CT image multi-region segmentation method, device, electronic equipment and storage medium

Abstract

This application discloses a method, apparatus, electronic device, and storage medium for multi-region segmentation of CT images, belonging to the field of CT image processing technology. The multi-region segmentation method for CT images includes: acquiring a CT image, which is a 3D image comprising multiple regions; inputting the CT image into a trained segmentation model to obtain coarse segmentation labels for the multiple regions; performing a dot product between the coarse segmentation labels and the CT image to obtain an image to be classified containing only foreground information; and inputting the image to be classified into a multi-label classification network to obtain a multi-label classification result. This method is simple, efficient, and can perform relatively accurate segmentation and classification, which is helpful for subsequent CT image interpretation.

Description

Technical Field

[0001] The present application belongs to the technical field of CT image processing, and specifically relates to a method, device, electronic device and storage medium for segmenting multiple regions of a CT image. Background Art

[0002] CT (computed tomography) is an imaging technique that uses X-rays and computer algorithms to create a cross-sectional image.

[0003] However, due to the complexity and diversity of CT images, how to accurately segment and classify images has become a challenging problem. Summary of the invention

[0004] The purpose of the present application is to provide a method, device, electronic device and storage medium for segmenting multiple regions of a CT image to perform segmentation and classification of three-dimensional CT images.

[0005] According to a first aspect of an embodiment of the present application, a method for segmenting multiple regions of a CT image is provided, and the method may include:

[0006] Acquire a CT image, where the CT image is a 3D image including multiple regions;

[0007] Input the CT image into the trained segmentation model to obtain multi-region coarse segmentation labels;

[0008] Perform dot product between the multi-region rough segmentation labels and the CT image to obtain an image to be classified that only contains foreground information;

[0009] Input the image to be classified into the multi-label classification network to obtain the multi-label classification result.

[0010] In some optional embodiments of the present application, inputting the CT image into a trained segmentation model includes:

[0011] The CT image is normalized to obtain an input image;

[0012] Feed the input image into the trained segmentation model.

[0013] In some optional embodiments of the present application, the CT image is normalized to obtain an input image, including:

[0014] Resampling the CT image with a preset spatial resolution to obtain a resampled image;

[0015] The CT value of the resampled image is normalized to obtain the input image.

[0016] In some optional embodiments of the present application, inputting the input image into the trained segmentation model includes:

[0017] Perform the first convolution on the input image to obtain the first layer feature map;

[0018] The first layer feature map is subjected to pooling processing and second convolution processing in sequence to obtain the second layer feature map;

[0019] The second layer feature map is subjected to pooling processing and the third convolution processing in sequence to obtain the third layer feature map;

[0020] The third layer feature map is sequentially subjected to pooling processing and the fourth convolution processing to obtain the first feature map;

[0021] Desample the first feature map and concatenate it with the third-layer feature map to obtain a first concatenated map;

[0022] Perform a first convolution fusion process on the first spliced ​​image to obtain a second feature map;

[0023] Performing a first upsampling process on the second feature map and splicing the second feature map with the second layer feature map to obtain a second spliced ​​map;

[0024] Perform a second convolution fusion process on the second spliced ​​image to obtain a third feature map;

[0025] Perform a second upsampling process on the third feature map and concatenate it with the first layer feature map to obtain a third concatenated map;

[0026] Perform a fifth convolution process on the third spliced ​​image to obtain a fourth feature map;

[0027] The fourth feature map is inversely transformed to obtain multi-region coarse segmentation labels.

[0028] In some optional embodiments of the present application, the multi-label classification network includes multiple densely connected blocks.

[0029] In some optional embodiments of the present application, the loss function of the segmentation model is a Dice loss function under a deep supervision loss mode.

[0030] In some optional embodiments of the present application, the loss function of the multi-label classification network is a binary cross entropy loss function.

[0031] According to a second aspect of an embodiment of the present application, a device for segmenting multiple regions of a CT image is provided, and the device may include:

[0032] An acquisition module, used for acquiring a CT image, where the CT image is a 3D image including multiple regions;

[0033] The segmentation module is used to input the CT image into the trained segmentation model to obtain multi-region coarse segmentation labels;

[0034] The dot product module is used to perform dot product between the multi-region coarse segmentation labels and the CT image to obtain an image to be classified containing only foreground information;

[0035] The classification module is used to input the image to be classified into the multi-label classification network to obtain the multi-label classification result.

[0036] According to a third aspect of an embodiment of the present application, an electronic device is provided, which may include:

[0037] processor;

[0038] a memory for storing processor-executable instructions;

[0039] The processor is configured to execute instructions to implement the method for segmenting multiple regions of a CT image as shown in any one of the embodiments of the first aspect.

[0040] According to a fourth aspect of an embodiment of the present application, a storage medium is provided. When instructions in the storage medium are executed by a processor of an information processing device or a server, the information processing device or the server implements a method for segmenting multiple regions of a CT image as shown in any one of the embodiments of the first aspect.

[0041] The above technical solution of the present application has the following beneficial technical effects:

[0042] The method of the embodiment of the present application first uses the segmentation model to segment the three-dimensional CT image to obtain multi-region rough segmentation labels, then performs dot product between the multi-region rough segmentation labels and the CT image to obtain an image to be classified containing only foreground information, and finally inputs the image to be classified into a multi-label classification network to obtain a multi-label classification result. The method is simple, efficient, and can perform segmentation and classification more accurately, which is helpful for subsequent CT image judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of a method for segmenting multiple regions of a CT image in an exemplary embodiment of the present application;

[0044] Figure 2 is a schematic diagram of the structure of a device for segmenting multiple regions of a CT image in an exemplary embodiment of the present application;

[0045] Figure 3 is a schematic diagram of the structure of an electronic device in an exemplary embodiment of the present application;

[0046] Figure 4 It is a schematic diagram of the hardware structure of an electronic device in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with specific implementations and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present application. In addition, in the following description, the description of known structures and technologies is omitted to avoid unnecessary confusion of the concepts of the present application.

[0048] The accompanying drawings show schematic diagrams of layer structures according to embodiments of the present application. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clarity. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0049] Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0050] In the description of the present application, it should be noted that the terms "first", "second" and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0051] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0052] In conjunction with the accompanying drawings, the CT image multi-region segmentation method, device, electronic device and storage medium provided in the embodiments of the present application are described in detail through specific embodiments and their application scenarios.

[0053] like Figure 1 As shown, in a first aspect of an embodiment of the present application, a method for segmenting multiple regions of a CT image is provided, and the method may include:

[0054] S110: Acquire a CT image, where the CT image is a 3D image including multiple regions;

[0055] S120: Input the CT image into the trained segmentation model to obtain multi-region rough segmentation labels;

[0056] S130: performing a dot product between the multi-region coarse segmentation labels and the CT image to obtain an image to be classified containing only foreground information;

[0057] S140: Input the image to be classified into a multi-label classification network to obtain a multi-label classification result.

[0058] The method of the embodiment of the present application first uses the segmentation model to segment the three-dimensional CT image to obtain multi-region rough segmentation labels, then performs dot product between the multi-region rough segmentation labels and the CT image to obtain an image to be classified containing only foreground information, and finally inputs the image to be classified into a multi-label classification network to obtain a multi-label classification result. The method is simple, efficient, and can perform segmentation and classification more accurately, which is helpful for subsequent CT image judgment.

[0059] This method solves the needs of multi-region precise segmentation and multi-label classification of CT images. It can assist clinicians in quickly and accurately making disease judgments in emergency medical imaging scenarios, saving a lot of manpower and material costs for subsequent rapid disposal and treatment. It can also assist auditors in industrial inspection scenarios to quickly achieve multi-region segmentation and multi-category judgment of different regions for industrial components with product defects or flaws, thereby improving the efficiency of industrial inspection.

[0060] In some embodiments, inputting the CT image into a trained segmentation model includes:

[0061] The CT image is normalized to obtain an input image;

[0062] Feed the input image into the trained segmentation model.

[0063] For example, the original 3D CT image may be first resampled to a spatial resolution of 1*1*1, and then normalized by a Z-score CT value to obtain an input image.

[0064] The segmentation model of this embodiment can be a segmentation network with a U-shaped architecture, so that the input image can be roughly segmented to obtain a rough multi-region segmentation result.

[0065] In some embodiments, the CT image is normalized to obtain an input image, including:

[0066] Resampling the CT image with a preset spatial resolution to obtain a resampled image;

[0067] The CT value of the resampled image is normalized to obtain the input image.

[0068] In some embodiments, inputting the input image into the trained segmentation model includes:

[0069] Perform the first convolution on the input image to obtain the first layer feature map;

[0070] The first layer feature map is subjected to pooling processing and second convolution processing in sequence to obtain the second layer feature map;

[0071] The second layer feature map is subjected to pooling processing and the third convolution processing in sequence to obtain the third layer feature map;

[0072] The third layer feature map is sequentially subjected to pooling processing and the fourth convolution processing to obtain the first feature map;

[0073] Desample the first feature map and concatenate it with the third-layer feature map to obtain a first concatenated map;

[0074] Perform a first convolution fusion process on the first spliced ​​image to obtain a second feature map;

[0075] Performing a first upsampling process on the second feature map and splicing the second feature map with the second layer feature map to obtain a second spliced ​​map;

[0076] Perform a second convolution fusion process on the second spliced ​​image to obtain a third feature map;

[0077] Perform a second upsampling process on the third feature map and concatenate it with the first layer feature map to obtain a third concatenated map;

[0078] Perform a fifth convolution process on the third spliced ​​image to obtain a fourth feature map;

[0079] The fourth feature map is inversely transformed to obtain multi-region coarse segmentation labels.

[0080] Exemplarily, the input image is first convolved twice to obtain the first layer feature map. The first layer feature map is pooled to reduce the size of the feature map, but at the same time increase the number of channels of the feature map. After the pooled feature map is convolved twice, the second layer feature map is obtained. After the second layer feature map is pooled, the size of the feature map is also reduced and the number of channels is increased, and two convolutions are performed to obtain the third layer feature map. After the third layer feature map is pooled and convolved twice, the size of the feature map is further reduced and the number of channels is increased, and finally the first feature map is obtained. The first feature map is the final hidden feature in the segmentation network encoder in the multi-task network framework. In order to better complete the multi-region segmentation, the first feature map is desampled and spliced ​​with the third layer feature map to strengthen the network's analysis of the previous and next features. Since the decoder of the segmentation network loses some information during the upsampling process, it is necessary to introduce the features of the encoder of the same layer to supplement it, which helps to avoid the problem of information loss in multiple convolutions. The spliced ​​features are convoluted and fused to obtain the second feature map. The second feature map is upsampled to increase the size of the feature map and spliced ​​with the second layer feature map of the same layer to enhance the information expression capability. After the concatenated features are convolved twice, the third feature map is obtained. After upsampling, the third feature map is concatenated with the first-layer feature map of the same layer, and finally the fourth feature map is obtained after two convolutions. The size of the fourth feature map is exactly the same as the input image size. The number of channels is converted to the number of channels of the input image size through the inverse transformation of one-hot encoding, and the multi-region coarse segmentation labels of the segmentation network can be obtained.

[0081] In some embodiments, the multi-label classification network includes a plurality of densely connected blocks.

[0082] In this embodiment, a new input image containing only foreground information is obtained by performing a dot product of the multi-region rough segmentation labels and the original input CT image. The image is sent to a multi-label classification network to finally output the multi-label classification results of different regions. The multi-label classification network is composed of different densely connected blocks, and each densely connected block adopts the idea of ​​sequential transmission of feature maps to strengthen the network's continuous learning of low-order and high-order latent space features. Specifically, the image is in the first densely connected block, and different feature maps are obtained through convolution operations. The first group of feature maps will fuse the equal-dimensional features in the segmentation network encoder, and the fused features will participate in the convolution operations of subsequent different groups of feature maps, so that the network can continue to learn the previous feature space. Finally, the feature map matrix of the block is output at the end of the densely connected block. After convolution and pooling operations, the size of the feature map is reduced and the number of feature map channels is increased. The newly generated feature map enters the next densely connected block, and the new feature map is obtained by the same convolution operation and feature splicing method, and finally convolution and pooling are performed to reduce the feature dimension. After passing through multiple densely connected blocks, the feature map passes through the fully connected layer and outputs the same number of predicted probability points in the same region, representing the normal and abnormal states of different segmented regions, and finally completes the multi-label classification task. In the multi-label classification network, the dense skip connection of multiple groups of feature maps in the previous stage will help the network learn the previous and current features, thereby avoiding gradient disappearance and effectively using features for subsequent multi-label prediction and classification tasks.

[0083] In some embodiments, the loss function of the segmentation model is a Dice loss function in a deep supervision loss mode.

[0084] In some embodiments, the loss function of the multi-label classification network is a binary cross entropy loss function.

[0085] The above embodiment is a multi-region segmentation and multi-label classification network based on a multi-task idea. It will effectively combine the feature layers in the segmentation task and the classification task. By comparing and combining the feature layers between different tasks, the network can learn the correlation between the classification task and the segmentation task, so that the performance is higher than that of a single subtask. In order to better converge the network, the network in the present invention adopts a weighted loss function. The loss function of the segmentation network adopts the Dice loss function under the deep supervision loss mode, and the loss function of the classification network adopts the binary cross entropy loss function. Specifically, the loss function Ls of the segmentation model is defined. Ls obtains the same space as the standard segmentation result by upsampling the first feature map, the second feature map, the third feature map and the fourth feature map, and then calculates the losses of the above feature maps respectively and defines them as LsA, LsB, LsC and LsD. Then the total loss function of the segmentation network is Ls=LsA+LsB+LsC+LsD. Each sub-loss function is defined as: L = 1-2*sum(yi*pi) / (sum(yi)+sum(pi)), yi represents the actual label of the i-th pixel, pi represents the predicted label of the i-th pixel, and sum is the sum calculation. The loss function of the classification network can perform a separate binary classification for each label and add the loss functions of all labels. Assuming there are N labels, the predicted value of the i-th label is pi, and the actual value is yi, then the binary cross entropy loss function of the i-th label is: Li = -[yi*log(pi)+(1-yi)*log(1-pi)], adding the loss functions of all labels, the total loss function of the multi-label classification network is Lc = sum(Li), where sum represents the sum of the loss functions of all labels. The final total loss function of the multi-task network is L = Ls+Lc. Therefore, the multi-task network algorithm of the above embodiment can effectively segment multi-region targets and realize multi-region multi-label classification.

[0086] By combining the encoder feature information of the segmentation network with the latent space features of the multi-label classification network, the progressive conduction of the feature map is used to help the network use the leading information, and the dual-task loss function is used to jointly constrain the convergence of the auxiliary network, thereby achieving accurate segmentation of multiple regions of CT images and accurate multi-label classification. The multi-task segmentation and classification network can realize the feature transfer and communication of the two subtasks, and better realize the model performance of different tasks by exchanging high-dimensional latent space features. At the same time, since the model can output segmentation and classification results at the same time, it greatly reduces the high parameter problem caused by the cascade of multiple models, which will effectively save the multi-region segmentation and multi-label classification reasoning time of CT images and improve the network judgment efficiency.

[0087] The above-mentioned embodiment method can be applied in multiple fields, such as multi-organ segmentation and multi-label classification of different organs in emergency scenarios in the medical field, and the use of CT images in industrial detection to achieve segmentation and multi-label classification analysis of different parts and components of industrial products. In addition to this, other fields can be fine-tuned and transformed according to actual conditions. For example: in materials science, CT images can be used to study the microstructure of materials, such as fibers, particles, etc. The segmentation technology of the above-mentioned embodiment is used to separate different material components, and then further morphological and physical analysis is performed; in geological exploration, CT images can be used to study the structure and composition of geological samples, and the segmentation technology of the above-mentioned embodiment is used to separate different rock components, and then further geological analysis is performed; in automotive engineering, CT images can be used to detect defects and damage of automotive parts, such as wheels, engine cylinders, etc., and the defective parts can be separated from the normal parts using the above-mentioned embodiment segmentation technology, and then further classification and analysis is performed.

[0088] Exemplary application scenarios in medical imaging:

[0089] The above embodiments can be applied to multi-organ segmentation and multi-label classification problems of different organs in emergency medicine scenarios, such as segmenting multiple organs in CT abdomen and simultaneously providing category prediction results of whether each organ has trauma.

[0090] Exemplary, applied to industrial detection application scenarios:

[0091] The above-mentioned embodiment method can be applied to CT images of industrial products to automatically segment different areas in different industrial products, and determine whether different areas have defects or damage based on the segmentation results, ultimately achieving a multi-label classification task of determining whether different sub-areas in industrial products are qualified.

[0092] It should be noted that the CT image multi-region segmentation method provided in the embodiment of the present application can be executed by a CT image multi-region segmentation device, or a control module in the CT image multi-region segmentation device for executing the CT image multi-region segmentation method. In the embodiment of the present application, the CT image multi-region segmentation method executed by the CT image multi-region segmentation device is taken as an example to illustrate the CT image multi-region segmentation device provided in the embodiment of the present application.

[0093] like Figure 2 As shown, in a second aspect of the embodiment of the present application, a device for segmenting multiple regions of a CT image is provided, and the device may include:

[0094] An acquisition module 210 is used to acquire a CT image, where the CT image is a 3D image including multiple regions;

[0095] The segmentation module 220 is used to input the CT image into the trained segmentation model to obtain multi-region rough segmentation labels;

[0096] A dot product module 230 is used to perform dot product between the multi-region coarse segmentation labels and the CT image to obtain an image to be classified containing only foreground information;

[0097] The classification module 240 is used to input the image to be classified into the multi-label classification network to obtain a multi-label classification result.

[0098] The CT image multi-region segmentation device in the embodiment of the present application can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), etc. The non-mobile electronic device can be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0099] The CT image multi-region segmentation device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0100] The CT image multi-region segmentation device provided in the embodiment of the present application can achieve Figure 1 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0101] Alternatively, if Figure 3 As shown, the embodiment of the present application also provides an electronic device 300, including a processor 301, a memory 302, and a program or instruction stored in the memory 302 and executable on the processor 301. When the program or instruction is executed by the processor 301, each process of the above-mentioned CT image multi-region segmentation method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0102] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0103] Figure 4 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of the present application.

[0104] The electronic device 400 includes but is not limited to components such as a radio frequency unit 401 , a network module 402 , an audio output unit 403 , an input unit 404 , a sensor 405 , a display unit 406 , a user input unit 407 , an interface unit 408 , a memory 409 , and a processor 410 .

[0105] Those skilled in the art will appreciate that the electronic device 400 may also include a power source (such as a battery) for supplying power to each component, and the power source may be logically connected to the processor 410 through a power management system, thereby implementing functions such as managing charging, discharging, and power consumption management through the power management system. Figure 4 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be described in detail here.

[0106] It should be understood that in the embodiment of the present application, the input unit 404 may include a graphics processor (GPU) 4041 and a microphone 4042, and the graphics processor 4041 processes the image data of a static picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 406 may include a display panel 4061, and the display panel 4061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 407 includes a touch panel 4071 and other input devices 4072. The touch panel 4071 is also called a touch screen. The touch panel 4071 may include two parts: a touch detection device and a touch controller. Other input devices 4072 may include, but are not limited to, a physical keyboard, a function key (such as a volume control button, a switch button, etc.), a trackball, a mouse, and a joystick, which will not be repeated here. The memory 409 may be used to store software programs and various data, including but not limited to applications and operating systems. The processor 410 may integrate an application processor and a modem processor, wherein the application processor mainly processes an operating system, a user interface, and applications, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 410.

[0107] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned CT image multi-region segmentation method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0108] The processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0109] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned CT image multi-region segmentation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0110] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0111] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0112] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0113] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for segmenting multiple regions of a CT image, characterized in that: include: Acquire a CT image, wherein the CT image is a 3D image including multiple regions; Inputting the CT image into a trained segmentation model to obtain multi-region rough segmentation labels; Performing a dot product between the multi-region coarse segmentation labels and the CT image to obtain an image to be classified containing only foreground information; Inputting the image to be classified into a multi-label classification network to obtain a multi-label classification result, wherein the multi-label classification network is composed of different densely connected blocks, and a feature map sequential conduction method is adopted in each densely connected block to enhance the continuous learning of low-order and high-order latent space features by the multi-label classification network; The step of inputting the CT image into the trained segmentation model comprises: normalizing the CT image to obtain an input image; and inputting the input image into the trained segmentation model; The step of normalizing the CT image to obtain an input image includes: firstly resampling the CT image to a 1*1*1 spatial resolution, and then normalizing the CT value of the CT image to obtain the input image by Z-score; The inputting of the input image into the trained segmentation model comprises: performing a first convolution process on the input image to obtain a first layer feature map; performing a pooling process and a second convolution process on the first layer feature map in sequence to obtain a second layer feature map, wherein the input image is firstly subjected to two convolution processes to obtain a first layer feature map, and the first layer feature map is subjected to a pooling operation to reduce the size of the feature map and increase the number of channels of the feature map; performing a pooling process and a third convolution process on the second layer feature map in sequence, and after the second layer feature map is subjected to a pooling operation, the size of the feature map is reduced and the number of channels is increased to obtain a third layer feature map; performing a pooling process and a fourth convolution process on the third layer feature map in sequence, and after the third layer feature map is subjected to a pooling operation and two convolution processes, the size of the feature map is further reduced and the number of channels is increased The number of channels is increased to obtain a first feature map; the first feature map is desampled and concatenated with the third-layer feature map to obtain a first concatenated map; the first convolution fusion process is performed on the first concatenated map to obtain a second feature map; the second feature map is performed on a first upsampling process and concatenated with the second-layer feature map to obtain a second concatenated map; the second convolution fusion process is performed on the second concatenated map to obtain a third feature map; the third feature map is performed on a second upsampling process and concatenated with the first-layer feature map to obtain a third concatenated map; the third concatenated map is performed on a fifth convolution process to obtain a fourth feature map, and the size of the fourth feature map is the same as the size of the input image; the fourth feature map is performed on an inverse transform of one-hot encoding to obtain a multi-region coarse segmentation label.

2. The method for segmenting multiple regions of a CT image according to claim 1, characterized in that: The loss function of the segmentation model is the Dice loss function under the deep supervision loss mode.

3. The method for segmenting multiple regions of a CT image according to claim 1, characterized in that: The loss function of the multi-label classification network is a binary cross entropy loss function.

4. A CT image multi-region segmentation device, characterized in that: include: An acquisition module, used for acquiring a CT image, wherein the CT image is a 3D image including multiple regions; A segmentation module, used for inputting the CT image into a trained segmentation model to obtain multi-region rough segmentation labels; A dot product module, used for performing a dot product between the multi-region coarse segmentation labels and the CT image to obtain an image to be classified containing only foreground information; A classification module, used for inputting the image to be classified into a multi-label classification network to obtain a multi-label classification result, wherein the multi-label classification network is composed of different densely connected blocks, and each densely connected block adopts a feature map sequential conduction method to enhance the continuous learning of low-order and high-order latent space features by the multi-label classification network; The segmentation module is also used to resample the CT image to a 1*1*1 spatial resolution and then perform Z-score CT value standardization to obtain an input image; The segmentation module is also used to perform a first convolution process on the input image to obtain a first-layer feature map; perform a pooling process and a second convolution process on the first-layer feature map in sequence to obtain a second-layer feature map, where the input image is firstly subjected to two convolutions to obtain a first-layer feature map, and the first-layer feature map is subjected to a pooling operation to reduce the size of the feature map and increase the number of channels of the feature map; the second-layer feature map is subjected to a pooling process and a third convolution process in sequence, and after the second-layer feature map is subjected to a pooling operation, the size of the feature map is reduced and the number of channels is increased to obtain a third-layer feature map; the third-layer feature map is subjected to a pooling process and a fourth convolution process in sequence, and after the third-layer feature map is subjected to a pooling operation and two convolutions, the size of the feature map is further reduced and the number of channels is increased to obtain a first-layer feature map. feature map; desampling the first feature map and splicing it with the third layer feature map to obtain a first spliced ​​map; performing a first convolution fusion process on the first spliced ​​map to obtain a second feature map; performing a first upsampling process on the second feature map and splicing it with the second layer feature map to obtain a second spliced ​​map; performing a second convolution fusion process on the second spliced ​​map to obtain a third feature map; performing a second upsampling process on the third feature map and splicing it with the first layer feature map to obtain a third spliced ​​map; performing a fifth convolution process on the third spliced ​​map to obtain a fourth feature map, the size of the fourth feature map being the same as the size of the input image; performing an inverse transform process of one-hot encoding on the fourth feature map to obtain a multi-region coarse segmentation label.

5. An electronic device, characterized in that: include: A processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method for segmenting multiple regions of a CT image as described in any one of claims 1 to 3.

6. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the method for segmenting multiple regions of a CT image as described in any one of claims 1 to 3 are implemented.

Images