Liver Eight-Segment Segmentation Model Based on Deep Learning, Its Training Method and Segmentation Method

Abstract

The present invention provides a liver eight-segment segmentation model based on deep learning, and its training method and segmentation method. Among them, the liver eight-segment segmentation model includes: an input layer for receiving input data to obtain a target sequence; a normalization layer for normalizing the target sequence; a linear layer for changing the dimension of the intermediate processing data of the current model; a convolutional layer for extracting features from the intermediate processing data of the current model; an activation layer for performing a non-linear transformation on the output of the convolutional layer using an activation function; and an output layer for calculating and processing the model using a preset loss function and outputting the model segmentation result. The liver eight-segment segmentation model based on deep learning, and its training method and segmentation method of the present invention can better adapt to different data characteristics and changes by using deep learning algorithms, realize automatic segmentation, improve the accuracy of segmentation, can quickly process and analyze large-scale medical image data, and improve the efficiency of liver eight-segment segmentation.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and particularly relates to a liver eight-segment segmentation model based on deep learning, a training method thereof, and a segmentation method thereof. Background Art

[0002] In medical imaging, liver segmentation is a key clinical task that can help doctors accurately diagnose liver diseases and formulate treatment plans. Traditionally, the main method used by doctors is the Couinaud segmentation method. However, in practical applications, it has the following disadvantages:

[0003] (1) The traditional Couinaud segmentation method is based on standard anatomical structures and is insufficient in dealing with the variability of liver anatomy among individuals;

[0004] (2) The traditional segmentation method relies on doctors' experience and manual operations, which has certain subjectivity and operational complexity. Differences in segmentation results may occur among different doctors, affecting the consistency of diagnosis and treatment. Summary of the Invention

[0005] In view of the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide a liver eight-segment segmentation model based on deep learning, a training method thereof, and a segmentation method thereof to solve the above problems.

[0006] In the first aspect, the present invention provides a liver eight-segment segmentation model based on deep learning, including:

[0007] An input layer for receiving input data to obtain a target sequence;

[0008] A normalization layer for normalizing the target sequence;

[0009] A linear layer for changing the dimension of the intermediate processing data of the current model;

[0010] A convolutional layer for extracting features from the intermediate processing data of the current model;

[0011] An activation layer for performing non-linear transformation on the output of the convolutional layer using an activation function;

[0012] An output layer for calculating and processing the model using a preset loss function and outputting the model segmentation result.

[0013] In the first aspect, the present invention provides a training method for a liver eight-segment segmentation model based on deep learning, which is applied to the liver eight-segment segmentation model based on deep learning, and the method includes:

[0014] Obtaining input images to obtain a training set and a validation set, where the input images include liver images with data labels;

[0015] Input the training set into the liver eight-segment segmentation model for testing, and use the validation set for verification to perform iterative training. Among them, when the number of iterations reaches the target iteration value, stop the iteration, or stop the iteration when the loss function reaches the target loss value;

[0016] After the iteration ends, output the finally trained liver eight-segment segmentation model based on the optimal model.

[0017] In a possible implementation manner of this application, training the liver eight-segment segmentation model specifically includes:

[0018] Perform image processing on the input image to obtain a target sequence, where the image processing method at least includes block operation and summation operation;

[0019] During the block operation, divide the input image into blocks according to a preset format, and use serialization operation to convert each block image into a one-dimensional vector;

[0020] During the summation operation, perform a summation operation based on the position information of each block image and the one-dimensional vector sequence to obtain the target sequence.

[0021] In a possible implementation manner of this application, during the block operation, it specifically includes:

[0022] Obtain the picture specification of the input image, where the input image specification is H*W, where H is the height of the input image and W is the width of the input image;

[0023] Perform block operation on the input image to obtain block images, where the block images are square images, the block image specification is P*P, and N = H*W / (P*P), P is the side length of the block image, and N is the number value of the block images;

[0024] Convert the block images into a one-dimensional vector N*D, where D is the sequence length value after serialization.

[0025] In a possible implementation manner of this application, training the liver eight-segment segmentation model specifically further includes:

[0026] Perform normalization processing on the target sequence and then change the dimension to obtain the first processed data;

[0027] Perform convolution operation and activation function processing on the first processed data to obtain the second processed data;

[0028] Process the second processed data using a structured state space model to obtain the third processed data;

[0029] Process the first processed data using an activation function, multiply it with the third processed data, and change the dimension to obtain the fourth processed data;

[0030] Restore the dimension of the feature map based on the fourth processed data to obtain the fifth processed data;

[0031] Perform upsampling and convolution operations based on the fifth processed data to restore the feature map and the target mask.

[0032] In a possible implementation manner of the present application, the loss function used when training the liver eight-segment segmentation model is the Dice-CE Loss function, and the ratio of the training set to the validation set is 4:1.

[0033] In a third aspect, the present invention provides a training system for a liver eight-segment segmentation model based on deep learning, and the system includes:

[0034] An acquisition module, configured to acquire input images to obtain a training set and a validation set, where the input images include liver images with data labels;

[0035] A training module, configured to input the training set into the liver eight-segment segmentation model for testing, verify it using the validation set, and perform iterative training. Among them, when the number of iterations reaches the target iteration value, the iteration stops, or when the loss function reaches the target loss value, the iteration stops;

[0036] An output module, configured to output the finally trained liver eight-segment segmentation model based on the optimal model after the iteration ends.

[0037] In a fourth aspect, the present invention provides an electronic device, and the electronic device includes: a processor and a memory;

[0038] The memory is used to store a computer program;

[0039] The processor is configured to execute the computer program stored in the memory, so that the electronic device executes the above-mentioned training method for a liver eight-segment segmentation model based on deep learning.

[0040] In a fifth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by an electronic device, it implements the above-mentioned training method for a liver eight-segment segmentation model based on deep learning.

[0041] In a sixth aspect, the present invention provides a segmentation method for a liver eight-segment segmentation model based on deep learning, characterized in that it is applied to a liver eight-segment segmentation model trained by any one of the above-mentioned training methods for a liver eight-segment segmentation model based on deep learning. Among them, the segmentation method includes:

[0042] Obtain the image to be segmented;

[0043] Input the image to be segmented into the trained liver eight-segment segmentation model for processing to obtain a segmentation result;

[0044] Output the target segmented image based on the segmentation result.

[0045] As described above, the liver eight-segment segmentation model based on deep learning and its training method and segmentation method of the present invention have the following beneficial effects:

[0046] (1) Improved accuracy: In the liver segmentation of the present invention, a large number of medical image data can be learned, and more accurate features can be extracted therefrom, so as to achieve more accurate liver segmentation. Compared with traditional rule-based or feature engineering methods, it can better adapt to different data features and changes, and improve the accuracy of segmentation;

[0047] (2) Improved automation: The present invention can realize the automatic analysis and processing of medical images, thereby reducing the work burden of doctors. Traditional liver segmentation requires doctors to manually draw regions or perform complex calculations, while the liver eight-segment segmentation model based on deep learning of the present invention can realize an automatic segmentation process, saving doctors' time and reducing the influence of human factors on the results;

[0048] (3) Enhanced applicability: The deep learning-based algorithm adopted by the present invention has strong generalization ability and can adapt to different types and qualities of medical image data. Therefore, it can be applied to various clinical scenarios. Whether it is different types of medical image data such as MRI, CT or ultrasound, the deep learning algorithm can achieve good liver segmentation results, thereby improving the applicability of the algorithm;

[0049] (4) Improved speed: Liver segmentation is a complex and time-consuming task, and the present invention can utilize the advantages of parallel computing to achieve rapid processing and analysis of large-scale medical image data, thereby improving the speed and efficiency of segmentation. Description of the Drawings

[0050] Figure 1 Shown is a schematic diagram of the scenario in an embodiment of the electronic device of the present invention;

[0051] Figure 2 Shown is a schematic diagram of the steps in an embodiment of the training method of the liver eight-segment segmentation model based on deep learning of the present invention;

[0052] Figure 3 Shown is a schematic diagram of the structure of the liver eight-segment segmentation model based on deep learning of the present invention;

[0053] Figure 4Schematic diagram of steps in an embodiment of the training method of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0054] Figure 5 Schematic diagram of steps in an embodiment of the training method of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0055] Figure 6 Schematic diagram of the structure in an embodiment of the training system of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0056] Figure 7 Schematic diagram of the structure in an embodiment of the electronic device according to the present invention;

[0057] Figure 8 Schematic diagram of steps in an embodiment of the segmentation method of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0058] Figure 9 Schematic diagram of the input image in an embodiment of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0059] Figure 10A Cross-sectional view of the segmentation result in an embodiment of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0060] Figure 10B Front view of the segmentation result in an embodiment of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0061] Figure 10C Side view of the segmentation result in an embodiment of the liver eight-segment segmentation model based on deep learning according to the present invention;

[0062] Figure 11 Cross-sectional view of the input image in an embodiment of the liver eight-segment segmentation model based on deep learning according to the present invention. Detailed implementation manners

[0063] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0064] It should be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of the present invention. Therefore, only the components related to the present invention are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and proportion of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0065] The following embodiments of the present invention provide a training method for a liver eight-segment segmentation model based on deep learning, which can be applied to an electronic device as shown in Figure 1 The electronic device described in the present invention may include a mobile phone 11 with a wireless charging function, a tablet computer 12, a laptop computer 13, a wearable device, a vehicle-mounted device, an augmented reality (AR) / virtual reality (VR) device, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc. The specific type of the electronic device is not limited in the embodiments of the present invention.

[0066] For example, the electronic device may be a station (STAION, ST) in a WLAN with a wireless charging function, a cellular phone with a wireless charging function, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA) device, a handheld device with a wireless charging function, a computing device or other processing device, a computer, a laptop computer, a handheld communication device, a handheld computing device, and / or other devices for communicating on a wireless system, as well as a next-generation communication system. For example, a mobile terminal in a 5G network, a mobile terminal in a future evolved public land mobile network (PLMN), or a mobile terminal in a future evolved non-terrestrial network (NTN), etc.

[0067] For example, the electronic device can communicate with a network and other devices via wireless communication. The above wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobilecommunication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), BT, GNSS, WLAN, NFC, FM, and / or IR technology, etc. The GNSS can include Global Positioning System (GPS), Global Navigation Satellite System (GLONASS), BeiDou navigation Satellite System (BDS), Quasi-Zenith Satellite System (QZSS), and / or Satellite Based Augmentation Systems (SBAS).

[0068] The technical solutions in the embodiments of the present invention will be described in detail below with reference to the accompanying drawings in the embodiments of the present invention.

[0069] Furthermore, Couinaud's liver segmental division method is an important technique based on the liver's anatomical structure. It is based on the distribution of the Glisson system within the liver. Here, Couinaud is the surname of the French anatomist Pierre Couinaud, which is used to describe the anatomical structure of the liver. Specifically, the Glisson system is a duct system composed of the branches of the hepatic portal vein, hepatic artery, and bile duct within the liver, presenting a dendritic distribution inside the liver. According to Couinaud's method, the liver is divided into "2" half livers, "4" zones, and "8" liver segments. This division is based on specific anatomical structures formed by the hepatic portal fissure, the middle hepatic fissure, and the left and right interlobar fissures, etc., dividing the liver into different functional regions. Specifically, the hepatic portal fissure divides the liver into "4" sectors, and each sector is further divided into upper and lower segments by the horizontal section of the left and right branches of the portal vein. The caudate lobe is numbered as segment I and exists as an autonomous segment. The middle hepatic fissure divides the liver into the left half liver and the right half liver, and the left and right interlobar fissures further subdivide the left and right half livers into the left external sector, the left internal sector, the right anterior sector, and the right posterior sector.

[0070] This liver segmental division method based on the Glisson system has important guiding significance in liver surgery. Since each liver segment receives a branch from the Glisson system, especially a branch of the portal vein system, clinicians can perform resection with the smallest possible scope according to the lesion situation to retain normal liver tissue as much as possible. In medical imaging, this method can also accurately locate liver segments by identifying hepatic veins, portal veins, and other structures. Therefore, Couinaud's liver segmental division method not only provides an important theoretical basis for liver anatomy but also provides practical guidance for clinical diagnosis and surgical operations.

[0071] However, with the further study of the anatomical structure of the intrahepatic ducts, it has been found that there are significant anatomical variations in the internal vascular structure of the liver. These anatomical variations may cause the traditional Couinaud segmentation method to be unable to accurately reflect the liver anatomical structure of some individuals. Therefore, it has become crucial to study and improve the background technology of the liver segmentation method. In response to this situation, the present invention began to explore using high-resolution medical imaging and deep learning techniques to improve the liver segmentation method. By using a large amount of medical imaging data, the deep learning model learns from the images and accurately segments each functional region of the liver. This deep learning-based liver segmentation method can better cope with the variations in the liver anatomical structure and provide more accurate anatomical guidance for clinical imaging diagnosis and liver surgery. Moreover, the development of this technology will bring important progress to the diagnosis and treatment of liver diseases, providing better medical services for patients, thus proposing the deep learning-based liver eight-segment segmentation model of the present invention and its training method and segmentation method.

[0072] Specifically, please refer to Figure 2 , in an embodiment of the invention, the training method of the liver eight-segment segmentation model based on deep learning of the present invention includes the following steps:

[0073] Step S202, obtaining input images to obtain a training set and a validation set, where the input images include liver images with data labels;

[0074] Step S204, inputting the training set into the liver eight-segment segmentation model for testing, and using the validation set for verification to perform iterative training, where when the number of iterations reaches the target iteration value, the iteration stops, or when the loss function reaches the target loss value, the iteration stops;

[0075] Step S206, after the iteration ends, outputting the finally trained liver eight-segment segmentation model based on the optimal model.

[0076] It should be noted that in this embodiment, the specifically described training method of the liver eight-segment segmentation model is applied to the liver eight-segment segmentation model based on deep learning, where, as Figure 3 shown, in an embodiment of the invention, the liver eight-segment segmentation model 30 based on deep learning specifically includes: an input layer 31 for receiving input data to obtain a target sequence; a normalization layer 32 for normalizing the target sequence; a linear layer 33 for changing the dimension of the intermediate processing data of the current model; a convolutional layer 34 for extracting features from the intermediate processing data of the current model; an activation layer 35 for performing a non-linear transformation on the output of the convolutional layer using an activation function; and an output layer 36 for calculating and processing the model using a preset loss function and outputting the model segmentation result.

[0077] Further, after constructing the liver eight-segment segmentation model based on the deep learning, it needs to be trained. First, an input image is obtained, where the input image is a liver image with data labels, and the input image is classified to obtain a training set and a validation set, which are specifically distinguished according to the ratio of "4:1". After obtaining the training set and the validation set, the liver eight-segment segmentation model is tested based on the training set, and the validation set is used for verification to perform multiple iterative trainings. Among them, stopping the iteration can be divided into two cases. One is to stop the iteration when the number of iterations reaches the target iteration value. For example, stop the iteration when it reaches "400" times. The other is to stop the iteration when the loss function reaches the target loss value. For example, stop the iteration when it reaches below the target loss value of "0.05". Correspondingly, the loss function used when training the liver eight-segment segmentation model is the Dice-CE Loss function. Among them, Dice Loss is a loss function used for image segmentation tasks and has been widely used in fields such as medical image segmentation. It can solve the problem that the cross-entropy loss function performs poorly on unbalanced data sets, while CE Loss is also a commonly used function in classification tasks and is a measure of the cross-entropy between the predicted probability distribution and the true label probability distribution. Therefore, in this embodiment, the two of them are combined as the loss function of the liver eight-segment segmentation model. Since it is a specific application of the prior art in this embodiment, the details will not be elaborated.

[0078] Further, after the iteration ends, the finally trained liver eight-segment segmentation model is output based on the optimal model. Among them, a model will be obtained after each iteration of training and verification. The optimal model is selected from the models of each iteration as the final target model, that is, the finally trained liver eight-segment segmentation model.

[0079] Further, in an embodiment of the invention, as Figure 4 shown, training the liver eight-segment segmentation model specifically includes the following steps:

[0080] Step S402, perform image processing on the input image to obtain a target sequence, where the image processing method at least includes block operation and summation operation;

[0081] In step S404, during the block operation, the input image is blocked according to a preset format, and each block image is converted into a one-dimensional vector by using a serialization operation;

[0082] In step S406, during the summation operation, a summation operation is performed on the position information of each block image and the one-dimensional vector sequence to obtain the target sequence.

[0083] It should be noted that in this embodiment, after obtaining the input image, image processing needs to be performed on it to obtain the target sequence. Among them, the processing process includes first dividing into blocks, then transposing the dimensions, and finally performing a summation operation to obtain the target sequence. Specifically, first, a block operation is performed on the input image. In an embodiment of the invention, first, the picture specification of the input image is obtained. The input image specification is H*W, where H is the height of the input image and W is the width of the input image. Then, the block operation is performed on the input image to obtain the block image. The block image is a square image, and the block image specification is P*P, and N = H*W / (P*P), where P is the side length of the block image and N is the number value of the block images. Further, after obtaining the block image, dimension conversion is performed on it. Specifically, the block image is converted into a one-dimensional vector N*D, where D is the sequence length value after serialization.

[0084] Further, after completing the block operation, a summation operation is performed. First, the position of each block image in the original image (input image) is obtained to obtain the corresponding position information, and then a summation operation is performed based on the position information of each block image and the one-dimensional vector sequence to obtain the target sequence.

[0085] Further, in an embodiment of the invention, as Figure 5 shown, training the liver eight-segment segmentation model specifically further includes the following steps:

[0086] Step S502, performing normalization processing on the target sequence and then changing the dimensions to obtain the first processed data;

[0087] Step S504, performing a convolution operation and activation function processing on the first processed data to obtain the second processed data;

[0088] Step S506, processing the second processed data using a structured state space model to obtain the third processed data;

[0089] Step S508, multiplying the first processed data after being processed by the activation function with the third processed data, and changing the dimensions to obtain the fourth processed data;

[0090] Step S510, restoring the feature map dimension based on the fourth processed data to obtain the fifth processed data;

[0091] Step S512, performing upsampling and convolution operations based on the fifth processed data to restore the feature map and the target mask.

[0092] It should be noted that in this embodiment, since it is described in the above embodiment that the liver eight-segment segmentation model based on deep learning includes, in addition to the input layer and the output layer, a normalization layer, a linear layer, a convolutional layer, and an activation layer, therefore, when each layer processes data, the corresponding processing methods are different. Specifically, after obtaining the target sequence, the target sequence is normalized (RMS Norm), and the dimension is changed (increased to three times the original) to obtain the first processed data X1 (N * 3D). Then, a convolutional operation (2D convolutional operation with padding of "1" and kernel size of "3") and an activation function (silu) are performed on the first processed data to obtain the second processed data X2 (N * 3D); the second processed data is processed using a selective state space model to obtain the third processed data X3 (N * 3D); the first processed data is processed using an activation function and then multiplied by the third processed data, and the dimension is changed (reduced to one time) to obtain the fourth processed data X4 (N * D); based on the fourth processed data, the feature map dimension is restored to obtain the fifth processed data X5 (H / P * W / P * D); based on the fifth processed data, upsampling and convolutional operations are performed to restore the feature map (H * W * K) and the target mask (Mask).

[0093] Furthermore, by continuously iteratively optimizing and training the liver eight-segment segmentation model with the input images (training set and validation set), a trained and optimal liver eight-segment segmentation model can finally be obtained for subsequent applications. Thus, in actual applications, a to-be-segmented image can be directly input, and after being processed by the trained liver eight-segment segmentation model in this embodiment, the corresponding segmentation result can be obtained based on the model output.

[0094] The embodiment of the present application also provides a training system for a liver eight-segment segmentation model based on deep learning. The training system for the liver eight-segment segmentation model based on deep learning can implement the training method for the liver eight-segment segmentation model based on deep learning described in the present application. However, the implementation devices for the training method for the liver eight-segment segmentation model based on deep learning described in the present application include, but are not limited to, the structures of the training system for the liver eight-segment segmentation model based on deep learning listed in this embodiment. Any structural deformation and substitution of the prior art made according to the principles of the present application are included in the protection scope of the present application.

[0095] Please refer to Figure 6 , in one embodiment, a training system 60 for a liver eight-segment segmentation model based on deep learning provided in this embodiment includes:

[0096] An acquisition module 61 for acquiring an input image to obtain a training set and a validation set, where the input image includes liver images with data labels;

[0097] A training module 62 for inputting the training set into the liver eight-segment segmentation model for testing, validating using the validation set, and performing iterative training. Wherein, when the number of iterations reaches the target iteration value, the iteration stops, or when the loss function reaches the target loss value, the iteration stops;

[0098] An output module 63 for, after the iteration ends, outputting the finally trained liver eight-segment segmentation model based on the optimal model.

[0099] Since the specific implementation manner of this embodiment corresponds to the foregoing method embodiment, the same details will not be repeated here. Those skilled in the art should also understand that Figure 6 The division of each module in the embodiment is only a logical function division. In actual implementation, it can be fully or partially integrated into one or more physical entities. And these modules can all be implemented in the form of software called by processing elements, or all in the form of hardware, or some modules can be implemented in the form of software called by processing elements, and some modules in the form of hardware.

[0100] In several embodiments provided by the present invention, it should be understood that the disclosed system, device or method can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of modules / units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of devices or modules or units can be in electrical, mechanical or other forms.

[0101] The modules / units described as separate components may or may not be physically separated, and the components displayed as modules / units may or may not be physical modules, that is, they can be located in one place, or distributed to multiple network units. Some or all of the modules / units can be selected according to actual needs to achieve the purpose of the embodiments of the present invention. For example, in each embodiment of the present invention, the functional modules / units can be integrated in one processing module, or each module / unit can exist physically alone, or two or more modules / units can be integrated in one module / unit.

[0102] Those of ordinary skill in the art should also be able to further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0103] Embodiments of the present invention also provide a computer-readable storage medium. Those of ordinary skill in the art can understand that all or part of the steps in the methods of the above embodiments can be completed by instructing a processor through a program. The said program can be stored in a computer-readable storage medium. The storage medium is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disc, and any combination thereof. The above storage medium can be any available medium that a computer can access or a data storage device such as a server or a data center that integrates one or more available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a digital video disc (DVD)), or a semiconductor medium (such as a solid-state disk (SSD)), etc.

[0104] Embodiments of the present invention also provide an electronic device. The electronic device includes a processor and a memory.

[0105] The memory is used to store a computer program.

[0106] The memory includes various media that can store program codes, such as ROM, RAM, magnetic disk, USB flash drive, memory card, or optical disc.

[0107] The processor is connected to the memory and is used to execute the computer program stored in the memory, so that the electronic device executes the above-mentioned training method of the liver eight-segment segmentation model based on deep learning.

[0108] Preferably, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0109] As Figure 7 shown, the electronic device of the present invention is presented in the form of a general computing device. The components of the electronic device may include, but are not limited to: one or more processors or processing units 71, a memory 72, and a bus 73 connecting different system components (including the memory 72 and the processing unit 71).

[0110] The bus 73 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus structures. By way of example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0111] The electronic device typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, removable and non-removable media.

[0112] The memory 72 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 721 and / or cache memory 722. The electronic device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, a storage system 723 may be used for reading and writing non-removable, non-volatile magnetic media ( Figure 7 not shown, commonly referred to as a "hard disk drive"). Although Figure 7Not shown in the figure, a disk drive for reading and writing a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) can be provided. In these cases, each drive can be connected to the bus 73 through one or more data medium interfaces. The memory 72 may include at least one program product having a set (such as at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0113] A program / utility 724 having a set (at least one) of program modules 7241 can be stored in, for example, the memory 72. Such program modules 7241 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. The program modules 7241 generally perform the functions and / or methods in the embodiments described in the present invention.

[0114] The electronic device can also communicate with one or more external devices (such as a keyboard, a pointing device, a display, etc.), can also communicate with one or more devices that enable a user to interact with the electronic device, and / or can communicate with any device that enables the electronic device to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 74. And, the electronic device can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN) and / or a public network, such as the Internet) through the network adapter 75. As Figure 7 shown, the network adapter 75 communicates with other modules of the electronic device through the bus 73. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0115] Furthermore, in an embodiment of the invention, the present invention also provides a segmentation method of a liver eight-segment segmentation model based on deep learning, which is applied to the liver eight-segment segmentation model trained by the training method of the liver eight-segment segmentation model based on deep learning according to any one of the claims. Among them, as Figure 8 shown, the segmentation method includes the following steps:

[0116] Step S802, obtaining an image to be segmented;

[0117] Step S804, inputting the image to be segmented into the trained liver eight-segment segmentation model for processing to obtain a segmentation result;

[0118] Step S806: Output a target segmented image based on the segmentation result.

[0119] It should be noted that in this embodiment, the actual application of the trained liver eight-segment segmentation model is specifically described. Among them, referring to Figure 9 , where Figure 9 the input image to be segmented is a three-dimensional image. The image to be segmented is input into the trained liver eight-segment segmentation model for processing to obtain a segmentation result, and then a corresponding target segmented image is output based on the segmentation result. As Figures 10A - 10C shown, where Figure 10A shows a cross-sectional view of the segmentation result, Figure 10B shows a front view of the segmentation result, Figure 10C shows a side view of the segmentation result. It should be noted that in order to better show the comparison before and after, Figure 11 a cross-sectional view of the image to be segmented is shown. By Figure 10A comparing with Figure 11 , it can be seen that the front and back segmentations of the trained liver eight-segment segmentation model in this embodiment can well distinguish the segmented results of the liver region after image segmentation.

[0120] The above embodiments are only illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A training method for a liver eight-segment segmentation model based on deep learning, characterized in that, The method includes: Obtaining input images to obtain a training set and a validation set, where the input images include liver images with data labels; Inputting the training set into the liver eight-segment segmentation model for testing, and using the validation set for verification for iterative training. Among them, when the number of iterations reaches the target iteration value, the iteration stops, or when the loss function reaches the target loss value, the iteration stops; After the iteration ends, output the finally trained liver eight-segment segmentation model based on the optimal model, where Training the liver eight-segment segmentation model specifically includes: performing image processing on the input images to obtain a target sequence. Among them, the image processing methods at least include block operation and summation operation; during block operation, the input images are blocked according to a preset format, and each block image is converted into a one-dimensional vector by serialization operation; during summation operation, a summation operation is performed based on the position information of each block image and the one-dimensional vector sequence to obtain the target sequence, where During block operation, it specifically includes: obtaining the picture specification of the input image, where the input image specification is H*W, where H is the height of the input image and W is the width of the input image; performing block operation on the input image to obtain block images, where the block images are square images, the block image specification is P*P, and N = H*W / (P*P), P is the side length of the block image, and N is the number value of the block images; converting the block images into N one-dimensional vectors N*D, where D is the length value of the serialized sequence; Training the liver eight-segment segmentation model specifically further includes: performing normalization processing on the target sequence and then changing the dimension to obtain the first processed data; performing convolution operation and activation function processing on the first processed data to obtain the second processed data; processing the second processed data using a structured state space model to obtain the third processed data; multiplying the first processed data after being processed by the activation function with the third processed data, and performing dimension change to obtain the fourth processed data; restoring the feature map dimension based on the fourth processed data to obtain the fifth processed data; performing upsampling and convolution operation based on the fifth processed data to restore the feature map and the target mask.

2. The training method of the liver eight-segment segmentation model based on deep learning according to claim 1, characterized in that, The loss function used during training the liver eight-segment segmentation model is the Dice-CE Loss function, and the ratio of the training set to the validation set is 4:

1.

3. A liver eight-segment segmentation model based on deep learning, characterized in that, Trained by using the training method of the liver eight-segment segmentation model based on deep learning according to any one of claims 1-2, the model includes: An input layer for receiving input data to obtain a target sequence; A normalization layer for normalizing the target sequence; A linear layer for changing the dimension of the intermediate processed data of the current model; A convolutional layer for extracting features from the intermediate processed data of the current model; An activation layer for performing non-linear transformation on the output of the convolutional layer by using an activation function; An output layer for calculating and processing the model by using a preset loss function and outputting the model segmentation result.

4. A training system for a liver eight-segment segmentation model based on deep learning, characterized in that, Includes: An acquisition module for obtaining input images to obtain a training set and a validation set, where the input images include liver images with data labels; A training module, configured to input the training set into the liver eight-segment segmentation model for testing, and use the validation set for validation and perform iterative training. Among them, when the number of iterations reaches the target iteration value, the iteration stops, or when the loss function reaches the target loss value, the iteration stops; An output module, configured to, after the iteration ends, output the finally trained liver eight-segment segmentation model based on the optimal model. Among them, training the liver eight-segment segmentation model specifically includes: performing image processing on the input image to obtain a target sequence, where the image processing method at least includes block operation and summation operation; during the block operation, the input image is divided into blocks according to a preset format, and each block image is converted into a one-dimensional vector by using serialization operation; during the summation operation, a summation operation is performed on the position information of each block image and the one-dimensional vector sequence to obtain the target sequence. Among them, during the block operation, it specifically includes: obtaining the picture specification of the input image, where the input image specification is H*W, where H is the height of the input image and W is the width of the input image; performing block operation on the input image to obtain block images, where the block images are square images, the block image specification is P*P, and N = H*W / (P*P), P is the side length of the block image, and N is the number value of the block images; converting the block images into N one-dimensional vectors N*D, where D is the length value of the serialized sequence; training the liver eight-segment segmentation model specifically further includes: performing normalization processing on the target sequence and then changing the dimension to obtain first processed data; performing convolution operation and activation function processing on the first processed data to obtain second processed data; processing the second processed data by using a structured state space model to obtain third processed data; multiplying the first processed data after being processed by the activation function and the third processed data, and performing dimension change to obtain fourth processed data; restoring the feature map dimension based on the fourth processed data to obtain fifth processed data; Performing upsampling and convolution operation based on the fifth processed data to restore and obtain a feature map and a target mask.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the training method of the liver eight-segment segmentation model based on deep learning according to any one of claims 1 to 2.

6. An electronic device, characterized in that, The electronic device includes: a processor and a memory; among them, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the training method of the liver eight-segment segmentation model based on deep learning according to any one of claims 1 to 2.

7. A segmentation method for a liver eight-segment segmentation model based on deep learning, characterized in that, Applied to the liver eight-segment segmentation model trained by the training method of the liver eight-segment segmentation model based on deep learning according to any one of claims 1-2, where the segmentation method includes: Obtaining an image to be segmented; Inputting the image to be segmented into the trained liver eight-segment segmentation model for processing to obtain a segmentation result; Outputting a target segmented image based on the segmentation result.

Images