A method for image three-dimensional tissue segmentation and measurement based on deep neural network

Through deep neural network and CT scanning technology, the fat, muscle and bone tissue of pig bodies are automatically segmented and measured, solving the problem that traditional methods require slaughtering and cannot predict growth conditions, and achieving rapid and accurate tissue grading and classification, suitable for The entire animal population.

CN112164073BActive Publication Date: 2024-11-26JIANGNAN UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010999256.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-09-22
Publication Date
2024-11-26
Estimated Expiration
2040-09-22

AI Technical Summary

Technical Problem

Traditional methods require manual dissection and separation of tissues after slaughtering in live animals, which cannot predict the growth status of animals and is not suitable for the three-dimensional tissue grading and classification of fat, muscle and bones of the entire animal population.

Method used

Using the three-dimensional imaging tissue segmentation and measurement method based on deep neural network, the CT bed segmentation network and visceral segmentation network are constructed through CT scan and construction, and the fat, muscle and bone tissue and their proportion of pigs are automatically segmented and measured, reducing the slaughtering. loss.

Benefits of technology

The rapid and accurate segmentation and determination of pig body tissue is achieved, suitable for breeding pigs of any shape and size, reduce unnecessary slaughter losses and predict animal growth conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112164073B_ABST
    Figure CN112164073B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for image three-dimensional tissue segmentation and measurement based on a deep neural network, including collecting CT images of living pigs, dividing them into training sets and test sets, and annotating the training sets; constructing a CT bed segmentation network and an internal organ segmentation network, and using the annotated training sets to train to obtain a CT bed segmentation model and an internal organ segmentation model; using the CT bed segmentation model and the internal organ segmentation model to predict their mask images and remove the CT bed and internal organs; combining the CT images of the living pigs to extract the fat, muscle and bone parts of the pig body and calculate the overall mass of the pig body and the proportion of each tissue. The fat, muscle and bone tissues of the breeding pig can be automatically, quickly and accurately segmented, and the method is applicable to breeding pigs of any shape and size.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision and deep learning, and in particular to a method for image three-dimensional tissue segmentation and measurement based on a deep neural network. Background Art

[0002] The grading and classification of farmed animals is generally based on the proportion of fat, lean meat and bone in the body. The traditional measurement method requires that bones, fat, muscles and other tissues be separated from the live body by manual dissection after slaughter, so as to calculate the proportion of the three. Although this method is accurate, it requires a certain amount of manpower and material resources, and its measurement results cannot predict the growth status of animals, so it cannot be applied to the entire animal population.

[0003] Computed Tomography (CT) is a method of obtaining internal microscopic structural information of an object under non-destructive conditions. Its main principle is to reflect the density of an object based on the degree of X-ray absorption by the object. CT is one of the earliest non-invasive imaging methods used in clinical medical diagnosis. It can display the complete three-dimensional information of an object, making the organ structure clearer and thus discovering some tiny lesions. With the development of computer technology and scientific and technological levels, the application field of CT technology has become more extensive. In animal husbandry production, combining CT with computer technology to determine the content of various tissues in living animals is one of the current research hotspots. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above-mentioned problems of the existing methods for monitoring fat content in living pig muscles, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is to provide a method for three-dimensional tissue segmentation and measurement based on deep neural networks. Only by performing a CT scan on a living pig can the weight and proportion of the pig's bones, muscles, and fat tissues be quickly and accurately segmented and measured, thereby reducing unnecessary losses caused by slaughtering pigs.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: collecting CT images of living pigs, dividing them into a training set and a test set, and annotating the training set; constructing a CT bed segmentation network and a viscera segmentation network, and using the annotated training set to train to obtain a CT bed segmentation model and a viscera segmentation model; using the CT bed segmentation model and the viscera segmentation model to predict their mask images and remove the CT bed and viscera; combining the CT images of the living pigs to extract the fat, muscle and bone parts of the pig body and calculate the overall mass of the pig body and the proportion of each tissue.

[0008] As a preferred solution of the image three-dimensional tissue segmentation and measurement method based on deep neural network described in the present invention, wherein: the construction of the CT bed segmentation network includes, the CT bed segmentation network is mainly composed of an encoder and a decoder, the encoder includes three submodules, each submodule includes two a×a convolution operations, a residual block, and a b×b maximum pooling; the decoder includes three submodules, each submodule includes a c×c deconvolution, two d×d convolution operations and a residual block, wherein each convolution operation in the network is followed by a nonlinear activation function, and the function of the convolution operation is expressed as follows:

[0009]

[0010] Among them: f(x,y) is the input image feature, h(i,j) is the convolution kernel, and g(x,y) is the output feature; the function of the defective block is expressed as follows:

[0011] x l+1 =x l +F(x)

[0012] Where: x l is the input of the residual block of this layer, F(x) is the output after a series of convolution operations, x l+1 is the output of the residual block; the "residual attention path" is used to connect the layers corresponding to the encoder and the decoder. The residual attention path consists of several residual blocks, each of which is composed of an e×e convolution kernel and an f×f convolution kernel that are accumulated in parallel. The number of residual blocks used is determined by the complexity of each layer; the final result is output using h×h convolution and the activation function of the neural network, where the activation function of the neural network is defined as follows:

[0013]

[0014] As a preferred solution of the method for three-dimensional tissue segmentation and measurement based on deep neural network of the present invention, the training of the CT bed segmentation model includes putting the manually marked CT bed and its original image into the CT bed segmentation network for training to obtain the CT bed segmentation model, and the loss function thereof is selected as binary cross entropy, which is defined as follows:

[0015]

[0016] Among them, X is the original image, Y is the manually annotated CT bed, is the image predicted by the network, and px is a pixel.

[0017] As a preferred embodiment of the method for three-dimensional tissue segmentation and measurement based on deep neural network described in the present invention, it is characterized in that: the construction of the visceral segmentation network includes: the visceral segmentation network includes a residual dense connection network and a bidirectional convolutional long short-term memory model, wherein the construction of the residual dense connection network consists of a series of residual dense connection blocks, downsampling and upsampling operations; the bidirectional convolutional long short-term memory model consists of two convolutional long short-term models in different directions, and the function definition of the convolutional long short-term model is as follows:

[0018]

[0019]

[0020]

[0021]

[0022]

[0023] Among them: * is the convolution operation, is the Hadamard product, X n , C n and H n is the input, cell state and hidden state, f n 、i n and n They are forget gate, input gate and output gate respectively, b i 、b f 、b c and b o is the bias, W is the weight matrix; the result of the residual dense connection network is used as the input of the bidirectional convolutional long short-term memory model, and finally the normalized index function is used to output the result.

[0024] As a preferred embodiment of the method for three-dimensional tissue segmentation and measurement based on deep neural network described in the present invention, the training of the visceral segmentation model includes putting the manually marked viscera and their original images into the visceral segmentation network for training to obtain the visceral segmentation model, and its loss function selects the same binary cross entropy.

[0025] As a preferred embodiment of the method for three-dimensional tissue segmentation and measurement based on deep neural network described in the present invention, the predicted CT bed mask image includes selecting any CT image in the remaining test set, inputting it into the CT bed segmentation model, and obtaining the CT bed mask image predicted by the network.

[0026] As a preferred solution of the method for three-dimensional tissue segmentation and measurement based on deep neural network described in the present invention, the predicted CT bed mask image is first inverted, and then an "AND" operation is performed with the original image, and the original image array and the mask image array are matched pixel by pixel. If the value of a pixel in the CT scanning bed image array is 1, the pixel value of the position in the original CT image array is set to 0 to obtain an image without the CT bed.

[0027] As a preferred embodiment of the method for three-dimensional tissue segmentation and measurement based on deep neural network of the present invention, the removal of internal organs includes: inputting the original CT image into the internal organ segmentation model for prediction to obtain a predicted internal organ mask map; performing an erosion operation on the internal organ mask map to remove noise points, wherein the erosion operation function is expressed as follows:

[0028]

[0029] in: is an erosion operation, x is the displacement of the set translation, S1 is the structural element used for the erosion, X1 is the original binary image set, and Z1 is the result image obtained after the erosion operation; the eroded image Z1 is expanded to restore its original size to obtain a new new visceral mask image, wherein the function of the expansion operation is expressed as follows:

[0030]

[0031] in: is a dilation operation, x is the displacement of the set translation, S2 is the structural element used for dilation, X2 is the image set before the dilation operation, and Z2 is the new visceral mask map obtained after the dilation; the viscera are removed from the image without the CT bed according to the new visceral mask map, the new visceral mask map is inverted, and then “AND” is performed with the image without the CT bed to obtain the image without the viscera.

[0032] As a preferred embodiment of the method for three-dimensional tissue segmentation and measurement based on deep neural network of the present invention, the extraction of fat, muscle and bone of the pig body includes extracting the fat, muscle and bone parts in the image after removing the internal organs according to the CT value range of the fat, muscle and bone respectively, and counting the number of voxel points of each.

[0033] As a preferred embodiment of the method for image three-dimensional tissue segmentation and measurement based on deep neural network described in the present invention, the calculation of the overall mass of the pig body and the proportion of each tissue includes calculating the weight of the fat, muscle and bone, and the calculation formula is as follows:

[0034] Wf=Fn×PS×ST×ρf

[0035] Wb=Bn×PS×ST×ρb

[0036] Wl=Ln×PS×ST×ρl

[0037] Where: Wf is the weight of fat, Wl is the weight of muscle, Wb is the weight of bone, Fn is the number of voxel points of fat, Bn is the number of voxel points of muscle, Ln is the number of voxel points of bone, PS is the pixel spacing of CT, ST is the slice thickness, ρf, ρb and ρl are the density of fat, muscle and bone respectively, and their values ​​are 0.92g / cm 3 , 1.06g / cm 3 , 1.4g / cm 3 ; The proportion formula of fat, muscle and bone is as follows:

[0038] W=Wf+Wb+Wl

[0039] Pf=Wf / W

[0040] Pb=Wb / W

[0041] Pl=Wl / W

[0042] Where: W is the total mass of the pig, ρf is the proportion of fat, ρb is the proportion of muscle, and ρl is the proportion of bones.

[0043] The beneficial effects of the present invention are as follows: the fat, muscle, bone and other tissues of the sow can be automatically, quickly and accurately segmented, and the method is applicable to sows of any shape and size; the 3D CT bed segmentation method proposed in the present invention can accurately segment the CT scanning bed and avoid changes that may occur in different positions of the CT bed under different pressures; the visceral segmentation model proposed in the present invention combines the residual dense connection network and the bidirectional convolutional long short-term memory model, which can take into account the semantic information inside the slice and the information between the slices, and can segment the visceral organs of the pig at one time, which is helpful to realize the three-dimensional tissue segmentation and measurement of CT images. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0045] Figure 1 This is a schematic flow chart of the method for image three-dimensional tissue segmentation and measurement based on a deep neural network according to the first embodiment of the present invention;

[0046] Figure 2 A cross-sectional view of a sow according to the method for three-dimensional tissue segmentation and measurement based on a deep neural network according to the second embodiment of the present invention;

[0047] Figure 3 A CT bed mask image predicted by the method for three-dimensional tissue segmentation and measurement based on a deep neural network according to the second embodiment of the present invention;

[0048] Figure 4 The image after removing the CT bed of the method for three-dimensional tissue segmentation and measurement based on a deep neural network described in the second embodiment of the present invention;

[0049] Figure 5 The visceral mask image of the image three-dimensional tissue segmentation and measurement method based on deep neural network described in the second embodiment of the present invention;

[0050] Figure 6 The image after removing the internal organs of the method for three-dimensional tissue segmentation and measurement based on a deep neural network described in the second embodiment of the present invention;

[0051] Figure 7 This is a diagram showing the effect of extracting fat tissue in the method for three-dimensional tissue segmentation and measurement based on a deep neural network according to the second embodiment of the present invention;

[0052] Figure 8 This is a diagram showing the effect of extracting muscle tissue according to the method for image three-dimensional tissue segmentation and measurement based on a deep neural network according to the second embodiment of the present invention;

[0053] Fig. 9 This is a diagram showing the effect of extracting bone tissue in the method for three-dimensional tissue segmentation and measurement based on deep neural networks described in the second embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0057] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0058] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0059] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] Example 1

[0061] Reference Figure 1 , which is the first embodiment of the present invention, provides a method for image three-dimensional tissue segmentation and measurement based on a deep neural network, comprising:

[0062] S1: Collect CT images of living pigs, divide them into training set and test set, and annotate the training set.

[0063] S2: Construct a CT bed segmentation network and an internal organ segmentation network, and use the labeled training set to train to obtain a CT bed segmentation model and an internal organ segmentation model. It should be noted that

[0064] The CT bed segmentation network is mainly composed of an encoder and a decoder. The encoder includes three submodules, each of which includes two a×a convolution operations, a residual block, and a b×b maximum pooling; the decoder includes three submodules, each of which includes a c×c deconvolution, two d×d convolution operations and a residual block. Each convolution operation in the network is followed by a nonlinear activation function. The function of the convolution operation is expressed as follows:

[0065]

[0066] Among them: f(x,y) is the input image feature, h(i,j) is the convolution kernel, and g(x,y) is the output feature; the function of the defective block is expressed as follows:

[0067] x l+1 =x l +F(x)

[0068] Where: x l is the input of the residual block of this layer, F(x) is the output after a series of convolution operations, x l+1is the output of the residual block; the "residual attention path" is used to connect the corresponding layers of the encoder and decoder. The residual attention path consists of several residual blocks, each of which is composed of an e×e convolution kernel and an f×f convolution kernel in parallel and then accumulated. The number of residual blocks used is determined by the complexity of each layer. In the present invention, 3, 2, and 1 residual blocks are used in the corresponding layers from top to bottom respectively; the final result is output using h×h convolution and the activation function (Sigmoid) of the neural network, where the activation function (Sigmoid) of the network is defined as follows:

[0069]

[0070] The visceral segmentation network includes a residual dense connection network and a bidirectional convolutional long short-term memory model. The construction of a residual dense connection network consists of a series of residual dense connection blocks, downsampling and upsampling operations; the bidirectional convolutional long short-term memory model consists of two convolutional long short-term models in different directions. The function definition of the convolutional long short-term model is as follows:

[0071]

[0072]

[0073]

[0074]

[0075] Among them: * is the convolution operation, is the Hadamard product, X n , C n and H n is the input, cell state and hidden state, f n 、i n and n They are respectively the forget gate, input gate and output gate, b i 、b f 、b c and b o is the bias, W is the weight matrix; the result of the residual dense connection network is used as the input of the bidirectional convolutional long short-term memory model, and finally the normalized exponential function is used to output the result.

[0076] Furthermore, training the CT bed segmentation model includes putting the manually labeled CT bed and its original image into the CT bed segmentation network for training to obtain the CT bed segmentation model, and its loss function selects binary cross entropy, which is defined as follows:

[0077]

[0078] Among them, X is the original image, Y is the manually annotated CT bed, is the image predicted by the network, and px is a pixel. Training the viscera segmentation model includes putting the manually marked viscera and their original images into the viscera segmentation network for training to obtain the viscera segmentation model, and its loss function selects the same binary cross entropy.

[0079] S3: Use the CT bed segmentation model and the viscera segmentation model to predict the mask image and remove the CT bed and viscera. It should be noted that:

[0080] Removing the CT bed includes first inverting the predicted CT bed mask image, then performing an AND operation with the original image, matching the original image array with the mask image array pixel by pixel, and if the value of a pixel in the CT scanning bed image array is 1, then setting the pixel value at that position in the original CT image array to 0 to obtain an image with the CT bed removed.

[0081] The removal of internal organs includes inputting the original CT image into the internal organ segmentation model for prediction to obtain a predicted internal organ mask map; performing an erosion operation on the internal organ mask map to remove noise points, wherein the erosion operation function is expressed as follows:

[0082]

[0083] in: is the corrosion operation, x is the displacement of the set translation, S1 is the structural element used for corrosion, X1 is the original binary image set, and Z1 is the result image obtained after the corrosion operation; the corroded image Z1 is expanded to restore its original size to obtain a new new visceral mask image, where the function of the expansion operation is expressed as follows:

[0084]

[0085] in: is the dilation operation, x is the displacement of the set translation, S2 is the structural element used for dilation, X2 is the image set before the dilation operation, and Z2 is the new visceral mask map obtained after the dilation; the viscera are removed from the image without the CT bed according to the new visceral mask map, the new visceral mask map is inverted, and then “AND” is performed with the image without the CT bed to obtain the image without the viscera.

[0086] S4: Combine the CT images of live pigs to extract the fat, muscle and bone parts of the pig and calculate the overall weight of the pig and the proportion of each tissue. It should be noted that

[0087] Extracting the fat, muscle and bone of the pig body includes extracting the fat, muscle and bone parts in the image without internal organs according to the CT value ranges of the fat, muscle and bone respectively, and counting the number of voxel points of each.

[0088] Furthermore, the weight of fat, muscle and bone is calculated as follows:

[0089] Wf=Fn×PS×ST×ρf

[0090] Wb=Bn×PS×ST×ρb

[0091] Wl=Ln×PS×ST×ρl

[0092] Where: Wf is the weight of fat, Wl is the weight of muscle, Wb is the weight of bone, Fn is the number of voxel points of fat, Bn is the number of voxel points of muscle, Ln is the number of voxel points of bone, PS is the pixel spacing of CT, which is 0.977*0.977mm 2 , ST is the slice thickness of 4.999 mm, ρf, ρb and ρl are the densities of fat, muscle and bone respectively; the proportion of fat, muscle and bone is as follows:

[0093] W=Wf+Wb+Wl

[0094] Pf=Wf / W

[0095] Pb=Wb / W

[0096] Pl=Wl / W

[0097] Where: W is the total mass of the pig, ρf is the proportion of fat, ρb is the proportion of muscle, and ρl is the proportion of bones.

[0098] Example 2

[0099] Reference Figures 2 to 9 , which is the second embodiment of the present invention, 40 healthy sows are selected and injected with sedatives, and the positions of the sows are fixed in a way that their limbs are separated and sent to a CT machine to take CT images. The CT image of a sow is regarded as a three-dimensional array T (x, y, z), and finally 40 three-dimensional arrays are generated to obtain CT image slices, and the image data of 10 sows are selected as a training set, and the image set of the remaining 30 pigs is used as a test set, and the CT bed and the pig's internal organs are manually marked on the training set; the marked internal organ training set is used to train the internal organ segmentation network to obtain the internal organ segmentation model and use the CT bed segmentation model to predict the CT bed mask image on the test set, and the CT bed is removed from the original image, and the fat, muscle, and bone parts of the pig body are extracted in combination with the pig's CT value, wherein the CT value range of fat is (-200, 0), and the obtained fat tissue is referenced Figure 7 The CT value range of muscle is (0,200), and the muscle tissue obtained is referenced Figure 8 The CT value range of bones is (230,1500), and the muscle tissue obtained is referenced Fig. 9; Based on the density information, the fat weight, muscle weight, bone weight and the proportion of each tissue in the pig body are calculated. The density values ​​of fat, muscle and bone are 0.92g / cm 3 , 1.06g / cm 3 , 1.4g / cm 3 , through calculation, the weights of fat, muscle and bone are 39.10kg, 62.82kg and 7.07kg respectively, so the total weight is 108.99kg. Through calculation, it can be obtained that the proportion of fat is 35.87%, the proportion of muscle is 57.64%, and the proportion of bone is 6.49%, realizing the quantitative analysis of fat, muscle, bone and other tissues of breeding pigs.

[0100] In order to better verify and illustrate the technical effect adopted in the method of the present invention, the artificial dissection method is selected for testing in this embodiment, and the test results are compared by means of scientific demonstration to verify the real effect of the method; the same 40 breeding pigs are dissected by the artificial dissection method, and the weight and proportion are calculated by dividing the fat, muscle and bone parts of the breeding pigs. The total mass of the breeding pigs is 107.71 kg, of which the fat accounts for 34.38%, the muscle accounts for 59.18% and the bone accounts for 6.44%. The analysis results of the same 40 breeding pigs by the two methods are shown in Table 1 below:

[0101] Table 1: Data analysis results

[0102]

[0103] As can be seen from Table 1, the total mass and proportion data obtained by the manual dissection method are not much different from the data obtained by the method of the present invention, and the average error does not exceed 2%. Therefore, it can be seen that the error in the segmentation and measurement of three-dimensional tissues using the method of the present invention is very small, and the results can be accurately measured. Compared with the manual dissection method, the method of the present invention can measure the breeding pigs without harming them, and its growth can be predicted, and it is suitable for the entire animal population.

[0104] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for image three-dimensional tissue segmentation and measurement based on deep neural network, characterized in that: include, Collect CT images of live pigs, divide them into training set and test set, and annotate the training set; Construct a CT bed segmentation network and a visceral segmentation network, and use the labeled training set to train to obtain a CT bed segmentation model and a visceral segmentation model; Using the CT bed segmentation model and the viscera segmentation model to predict the mask image and remove the CT bed and the viscera; Extracting the fat, muscle and bone parts of the pig body in combination with the CT image of the living pig and calculating the overall weight of the pig body and the proportion of each tissue; The construction of the CT bed segmentation network includes: The CT bed segmentation network is mainly composed of an encoder and a decoder. The encoder includes three submodules, each of which includes two a×a convolution operations, a residual block, and a b×b maximum pooling; the decoder includes three submodules, each of which includes a c×c deconvolution, two d×d convolution operations and a residual block, where each convolution operation in the network is followed by a nonlinear activation function, and the function of the convolution operation is expressed as follows: Among them: f(x,y) is the input image feature, h(i,j) is the convolution kernel, and g(x,y) is the output feature; the function of the defective block is expressed as follows: x l+1 =x l +F(x) Where: x l is the input of the residual block of this layer, F(x) is the output after a series of convolution operations, x l+1 is the output of the residual block; the "residual attention path" is used to connect the layers corresponding to the encoder and the decoder. The residual attention path consists of several residual blocks, each of which is composed of an e×e convolution kernel and an f×f convolution kernel that are accumulated in parallel. The number of residual blocks used is determined by the complexity of each layer; the final result is output using h×h convolution and the activation function of the neural network, where the activation function of the neural network is defined as follows: Building a visceral segmentation network includes: The visceral segmentation network includes a residual dense connection network and a bidirectional convolutional long short-term memory model, wherein the construction of the residual dense connection network consists of a series of residual dense connection blocks, downsampling and upsampling operations; the bidirectional convolutional long short-term memory model consists of two convolutional long short-term models in different directions, and the function definition of the convolutional long short-term model is as follows: Among them: * is the convolution operation, is the Hadamard product, X n , C n and H n is the input, cell state and hidden state, f n 、i n and n They are respectively the forget gate, input gate and output gate, b i , b f , b c and b o is the bias, W is the weight matrix; the result of the residual dense connection network is used as the input of the bidirectional convolutional long short-term memory model, and finally the normalized index function is used to output the result.

2. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: Training the CT bed segmentation model includes, The manually marked CT bed and its original image are put into the CT bed segmentation network for training to obtain a CT bed segmentation model, and its loss function is selected as binary cross entropy, which is defined as follows: Among them, X is the original image, Y is the manually annotated CT bed, is the image predicted by the network, and px is a pixel.

3. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: Training the visceral segmentation model includes, The manually marked viscera and their original images are put into the viscera segmentation network for training to obtain the viscera segmentation model, and the loss function thereof selects the same binary cross entropy.

4. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: Predicted CT bed mask includes, Any CT image is selected from the remaining test sets and input into the CT bed segmentation model to obtain the CT bed mask predicted by the network.

5. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: The removing of the CT bed comprises: First, the predicted CT bed mask image is inverted, and then an AND operation is performed with the original image, and the original image array and the mask image array are matched pixel by pixel. If the value of a pixel in the CT scanning bed image array is 1, the pixel value of the position in the original CT image array is set to 0, and an image without the CT bed is obtained.

6. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: Evisceration includes: The original CT image is input into the visceral segmentation model for prediction to obtain a predicted visceral mask map; the visceral mask map is corroded to remove noise points, wherein the corrosion operation function is expressed as follows: in: is an erosion operation, x is the displacement of the set translation, S1 is the structural element used for the erosion, X1 is the original binary image set, and Z1 is the result image obtained after the erosion operation; the eroded image Z1 is expanded to restore its original size to obtain a new new visceral mask image, wherein the function of the expansion operation is expressed as follows: in: is a dilation operation, x is the displacement of the set translation, S2 is the structural element used for dilation, X2 is the image set before the dilation operation, and Z2 is the new visceral mask map obtained after the dilation; the viscera are removed from the image without the CT bed according to the new visceral mask map, the new visceral mask map is inverted, and then the new visceral mask map is ANDed with the image without the CT bed to obtain the image without the viscera.

7. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: The extraction of fat, muscle and bone from the pig body includes: According to the CT value ranges of the fat, muscle and bone respectively, the fat, muscle and bone parts are extracted from the image without internal organs, and the number of voxel points of each is counted.

8. The method for three-dimensional tissue segmentation and measurement based on deep neural network according to claim 1, characterized in that: The calculation of the overall weight of the pig and the proportion of each tissue includes: The weight of the fat, muscle and bone is calculated as follows: Wf=Fn×PS×ST×ρf Wb=Bn×PS×ST×ρb Wl=Ln×PS×ST×ρl Where: Wf is the weight of fat, Wl is the weight of muscle, Wb is the weight of bone, Fn is the number of voxel points of fat, Bn is the number of voxel points of muscle, Ln is the number of voxel points of bone, PS is the pixel spacing of CT, ST is the slice thickness, ρf, ρb and ρl are the densities of fat, muscle and bone respectively; the proportion formula of fat, muscle and bone is as follows: W=Wf+Wb+Wl Pf=Wf / W Pb=Wb / W Pl=Wl / W Where: W is the total mass of the pig, ρf is the proportion of fat, ρb is the proportion of muscle, and ρl is the proportion of bones.