is a lack of large-scale and high-quality medical segmentation image data, a medical image segmentation model with higher segmentation accuracy can be trained through the above method.

Further, based on the above embodiments, the medical image segmentation model also includes: a spatial multi-scale information feature extraction module, and the update module 703 is specifically used to: use the medical segmentation image data to segment the medical image. The parameters of the segmentation adapter, the image decoder and the spatial multi-scale information feature extraction module in the model are updated to obtain a trained medical image segmentation model.

Please refer to Figure 8. Figure 8 is a structural block diagram of an electronic device provided by an embodiment of the present application. The electronic device 800 includes: at least one processor 801, at least one communication interface 802, at least one memory 803 and at least one communication bus 804. . Among them, the communication bus 804 is used to realize direct connection communication between these components, the communication interface 802 is used to communicate signaling or data with other node devices, and the memory 803 stores machine-readable instructions executable by the processor 801. When the electronic device 800 is running, the processor 801 and the memory 803 communicate through the communication bus 804, and when the machine-readable instructions are called by the processor 801, the above-mentioned medical image segmentation method or the training method of the medical image segmentation model is executed.

For example, the processor 801 in the embodiment of the present application reads the computer program from the memory 803 through the communication bus 804 and executes the computer program. As an implementation manner, the following method can be implemented: Step S101: Obtain the medical image to be processed. Step S102: Input the medical image to be processed into the pre-trained medical image segmentation model, and obtain the segmentation result output by the medical image segmentation model. As another implementation manner, the following method can be implemented: Step S501: Obtain medical segmentation image data and the medical image segmentation model to be trained. Step S502: Load corresponding pre-training parameters for the natural image segmentation module. Step S503: Use the medical segmentation image data to update the parameters of the segmentation adapter and the image decoder in the medical image segmentation model to obtain a trained medical image segmentation model.

The processor 801 includes one or more integrated circuit chips, which may be integrated circuit chips, and have signal processing capabilities. The above-mentioned processor 801 can be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU for short), a micro control unit (Micro Controller Unit, MCU for short), a network processor (Network Processor, NP for short), or other conventional processors. ; It can also be a dedicated processor, including a neural network processor (Neural-network Processing Unit, referred to as NPU), a graphics processor (Graphics Processing Unit, referred to as GPU), a digital signal processor (Digital Signal Processor, referred to as DSP), a dedicated Integrated circuits (Application Specific Integrated Circuits, ASIC for short), Field Programmable Gate Array (FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components. Moreover, when there are multiple processors 801, some of them may be general-purpose processors and another part may be special-purpose processors.

The memory 803 includes one or more, which may be, but is not limited to, random access memory (RandomAccess Memory, RAM for short), read-only memory (Read Only Memory, ROM for short), programmable read-only memory (Programmable Read-Only) Memory (PROM for short), Erasable Programmable Read-Only Memory (EPROM for short), Electrically Erasable Programmable Read-Only Memory (EEPROM for short), etc.

It can be understood that the structure shown in FIG. 8 is only illustrative, and the electronic device 800 may also include more or fewer components than shown in FIG. 8 , or have a different configuration than that shown in FIG. 8 . Each component shown in Figure 8 can be implemented in hardware, software, or a combination thereof. In this embodiment of the present application, the electronic device 800 may be, but is not limited to, a desktop computer, a laptop computer, a smartphone, a smart wearable device, a vehicle-mounted device, or other physical device, or it may be a virtual device such as a virtual machine. In addition, the electronic device 800 is not necessarily a single device, but can also be a combination of multiple devices, such as a server cluster, etc.

Embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium stores computer program instructions. When the computer program instructions are run by a computer, they cause the computer to perform the medical treatment described in the foregoing method embodiments. Image segmentation method or training method of medical image segmentation model.

In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the coupling or direct coupling or communication connection between each other shown or discussed may be through some communication interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical or other forms.

In addition, units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Furthermore, each functional module in each embodiment of the present application can be integrated together to form an independent part, each module can exist alone, or two or more modules can be integrated to form an independent part.

It should be noted that if functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes.

In this document, relational terms such as first, second, etc. are used merely to distinguish one entity or operation from another entity or operation and do not necessarily require or imply the existence of any such entity or operation between these entities or operations. Actual relationship or sequence.

The above descriptions are only examples of the present application and are not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included in the protection scope of this application.

Claims (10)

1. A medical image segmentation method, comprising:

acquiring a medical image to be processed;

inputting the medical image to be processed into a medical image segmentation model trained in advance to obtain a segmentation result output by the medical image segmentation model;

the medical image segmentation model comprises a natural image segmentation module, wherein the natural image segmentation module comprises an image encoder, an image decoder and a segmentation adapter, the segmentation adapter is embedded into the image encoder, and training of the image encoder is achieved by updating parameters of the segmentation adapter.

2. The medical image segmentation method according to claim 1, wherein the inputting the medical image to be processed into a pre-trained medical image segmentation model to obtain the segmentation result output by the medical image segmentation model comprises:

Inputting the medical image to be processed into the image encoder to obtain an encoding result output by the image encoder;

and inputting the coding result into the image decoder to obtain the segmentation result output by the image decoder.

3. The medical image segmentation method as set forth in claim 1, further comprising:

training the medical image segmentation model by using the following steps:

acquiring medical segmentation image data and a medical image segmentation model to be trained;

loading corresponding pre-training parameters for the natural image segmentation module;

and updating parameters of the segmentation adapter and the image decoder in the medical image segmentation model by using the medical segmentation image data to obtain a trained medical image segmentation model.

4. The medical image segmentation method as set forth in claim 1, wherein the medical image segmentation model further comprises: a spatial multi-scale information feature extraction module;

inputting the medical image to be processed into a medical image segmentation model trained in advance to obtain a segmentation result output by the medical image segmentation model, wherein the method comprises the following steps:

Inputting the medical image to be processed into the spatial multi-scale information feature extraction module to obtain feature data output by the spatial multi-scale information feature extraction module;

inputting the medical image to be processed and the characteristic data into the image encoder to obtain an encoding result output by the image encoder;

and inputting the coding result into the image decoder to obtain the segmentation result output by the image decoder.

5. The medical image segmentation method as set forth in claim 4, further comprising:

training the medical image segmentation model by using the following steps:

acquiring medical segmentation image data and a medical image segmentation model to be trained;

loading corresponding pre-training parameters for the natural image segmentation module;

and updating parameters of the segmentation adapter, the image decoder and the spatial multi-scale information feature extraction module in the medical image segmentation model by using the medical segmentation image data to obtain a trained medical image segmentation model.

6. A method of training a medical image segmentation model, comprising:

Acquiring medical segmentation image data and a medical image segmentation model to be trained, wherein the medical image segmentation model comprises a natural image segmentation module, the natural image segmentation module comprises an image encoder, an image decoder and a segmentation adapter, and the segmentation adapter is embedded in the image encoder;

loading corresponding pre-training parameters for the natural image segmentation module;

and updating parameters of the segmentation adapter and the image decoder in the medical image segmentation model by using the medical segmentation image data to obtain a trained medical image segmentation model.

7. A medical image segmentation apparatus, comprising:

the first acquisition module is used for acquiring a medical image to be processed;

the input module is used for inputting the medical image to be processed into a medical image segmentation model trained in advance to obtain a segmentation result output by the medical image segmentation model;

the medical image segmentation model comprises a natural image segmentation module, wherein the natural image segmentation module comprises an image encoder, an image decoder and a segmentation adapter, the segmentation adapter is embedded into the image encoder, and training of the image encoder is achieved by updating parameters of the segmentation adapter.

8. A training device for a medical image segmentation model, comprising:

the medical image segmentation module comprises a natural image segmentation module, wherein the natural image segmentation module comprises an image encoder, an image decoder and a segmentation adapter, and the segmentation adapter is embedded in the image encoder;

the loading module is used for loading corresponding pre-training parameters aiming at the natural image segmentation module;

and the updating module is used for updating parameters of the segmentation adapter and the image decoder in the medical image segmentation model by utilizing the medical segmentation image data to obtain a trained medical image segmentation model.

9. An electronic device, comprising: a processor, a memory, and a bus;

the processor and the memory complete communication with each other through the bus;

the memory stores computer program instructions executable by the processor, the processor invoking the computer program instructions to be able to perform the medical image segmentation method according to any of claims 1-5 or the training method of the medical image segmentation model according to claim 6.

10. A computer readable storage medium storing computer program instructions which, when executed by a computer, cause the computer to perform the medical image segmentation method according to any one of claims 1-5 or the training method of the medical image segmentation model according to claim 6.