CN114332133A - New coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net - Google Patents
New coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net Download PDFInfo
- Publication number
- CN114332133A CN114332133A CN202210009185.7A CN202210009185A CN114332133A CN 114332133 A CN114332133 A CN 114332133A CN 202210009185 A CN202210009185 A CN 202210009185A CN 114332133 A CN114332133 A CN 114332133A
- Authority
- CN
- China
- Prior art keywords
- module
- image
- features
- new coronary
- coronary pneumonia
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 206010035664 Pneumonia Diseases 0.000 title claims abstract description 38
- 230000011218 segmentation Effects 0.000 title claims abstract description 33
- 208000015181 infectious disease Diseases 0.000 claims abstract description 21
- 230000002776 aggregation Effects 0.000 claims abstract description 14
- 238000004220 aggregation Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 38
- 230000007246 mechanism Effects 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 11
- 238000010586 diagram Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 11
- 230000005284 excitation Effects 0.000 claims description 10
- 238000003709 image segmentation Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 9
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000004927 fusion Effects 0.000 claims description 5
- 210000004072 lung Anatomy 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 4
- 239000011800 void material Substances 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000001125 extrusion Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 239000000284 extract Substances 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 6
- 208000025721 COVID-19 Diseases 0.000 abstract description 5
- 238000002591 computed tomography Methods 0.000 abstract description 5
- 238000004590 computer program Methods 0.000 description 8
- 230000035945 sensitivity Effects 0.000 description 7
- 238000003860 storage Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 208000032376 Lung infection Diseases 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a new coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net. Secondly, a feature aggregation module is introduced, and a bilinear interpolation method is adopted, so that image features of different levels are fused, an expression with higher discrimination capability is obtained, and the segmentation precision of the network is further improved. The method can better capture the characteristics of the new coronary pneumonia infection area in the CT image on the COVID-19-CT-Scans data set, has good segmentation effect, and is obviously improved compared with the original CE-Net network and other segmentation algorithms on the whole.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a new coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net.
Background
Although the existing deep learning algorithm achieves good results in the aspect of processing the new coronary pneumonia image, the related work of segmenting the new coronary pneumonia infection region in the image is still less, because the following difficulties exist in segmenting the infection region from the two-dimensional CT image: 1) there are large differences in the location, size, shape of the infection in different two-dimensional CT images, which often leads to false negative detections. 2) The infected area has low contrast with the normal area. 3) The boundaries of the infected area are often obscured and it is difficult to obtain a sufficiently accurate label.
Disclosure of Invention
In order to make up for the blank and the defects of the prior art, the invention provides a new coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net, which are used for improving the segmentation precision of new coronary pneumonia infected areas.
Firstly, an attention mechanism SE module is added in an encoding stage to introduce global context information, the receptive field of a feature extraction stage is enhanced, the weight of a target related feature channel is increased, and therefore the segmentation capability of a small target is improved. Secondly, a feature aggregation module is introduced, and a bilinear interpolation method is adopted, so that image features of different levels are fused, an expression with higher discrimination capability is obtained, and the segmentation precision of the network is further improved. The method can better capture the characteristics of the new coronary pneumonia infection area in the CT image on the COVID-19-CT-Scans data set, has good segmentation effect, and is obviously improved compared with the original CE-Net network and other segmentation algorithms on the whole.
The invention specifically adopts the following technical scheme:
a new coronary pneumonia CT image infected area segmentation method based on improved CE-Net is characterized by comprising the following steps:
step S1: preprocessing data of the data set, performing image enhancement on all CT images, finding out the outline of the lung parenchyma, and cutting the part except the outline;
step S2: inputting the preprocessed image obtained in the step S1 into a coding part of the network, and respectively extracting basic features of the image through a residual block ResNet and an attention mechanism module SE;
step S3: inputting the features obtained in step S2 into a Dense void Convolution (DAC) and a Residual multi-kernel pool (RMP) for capturing more advanced features and retaining more spatial information;
step S4: inputting the features of different scales obtained in the step S2 into a feature fusion module;
step S5: adding the features obtained in the step S3 and the features fused in the step S4, inputting the added features into a decoder part of the network, and obtaining a segmented result through up-sampling and deconvolution processing;
step S6: and optimizing the image segmentation model through a loss function.
The provided image infection region segmentation model comprises an improvement on an original CE-Net model, and the core of the image infection region segmentation model comprises an attention-mechanism squeezing and Excitation module (SE) and a Feature Aggregation Module (FAM).
Further, in step S1, the contrast of the image is enhanced by using a contrast-limited adaptive histogram equalization algorithm, so that the infected area is more easily distinguished from the normal area, and the contour of the lung parenchyma is found by using a canny algorithm, and the parts other than the contour are cut out, so as to reduce the influence of the irrelevant parts to the maximum extent.
Further, in step S2, the encoding portion of the network includes three portions, the first portion uses 1 convolution with 3 × 3 to extract shallow features F0, the second portion uses 4 pre-trained ResNet modules to extract deep features, and the third portion adds an attention mechanism module behind the ResNet module to introduce global context information, enhance the receptive field during the feature extraction stage, and increase the weight of the target-related feature channel.
Further, in step S2, the residual block ResNet takes the shallow feature F0 as an input feature, passes through two convolution kernels with a size of 3 × 3, and then passes through shortcut to perform superposition output on the input and output; the attention mechanism module SE is divided into two operations: and extruding and exciting, wherein the extruding operation carries out global average pooling operation on the input feature map, so that each channel has global information, and the mathematical formula is expressed as follows:
where X is the input signature, i.e., the output of each residual block, H, W, C represents the height, width, and number of channels of the signature, respectively;
the excitation operation stage is used for obtaining the interdependence relation among all channels of the characteristic diagram, the operation firstly inputs the vectors obtained by extrusion into a full connection layer to obtain the vectors of 1 multiplied by (C/r), r is a set constant and is activated by using a ReLu function, then the number of the channels is expanded from C/r to C through a full connection layer, and then the weight coefficient s of the channels is calculated through a Sigmoid function, thereby realizing the excitation operation, wherein the calculation formula is as follows:
s＝Fex(z,W)＝σ(g(z,W))＝σ(W2δ(W1z))
(2)
where σ (-) is sigmoid activation function, δ (-) is ReLu function, w1，w2Convolution kernels of two fully-connected layers; and finally, multiplying the weight coefficient by the corresponding channel number to obtain a result characteristic diagram.
Further, in step S3, the dense hole convolution DAC has 4 cascaded branches, which are increased from 1 to 1, 3 and 5 as the number of atrous convolutions is gradually increased, the acceptance domain of each branch is 3, 7, 9 and 19, a 1 × 1 convolution is applied to each branch to correct linear activation, and the DAC block extracts the features of objects with different sizes by combining the atrous convolutions with different atrous rates;
the residual multi-core pool RMP module is provided with 4 receiving domains with different sizes, namely 2 multiplied by 2,3 multiplied by 3, 5 multiplied by 5 and 6 multiplied by 6, four convolution kernels with different sizes obtain 4 different feature information, 1 multiplied by 1 convolution is added after each layer of pooling, features with the same size as the original features are obtained through linear interpolation, and finally the original features and the features obtained through interpolation are connected.
Further, in step S4, the feature fusion module FAM fuses convolution blocks with different sizes obtained in the encoding process by using a bilinear interpolation method, so as to achieve the purpose of feature reuse.
Further, in step S6, the loss function is a combination of a cross entropy loss function and a die coefficient loss function, which is specifically expressed as follows:
wherein Y & ltY & gt 1, Y2 & ltY & gt, Yb & ltY & gt represents a true value,
the prediction probability is represented, N represents the batch size, sigma (·) corresponds to a sigmoid activation function, and the value of alpha is 0.5.
A new coronary pneumonia CT image infected area segmentation system based on improved CE-Net is characterized in that: based on a computer system, the adopted image segmentation model comprises: the device comprises an encoding module, a context extraction module and a decoding module;
after a new coronary pneumonia data set is preprocessed and input into the coding module, the preprocessed new coronary pneumonia data set passes through a convolution kernel with the size of 3 x 3, then passes through 4 ResNet modules, each ResNet module needs to be extruded and excited through an attention mechanism SE module, and then passes through a dense void convolution DAC (digital-to-analog converter) and a residual multi-core pool RMP (residual multi-core pool) of a context extraction module to be used for capturing more advanced features and reserving more spatial information;
the decoding module consists of an upper sampling layer and a characteristic aggregation module; the upper sampling layer is composed of an deconvolution layer with the size of 3 multiplied by 3 and the step length of 2, the size of an output feature map is consistent with that of the feature map in the corresponding coding process, jump connection with a feature aggregation module is accessed, finally, the new coronary pneumonia infection area and the background image are classified through a Sigmoid activation function, and the segmentation result of the infection area is output.
Compared with the prior art, the invention and the optimized scheme thereof firstly add an attention mechanism SE module in the coding process to introduce global context information, so that the model better pays attention to the relevant characteristics of an infected area in the learning process, and then add a characteristic aggregation module on the original structure of Ce-Net, and the module fully fuses the spatial information of high and low layers to obtain the characteristics with more discrimination capability, thereby obtaining better image segmentation effect.
Drawings
FIG. 1 is a schematic topological structure diagram of a new coronary pneumonia CT image infection region segmentation model based on improved CE-Net according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a topology of an attention mechanism module according to an embodiment of the present invention;
fig. 3 is a schematic topology diagram of a feature aggregation module according to an embodiment of the present invention;
FIG. 4 is a schematic flowchart of a method for segmenting an infected area of a new coronary pneumonia CT image based on improved CE-Net according to an embodiment of the present invention;
FIG. 5 is a graph of comparison results of test images under different algorithms according to an embodiment of the present invention.
Detailed Description
In order to make the features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail as follows:
it should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Referring to fig. 1, fig. 1 is a schematic view of a topological structure of a new coronary pneumonia CT image segmentation model according to an embodiment of the present invention.
The inventor researches and finds that the existing medical image segmentation methods tend to improve the segmentation capability of the network by extracting the image boundary to obtain higher evaluation indexes, but the existing methods hardly consider how to improve the segmentation effect of small targets, so that the embodiment of the invention provides a new coronary pneumonia CT image infected region segmentation model based on improved CE-Net to solve the problems.
The embodiment of the invention provides an image segmentation model as shown in fig. 1, wherein the model comprises 3 stages, namely an encoding stage, a context extraction module and a decoding stage. The method comprises the steps of preprocessing a new coronary pneumonia data set, inputting the preprocessed new coronary pneumonia data set into a network coding part, performing convolution kernel with the size of 3 x 3, and then passing through 4 ResNet modules, wherein each ResNet module needs to be extruded and excited through an attention mechanism SE module. And then go through context extraction modules DAC and RMP to capture more advanced features and retain more spatial information. The decoding part consists of an up-sampling layer and a characteristic aggregation module. The up-sampling layer is composed of a deconvolution layer with the size of 3 x 3 and the step length of 2, and the size of the output characteristic graph is consistent with that of the characteristic graph in the corresponding coding process. However, the up-sampling results in a loss of characteristic information of the partially infected area, and thus a jump connection with a characteristic aggregation module is accessed. And finally, classifying the new coronary pneumonia infection area and the background image through a Sigmoid activation function, and outputting an infection area segmentation result.
Referring to fig. 2 and 3, fig. 2 is a schematic diagram of a topology of an attention mechanism module according to the present invention; fig. 3 is a schematic diagram of a topology structure of a feature aggregation module according to the present invention.
In one embodiment, the attention mechanism module SE is divided into two operations, squeeze (squeeze) and excitation (excitation), and the squeeze operation performs a global average pooling operation on the input feature map, so that each channel has global information, which is expressed by the following mathematical formula:
where X is the input feature map (output of each residual block), H, W, C represents the height, width and number of channels of the feature map respectively
The excitation operation stage can acquire the interdependence relation between the channels of the characteristic diagram. Firstly, inputting a vector obtained by extrusion into a full connection layer to obtain a vector of 1 multiplied by (C/r) (the model r is set as 16), activating by using a ReLu function, then expanding the number of channels from C/r to C through the full connection layer, and then calculating a weight coefficient s of the channels through a Sigmoid function to realize excitation operation, wherein the calculation formula is as follows:
s＝Fex(z,W)＝σ(g(z,W))＝σ(W2δ(W1z))
where σ (-) is sigmoid activation function, δ (-) is ReLu function, w1，w2Convolution kernels of two fully connected layers. And finally, multiplying the weight coefficient by the corresponding channel number to obtain a result characteristic diagram.
As shown in fig. 3, three types of lines represent different operations, a straight line represents a connection between the same sizes, a curved line represents an upward linear interpolation, and a dotted curve represents a downward interpolation. The dark and light volume blocks represent the three inputs and outputs of the module. The main principle of the characteristic aggregation module is to fuse convolution blocks with different sizes by utilizing a bilinear interpolation method so as to achieve the purpose of characteristic reuse. The main process is as follows: firstly, the size obtained by the coding process is 562X 128 and 282The x 256 convolution block interpolates twice and four times, respectively, up to 1122X 128 and 1122Multiplying 256 convolution blocks, and combining the two interpolated convolution blocks with the input size of 112 by concat2The x 64 convolved blocks are fused together to get a size of 1122Volume block of x 448. Then obtaining the size by the same methodIs 562X 448 and 282A convolutional block of x 448, which is,
in an implementation manner, an embodiment of the present invention further provides a segmentation method applied to a new coronary pneumonia infection area, which is described in detail with reference to fig. 4 as follows:
step S1: preprocessing data of the data set, namely performing image enhancement on all CT images, finding out the outline of lung parenchyma, and cutting the parts except the outline to reduce the influence of irrelevant parts to the maximum extent;
step S2: inputting the preprocessed image obtained in step S1 into the encoding part of the network, and extracting the basic features of the graph through the residual block ResNet and the attention mechanism module SE respectively, where the expression is as follows:
F0＝Conv3×3(P)
Fi＝fex(fsq(fre(Fi-1))
where P is the preprocessed image, F0For shallow extracted features, Conv3×3Is a convolution kernel of size 3 × 3, freIs the ResNet module feature extraction function, fsqIs a function of the squeeze operation in the attention mechanism, fexIs an excitation function of operation, FiIs the i-th layer output (i ═ 1,2,3,4) of the encoding section;
step S3: inputting the features obtained in step S2 into a Dense void Convolution (DAC) and a Residual multi-kernel pool (RMP) for capturing more advanced features and retaining more spatial information, where the expression is specifically as follows:
FDAC＝Conv3×3(rate＝1)(F4)+Conv3×3(rate＝3)(F4)+Conv3×3(rate＝5)(F4)
FRMP＝maxpool2×2(FDAC)+maxpool3×3(FDAC)+maxpool5×5(FDAC)+maxpool6×6(FDAC)
wherein FDAC，FRMPThe outputs of DAC module and RMP module, maxpooli×iTo representFor a maximum pooling function with convolution kernel of i × i, Conv3×3(rate＝j)A 3 × 3 convolution with step size j;
step S4: inputting the features with different scales obtained in the step S2 into a feature fusion module, wherein the expression is specifically as follows;
FFAM＝(F1+fI(F2)+fI(F3)
FFAMfor the fused features, FiIs the i-th layer output of the coding part, fIIs an interpolation function;
step S5: adding the features obtained in the step S3 and the features fused in the step S4, inputting the added features into a decoder part of the network, and obtaining a segmented result through upsampling and deconvolution, wherein the specific expression is as follows:
Fdst＝fTranconv(FFAM+FRMP)
wherein ftroncv represents the deconvolution function and Fdst represents the output;
step S6: optimizing the image segmentation model by a loss function;
wherein Y & ltY & gt 1, Y2 & ltY & gt, Yb & ltY & gt represents a true value,
the prediction probability is represented, N represents the batch size, sigma (·) corresponds to a sigmoid activation function, and the value of alpha is 0.5.
In order to better illustrate the effectiveness of the present invention, the embodiment of the present invention also adopts a comparison experiment mode to compare the segmentation effects. The example of the present invention uses the data set using COVID-19-CT-Scans. The data set consisted of 1600 two-dimensional CT images, all of which were collected by the chinese radiology association. The radiologist segments the CT image using different labels to identify areas of lung infection. 1456 of the sheets were used as training sets and 144 of the sheets were used as test sets. The images used in the present example were pre-processed and then resized to 224 x 224 for training. All networks used by the present example are implemented based on a pytorech framework. α of the loss function is set to 0.5; the learning rate was set to 0.0001; the epochs are set to 100 and the model is saved once every 5 epochs. The batch size batch _ size is set to 8, limited by the GPU size. The optimizer selects the RMSprop algorithm for optimization. The Dice Similarity Coefficient (DSC), Sensitivity (SEN), Specificity (SPEC), and mean cross-over ratio (MIOU) were used to evaluate model performance.
The performance of the model is tested by using 144 pictures in the COVID-19-CT-Scans data set, models such as FCN, DeepLabv3+, UNet + + and CE-Net are selected in a comparison experiment to be compared with the experimental results of the invention, the experimental results are shown in Table 1, the performances of the Dice Similarity Coefficient (DSC), Sensitivity (SEN), Specificity (SPEC), average cross-over-parallel ratio (MIOU) and the like of the COVID-19-CT-Scans data set respectively reach 74.32%, 84.25%, 99.14% and 80.34%, compared with a Ce-Net network, the performances of the Dice similarity coefficient, specificity and average cross-over-parallel ratio are respectively improved by 1.91%, 0.16% and 1.26%, although the sensitivity of the invention is not as good as that of the CE-Net, the other three performance indexes are obviously better than the CE-Net, and the whole is better. The performance comparison result shows that the accuracy of segmentation is higher on the premise of ensuring the sensitivity, and the segmentation of the new coronary pneumonia infected area can be better carried out.
TABLE 1 comparison of the Performance of different networks (%)
| Algorithm | DSC | SEN | SPEC | MIOU |
| FCN | 63.03 | 68.08 | 99.14 | 74.34 |
| DeepLabV3+ | 62.95 | 72.72 | 98.86 | 74.11 |
| UNet++ | 67.14 | 83.52 | 98.40 | 75.96 |
| CE-Net | 72.41 | 85.17 | 98.98 | 79.08 |
| Scheme of the embodiment | 74.32 | 84.25 | 99.14 | 80.34 |
As shown in fig. 5, this embodiment selects 5 images in the test set, and uses the above 5 models to segment them, and the result is shown in fig. 5. The first column is an original new coronary pneumonia CT image, the second column is a label image, and the third column to the seventh column are FCN, DeepLabV3+, UNet + +, CE-Net and the segmentation result of the invention in sequence. Comparing the first and second lines of results in fig. 4, it can be seen that when a single object is segmented, the result obtained by the method of the present invention is the closest to the real tag, and the results of the other four methods have larger differences in size, contour, etc. than the real tag. Meanwhile, due to the incompleteness of feature extraction, the other four methods have the phenomenon of wrong segmentation, and the target is segmented in an uninfected area, but the phenomenon does not occur in the method. Comparing the third, fourth and fifth lines shows that the result of the present invention is superior to other methods when segmenting multiple objects. Especially when small objects are segmented, compared with other four methods, the method can segment each small object correctly, and the segmented result is closest to a real label in terms of shape and size.
In conclusion, the invention provides an improved Ce-Net network model aiming at the problems that the infected region in the new coronary pneumonia CT image is difficult to segment due to unobvious features. According to the network, attention mechanism SE modules are added in the coding process to introduce global context information, so that the model pays attention to relevant features of an infected area better in the learning process, and then a feature aggregation module is added on the original structure of Ce-Net, so that the module is fully integrated with high-level and low-level spatial information, the features with discrimination capability are obtained, and a better segmentation effect is obtained.
The above programming scheme provided by this embodiment can be stored in a computer readable storage medium in a coded form, and implemented in a computer program manner, and inputs basic parameter information required for calculation through computer hardware, and outputs a calculation result.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each flow of the flowcharts, and combinations of flows in the flowcharts, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows.
The foregoing is directed to preferred embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow. However, any simple modification, equivalent change and modification of the above embodiments according to the technical essence of the present invention are within the protection scope of the technical solution of the present invention.
The present invention is not limited to the above preferred embodiments, and other various methods and systems for segmenting an infected area of new coronary pneumonia CT image based on improved CE-Net can be derived by anyone based on the teaching of the present invention.
Claims (8)
1. A new coronary pneumonia CT image infected area segmentation method based on improved CE-Net is characterized by comprising the following steps:
step S1: preprocessing data of the data set, performing image enhancement on all CT images, finding out the outline of the lung parenchyma, and cutting the part except the outline;
step S2: inputting the preprocessed image obtained in the step S1 into a coding part of the network, and respectively extracting basic features of the image through a residual block ResNet and an attention mechanism module SE;
step S3: inputting the features obtained in the step S2 into a dense hole convolution DAC and a residual multi-core pool RMP for capturing more advanced features and reserving more spatial information;
step S4: inputting the features of different scales obtained in the step S2 into a feature fusion module;
step S5: adding the features obtained in the step S3 and the features fused in the step S4, inputting the added features into a decoder part of the network, and obtaining a segmented result through up-sampling and deconvolution processing;
step S6: and optimizing the image segmentation model through a loss function.
2. The method for segmenting the infection area of the new coronary pneumonia CT image based on the improved CE-Net in claim 1, wherein: in step S1, the contrast of the image is enhanced by using a contrast-limited adaptive histogram equalization algorithm to make the infected area more easily distinguishable from the normal area, and the contour of the lung parenchyma is found by using a canny algorithm, and the parts other than the contour are cut out to reduce the influence of the irrelevant parts to the maximum extent.
3. The method for segmenting the infection area of the new coronary pneumonia CT image based on the improved CE-Net in claim 1, wherein: in step S2, the encoding portion of the network includes three portions, the first portion uses 1 convolution with 3 × 3 to extract shallow features F0, the second portion uses 4 pretrained ResNet modules to extract deep features, and the third portion adds an attention mechanism module behind the ResNet module to introduce global context information, to enhance the receptive field during the feature extraction phase, and to increase the weight of the target-related feature channel.
4. The method for segmenting the infection area of the new coronary pneumonia CT image based on the improved CE-Net in claim 3, wherein: in step S2, the residual block ResNet takes the shallow feature F0 as an input feature, and performs superposition output on the input and output through two convolution kernels with the size of 3 × 3 and short cut; the attention mechanism module SE is divided into two operations: and extruding and exciting, wherein the extruding operation carries out global average pooling operation on the input feature map, so that each channel has global information, and the mathematical formula is expressed as follows:
where X is the input signature, i.e., the output of each residual block, H, W, C represents the height, width, and number of channels of the signature, respectively;
the excitation operation stage is used for obtaining the interdependence relation among all channels of the characteristic diagram, the operation firstly inputs the vectors obtained by extrusion into a full connection layer to obtain the vectors of 1 multiplied by (C/r), r is a set constant and is activated by using a ReLu function, then the number of the channels is expanded from C/r to C through a full connection layer, and then the weight coefficient s of the channels is calculated through a Sigmoid function, thereby realizing the excitation operation, wherein the calculation formula is as follows:
s＝Fex(z,W)＝σ(g(z,W))＝σ(W2δ(W1z))
(2)
where σ (-) is sigmoid activation function, δ (-) is ReLu function, w1，w2Convolution kernels of two fully-connected layers; finally, multiplying the weight coefficientAnd obtaining a result characteristic diagram according to the corresponding channel number.
5. The method for segmenting the infection area of the new coronary pneumonia CT image based on the improved CE-Net in claim 1, wherein: in step S3, the dense hole convolution DAC has 4 cascaded branches, which are increased from 1 to 1, 3 and 5 as the number of atrous convolutions gradually increases, the acceptance domain of each branch is 3, 7, 9 and 19, a 1 × 1 convolution correction linear activation is applied to each branch, and the DAC block extracts the features of objects with different sizes by combining the atrous convolutions with different atrous rates;
the residual multi-core pool RMP module is provided with 4 receiving domains with different sizes, namely 2 multiplied by 2,3 multiplied by 3, 5 multiplied by 5 and 6 multiplied by 6, four convolution kernels with different sizes obtain 4 different feature information, 1 multiplied by 1 convolution is added after each layer of pooling, features with the same size as the original features are obtained through linear interpolation, and finally the original features and the features obtained through interpolation are connected.
6. The method for segmenting the infection area of the new coronary pneumonia CT image based on the improved CE-Net in claim 1, wherein: in step S4, the feature fusion module FAM fuses convolution blocks with different sizes obtained in the encoding process by using a bilinear interpolation method, so as to achieve the purpose of feature reuse.
7. The method for segmenting the infection area of the new coronary pneumonia CT image based on the improved CE-Net in claim 1, wherein: in step S6, the loss function is a combination of a cross entropy loss function and a die coefficient loss function, which is expressed as follows:
wherein Y & ltY & gt 1, Y2 & ltY & gt, Yb & ltY & gt represents a true value,
the prediction probability is represented, N represents the batch size, sigma (·) corresponds to a sigmoid activation function, and the value of alpha is 0.5.
8. A new coronary pneumonia CT image infected area segmentation system based on improved CE-Net is characterized in that: based on a computer system, the adopted image segmentation model comprises: the device comprises an encoding module, a context extraction module and a decoding module;
after a new coronary pneumonia data set is preprocessed and input into the coding module, the preprocessed new coronary pneumonia data set passes through a convolution kernel with the size of 3 x 3, then passes through 4 ResNet modules, each ResNet module needs to be extruded and excited through an attention mechanism SE module, and then passes through a dense void convolution DAC (digital-to-analog converter) and a residual multi-core pool RMP (residual multi-core pool) of a context extraction module to be used for capturing more advanced features and reserving more spatial information;
the decoding module consists of an upper sampling layer and a characteristic aggregation module; the upper sampling layer is composed of an deconvolution layer with the size of 3 multiplied by 3 and the step length of 2, the size of an output feature map is consistent with that of the feature map in the corresponding coding process, jump connection with a feature aggregation module is accessed, finally, the new coronary pneumonia infection area and the background image are classified through a Sigmoid activation function, and the segmentation result of the infection area is output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210009185.7A CN114332133A (en) | 2022-01-06 | 2022-01-06 | New coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210009185.7A CN114332133A (en) | 2022-01-06 | 2022-01-06 | New coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114332133A true CN114332133A (en) | 2022-04-12 |
Family
ID=81024714
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210009185.7A Pending CN114332133A (en) | 2022-01-06 | 2022-01-06 | New coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114332133A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114882282A (en) * | 2022-05-16 | 2022-08-09 | 福州大学 | Neural network prediction method for colorectal cancer treatment effect based on MRI and CT images |
| CN116543167A (en) * | 2023-07-04 | 2023-08-04 | 真健康（北京）医疗科技有限公司 | CT image segmentation method and device |
| CN116563285A (en) * | 2023-07-10 | 2023-08-08 | 邦世科技(南京)有限公司 | Focus characteristic identifying and dividing method and system based on full neural network |
| CN116702065A (en) * | 2023-05-30 | 2023-09-05 | 浙江时空智子大数据有限公司 | Method and system for monitoring ecological treatment pollution of black and odorous water based on image data |
| WO2024046142A1 (en) * | 2022-08-30 | 2024-03-07 | Subtle Medical, Inc. | Systems and methods for image segmentation of pet/ct using cascaded and ensembled convolutional neural networks |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233117A (en) * | 2020-12-14 | 2021-01-15 | 浙江卡易智慧医疗科技有限公司 | New coronary pneumonia CT detects discernment positioning system and computing equipment |
| WO2021104056A1 (en) * | 2019-11-27 | 2021-06-03 | 中国科学院深圳先进技术研究院 | Automatic tumor segmentation system and method, and electronic device |
| CN112927240A (en) * | 2021-03-08 | 2021-06-08 | 重庆邮电大学 | CT image segmentation method based on improved AU-Net network |
| CN113160226A (en) * | 2021-05-24 | 2021-07-23 | 苏州大学 | Two-way guide network-based classification segmentation method and system for AMD lesion OCT image |
-
2022
- 2022-01-06 CN CN202210009185.7A patent/CN114332133A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021104056A1 (en) * | 2019-11-27 | 2021-06-03 | 中国科学院深圳先进技术研究院 | Automatic tumor segmentation system and method, and electronic device |
| CN112233117A (en) * | 2020-12-14 | 2021-01-15 | 浙江卡易智慧医疗科技有限公司 | New coronary pneumonia CT detects discernment positioning system and computing equipment |
| CN112927240A (en) * | 2021-03-08 | 2021-06-08 | 重庆邮电大学 | CT image segmentation method based on improved AU-Net network |
| CN113160226A (en) * | 2021-05-24 | 2021-07-23 | 苏州大学 | Two-way guide network-based classification segmentation method and system for AMD lesion OCT image |
Non-Patent Citations (1)
| Title |
|---|
| 邱纯乾: ""基于改进CENet的新冠肺炎CT图像感染区域分割"", 《传感器与微系统》, 30 November 2023 (2023-11-30) * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114882282A (en) * | 2022-05-16 | 2022-08-09 | 福州大学 | Neural network prediction method for colorectal cancer treatment effect based on MRI and CT images |
| WO2024046142A1 (en) * | 2022-08-30 | 2024-03-07 | Subtle Medical, Inc. | Systems and methods for image segmentation of pet/ct using cascaded and ensembled convolutional neural networks |
| CN116702065A (en) * | 2023-05-30 | 2023-09-05 | 浙江时空智子大数据有限公司 | Method and system for monitoring ecological treatment pollution of black and odorous water based on image data |
| CN116702065B (en) * | 2023-05-30 | 2024-04-16 | 浙江时空智子大数据有限公司 | Method and system for monitoring ecological treatment pollution of black and odorous water based on image data |
| CN116543167A (en) * | 2023-07-04 | 2023-08-04 | 真健康（北京）医疗科技有限公司 | CT image segmentation method and device |
| CN116543167B (en) * | 2023-07-04 | 2023-09-05 | 真健康（北京）医疗科技有限公司 | CT image segmentation method and device |
| CN116563285A (en) * | 2023-07-10 | 2023-08-08 | 邦世科技(南京)有限公司 | Focus characteristic identifying and dividing method and system based on full neural network |
| CN116563285B (en) * | 2023-07-10 | 2023-09-19 | 邦世科技(南京)有限公司 | Focus characteristic identifying and dividing method and system based on full neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108647585B (en) | Traffic identifier detection method based on multi-scale circulation attention network | |
| CN107564025B (en) | Electric power equipment infrared image semantic segmentation method based on deep neural network | |
| CN114332133A (en) | New coronary pneumonia CT image infected area segmentation method and system based on improved CE-Net | |
| CN109840556B (en) | Image classification and identification method based on twin network | |
| CN108664981B (en) | Salient image extraction method and device | |
| CN111145181B (en) | Skeleton CT image three-dimensional segmentation method based on multi-view separation convolutional neural network | |
| CN113139543B (en) | Training method of target object detection model, target object detection method and equipment | |
| CN111914654B (en) | Text layout analysis method, device, equipment and medium | |
| CN111461213A (en) | Training method of target detection model and target rapid detection method | |
| CN116645592B (en) | Crack detection method based on image processing and storage medium | |
| CN114266794B (en) | Pathological section image cancer region segmentation system based on full convolution neural network | |
| CN114758137B (en) | Ultrasonic image segmentation method and device and computer readable storage medium | |
| CN110852327A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| CN113762265A (en) | Pneumonia classification and segmentation method and system | |
| CN114266894A (en) | Image segmentation method and device, electronic equipment and storage medium | |
| CN114241388A (en) | Video instance segmentation method and segmentation device based on space-time memory information | |
| CN113870286A (en) | Foreground segmentation method based on multi-level feature and mask fusion | |
| US20220262006A1 (en) | Device for detecting an edge using segmentation information and method thereof | |
| CN113657225B (en) | Target detection method | |
| CN111738069A (en) | Face detection method and device, electronic equipment and storage medium | |
| CN116542988A (en) | Nodule segmentation method, nodule segmentation device, electronic equipment and storage medium | |
| CN115937693A (en) | Road identification method and system based on remote sensing image | |
| CN110555462A (en) | non-fixed multi-character verification code identification method based on convolutional neural network | |
| CN115424250A (en) | License plate recognition method and device | |
| CN113657196A (en) | SAR image target detection method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |