CN117333736A - Res2 net-based liver fibrosis detection model training method and system - Google Patents
Res2 net-based liver fibrosis detection model training method and system Download PDFInfo
- Publication number
- CN117333736A CN117333736A CN202311286872.4A CN202311286872A CN117333736A CN 117333736 A CN117333736 A CN 117333736A CN 202311286872 A CN202311286872 A CN 202311286872A CN 117333736 A CN117333736 A CN 117333736A
- Authority
- CN
- China
- Prior art keywords
- model
- training
- liver
- res2net
- layer
- 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
- 238000012549 training Methods 0.000 title claims abstract description 86
- 208000019425 cirrhosis of liver Diseases 0.000 title claims abstract description 51
- 238000001514 detection method Methods 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000011176 pooling Methods 0.000 claims abstract description 42
- 206010019668 Hepatic fibrosis Diseases 0.000 claims abstract description 8
- 210000004185 liver Anatomy 0.000 claims description 61
- 230000006870 function Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 11
- 210000002966 serum Anatomy 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 230000003176 fibrotic effect Effects 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 230000003902 lesion Effects 0.000 abstract description 7
- 230000006872 improvement Effects 0.000 abstract description 3
- 239000003814 drug Substances 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 5
- 206010016654 Fibrosis Diseases 0.000 description 4
- 230000004761 fibrosis Effects 0.000 description 4
- 208000019423 liver disease Diseases 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 210000005228 liver tissue Anatomy 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- MXBCYQUALCBQIJ-RYVPXURESA-N (8s,9s,10r,13s,14s,17r)-13-ethyl-17-ethynyl-11-methylidene-1,2,3,6,7,8,9,10,12,14,15,16-dodecahydrocyclopenta[a]phenanthren-17-ol;(8r,9s,13s,14s,17r)-17-ethynyl-13-methyl-7,8,9,11,12,14,15,16-octahydro-6h-cyclopenta[a]phenanthrene-3,17-diol Chemical compound OC1=CC=C2[C@H]3CC[C@](C)([C@](CC4)(O)C#C)[C@@H]4[C@@H]3CCC2=C1.C1CC[C@@H]2[C@H]3C(=C)C[C@](CC)([C@](CC4)(O)C#C)[C@@H]4[C@@H]3CCC2=C1 MXBCYQUALCBQIJ-RYVPXURESA-N 0.000 description 1
- 206010061818 Disease progression Diseases 0.000 description 1
- 208000035965 Postoperative Complications Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000740 bleeding effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000005750 disease progression Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000007542 hardness measurement Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000002113 ultrasound elastography Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30056—Liver; Hepatic
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a liver fibrosis detection model training method and system based on Res2net, and relates to the technical field of medicine, wherein the method comprises the steps of determining a training data set; constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain a hepatic fibrosis detection model; and training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model. The invention can comprehensively utilize multi-mode data to more comprehensively capture the characteristics of hepatic fibrosis lesions, and secondly, by improving the Res2net model, introducing an index weighted model and adding a global average pooling layer and a convolution layer, the understanding capability of the model on complex image data is improved, the hepatic fibrosis lesions can be more accurately identified and positioned by the improvement, and the detection accuracy is effectively improved.
Description
Technical Field
The invention relates to the technical field of medicine, in particular to a liver fibrosis detection model training method and system based on Res2 net.
Background
Liver disease has become a significant cause of influence on human health worldwide. The progress of various liver diseases is accompanied by fibrosis of the liver, and fibrosis of the liver is accompanied by changes in the elasticity of the liver. Traditional liver fibrosis diagnosis requires liver tissue acquisition through a puncture needle, and the operation process is not only invasive, but also local liver tissue acquisition cannot completely represent the disease progression degree of the whole liver. In addition, some patients may be at risk of postoperative complications such as pain, bleeding, etc. after the puncture. Although there have been developed foreign serum models (such as APRI, FIB-4) and advanced liver hardness detection devices (e.g. ultrasound elastography), the accuracy of liver fibrosis prediction results by the above method is low, and the AUC value of the serum model applied to domestic patient data is usually about 0.7. While the accuracy of the liver hardness testing device is relatively high, measurement errors are likely to occur, thus also affecting its usefulness.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a liver fibrosis detection model training method and system based on Resnet, which are used for solving the problem of low accuracy of the existing liver fibrosis detection.
The technical scheme of the invention is as follows:
liver fibrosis detection model training method based on Resnet
Determining a training data set, wherein the training data set comprises liver sample CT image data, and serum index data and liver fibrosis degree data corresponding to the liver sample CT image data, and the number of each type of training samples in the training data set is consistent;
constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain a hepatic fibrosis detection model;
and training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model.
Preferably, the method further comprises pre-processing liver sample image data, the pre-processing comprising:
extracting liver areas from all CT images by adopting a liver segmentation algorithm;
correcting the segmented liver region, and reducing the noise influence of the image acquisition process on the image;
and cutting the liver image through the 3D connected domain, reducing irrelevant areas, and then scaling the liver image to a preset specification.
Preferably, the preprocessing further comprises:
manually labeling the fibrotic degree livers of the liver images;
and determining the proportion of the liver images corresponding to each degree, and carrying out data enhancement processing on the liver images according to the proportion.
Preferably, the improving the Res2net model includes: the channel attention extraction module and the spatial attention module are added after the Res2net model is added to the first layer and the second layer, the global average pooling layer is added before the third layer, and the output of the original image and the first spatial attention module is led to the third layer.
Preferably, the channel attention module comprises a full-pooling layer, an average pooling layer, two layers of perceptrons and an activation function, the characteristics processed by the first layer are respectively input into the full-pooling layer and the average pooling layer, the outputs of the full-pooling layer and the average pooling layer are simultaneously input into the perceptrons at two sides, and the outputs of the two layers of perceptrons are multiplied by the characteristics processed by the first layer through the activation function after being added.
Preferably, the spatial attention module comprises an average pooling and maximum pooling layer, a convolution layer and an activation function.
Preferably, the method of training the liver fibrosis detection model using the training dataset comprises:
determining training parameters, wherein determining the training parameters comprises using random gradient descent as an optimizer during training, setting initial power to 0.9, learning rate to 0.001, and weight attenuation coefficient to 0.0001;
and obtaining a hepatic fibrosis detection model based on the training parameters and the training data set for the improved Res2net model until the model error is smaller than a preset value.
A Res2net based liver fibrosis detection model training system comprising:
the system comprises a determining module, a processing module and a processing module, wherein the determining module is used for determining a training data set, the training data set comprises liver sample CT image data, serum index data and liver fibrosis degree data corresponding to the liver sample CT image data, and the number of each type of training sample in the training data set is consistent;
the construction module is used for constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain an initial liver fibrosis detection model;
and the training module is used for training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model.
A computer storage medium having stored thereon a computer program which, when executed by a processor, implements a method as claimed in any one of the preceding claims.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing a method as claimed in any one of the preceding claims when executing the computer program.
The beneficial effects of the invention are as follows: the liver fibrosis detection model training method and system based on Res2net can comprehensively utilize multi-mode data, including CT/MRI image data and serum index data of liver samples, so that the characteristics of liver fibrosis lesions can be more comprehensively captured, and secondly, by improving the Res2net model, introducing an index weighting model and adding a global average pooling layer and a convolution layer, the understanding capability of the model on complex image data is improved, the liver fibrosis lesions can be more accurately identified and positioned by improvement, and the detection accuracy is effectively improved. In addition, the strategy that the quantity of training samples of each type in the training data set is consistent is adopted, so that the problem of unbalance of the samples is avoided, and the utilization rate of the data is improved. The liver fibrosis detection model obtained by training the method has higher precision and robustness, and can be widely applied to diagnosis and treatment of clinical liver diseases.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. Like elements or portions are generally identified by like reference numerals throughout the several figures. In the drawings, elements or portions thereof are not necessarily drawn to scale.
FIG. 1 is a flowchart of a liver fibrosis detection model training method based on Res2net provided by an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a liver fiber detection model according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an improved Res2net model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a channel attention module and a spatial attention module according to an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of a Res2 net-based liver fibrosis detection model training system according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a computer device according to an embodiment of the present invention.
Detailed Description
Embodiments of the technical scheme of the present invention will be described in detail below with reference to the accompanying drawings. The following examples are only for more clearly illustrating the technical aspects of the present invention, and thus are merely examples, and are not intended to limit the scope of the present invention.
It is noted that unless otherwise indicated, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
Example 1
As shown in fig. 1 and 2, the method for training a liver fibrosis detection model based on Res2net provided by the embodiment of the present invention includes: determining a training data set, wherein the training data set comprises liver sample CT image data, and serum index data and liver fibrosis degree data corresponding to the liver sample CT image data, and the number of each type of training samples in the training data set is consistent; constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain a hepatic fibrosis detection model; and training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model.
The liver fibrosis detection model training method based on the Res2net can comprehensively utilize multi-mode data, including CT/MRI image data and serum index data of liver samples, so that the characteristics of liver fibrosis lesions are more comprehensively captured, and secondly, by improving the Res2net model, introducing an index weighting model and adding a global average pooling layer and a convolution layer, the understanding capability of the model on complex image data is improved, the liver fibrosis lesions can be more accurately identified and positioned by improvement, and the detection accuracy is effectively improved. In addition, the strategy that the quantity of training samples of each type in the training data set is consistent is adopted, so that the problem of unbalance of the samples is avoided, and the utilization rate of the data is improved. The liver fibrosis detection model obtained by training the method has higher precision and robustness, and can be widely applied to diagnosis and treatment of clinical liver diseases.
In this embodiment, the weighting model is as follows:
h(x)=h θ (x)=θ 0 +θ 1 x 1 +θ 2 x 2 +θ 3 x 3 +…+θ n x n
θ in the model n Represents the serum index, theta in the nth 0 Representing the error. Since in actual operation x 0 When=1, x 0 Can be regarded as theta 0 x 0 Thus, the initial security value calculation model can be rewritten as:
in this embodiment, the method further comprises preprocessing the liver sample image data, the preprocessing comprising: extracting liver areas from all CT images by adopting a liver segmentation algorithm; correcting the segmented liver region, and reducing the noise influence of the image acquisition process on the image; clipping the liver image through the 3D connected domain, reducing irrelevant areas, and then scaling the liver image to a preset specification; manually labeling the fibrotic degree livers of the liver images; and determining the proportion of the liver images corresponding to each degree, and carrying out data enhancement processing on the liver images according to the proportion.
Specifically, a liver segmentation algorithm is adopted to process all CT images, liver regions are extracted, other irrelevant regions are removed, the segmented liver regions are corrected, influences such as noise and artifacts possibly existing in an image acquisition process are reduced, the image quality and accuracy are improved, the liver images are cut through a 3D connected domain technology, the irrelevant regions are removed, and only the parts relevant to liver lesions are reserved. Therefore, the data volume can be reduced, the efficiency of subsequent processing is improved, the cut liver images are scaled to the preset specification size so as to ensure the consistency and stability of an input model, each liver image is manually marked, the fibrosis degree of each liver image is determined, label information is provided for the supervised learning of the subsequent model, the proportion of the liver image corresponding to each fibrosis degree is determined, and the data enhancement processing is carried out on the images according to the proportion. The data expansion can be performed by adopting technologies such as rotation, translation, scaling, overturning and the like, so that the diversity of training samples is increased, and the generalization capability of the model is improved.
Through the preprocessing steps, liver image data can be better processed, interference and noise are reduced, data quality is improved, and reliable input and basis are provided for subsequent model training and liver fibrosis detection.
As shown in fig. 3, in this embodiment, the modification of the Res2net model includes: the channel attention extraction module and the spatial attention module are added after the Res2net model is added to the first layer and the second layer, the global average pooling layer is added before the third layer, and the output of the original image and the first spatial attention module is led to the third layer.
By adding the channel attention extraction module and the space attention module after the first layer and the second layer of the Res2net model, the relevance among channels and the importance in space can be enhanced in the feature extraction stage, so that the perceptibility of the model to different scales and features is improved, invalid features are restrained, and the network is focused on the feature extraction of the target area. In addition, a global average pooling layer is introduced before the third layer, the original image is introduced into the third layer, so that original data can be effectively reserved, data loss is reduced, and meanwhile, deeper data can be obtained by combining the output of the second attention module fused with the first attention module, so that overfitting is reduced, generalization capability of a model is improved, and feature extraction capability is further improved.
As shown in fig. 4, in this embodiment, the channel attention module includes a full pooling layer, an average pooling layer, two-layer perceptron and an activation function, and the spatial attention module includes an average pooling and maximum pooling layer, a convolution layer and an activation function.
Specifically, in the channel attention module, the output U (size h×w×c) of the first layer is subjected to global maximum pooling and average pooling to obtain two feature graphs of 1×1×c, the results are respectively input into two layers of perceptrons (MLPs) with shared weights, the output features are added and subjected to an activation function to generate a channel attention weight MC., and finally the input feature graph is multiplied by the channel attention weight to generate a channel attention feature graph F', wherein the calculation process is as follows:
Z avg =AvgPool(U),
Z max =MaxPool(U),
M C =σ[f 2 (δ(f 1 (Z avg )))+f 2 (δ(f 1 (Z max )))]
F′=M C ×U,
wherein, avgPool represents global average pooling, and the pixel average value of each channel is calculated; maxPool represents global max pooling, preserving the pixel max of the feature map for each channel; f (f) 1 Representing the full connection layer of the input channel C and the output channel C/16; f (f) 2 Representing the input channel C/16, the full connection layer of the output channel C; delta represents a ReLU function; sigma represents Sigmoid function.
Specifically, in the spatial attention module, the channel attention feature map F' is used as input to perform average pooling and maximum pooling of channel dimensions to obtain two feature maps of h×wx1, then the two feature maps are spliced and pass through a convolution layer of 7×7, and the result passes through a Sigmoid function to obtain a spatial attention weight Ms. of a two-dimensional single channel, and the spatial attention weight is multiplied by the channel attention feature map to generate a feature map f″ with spatial and channel attention at the same time, and the calculation process is as follows:
Z AVG =AVGPool(U),
Z MAX =MAXPool(U),
M s =σ(f 3 (Z AVG ,Z MAX )),
F″=M s ×F″,
wherein AVGPool represents channel dimension average pooling, characteristics for each channelAveraging the pixel addition corresponding to the graph; MAXPool represents the maximum pooling of channel dimensions, and the maximum pixel value of the corresponding position of each channel feature map is reserved; f (f) 3 Representing a convolution.
In this embodiment, the method for training the improved Res2net model by using the training data set includes: determining training parameters, wherein determining the training parameters comprises using random gradient descent as an optimizer during training, setting initial power to 0.9, learning rate to 0.001, and weight attenuation coefficient to 0.0001; and obtaining a hepatic fibrosis detection model based on the training parameters and the training data set for the improved Res2net model until the model error is smaller than a preset value.
Example 2
As shown in fig. 5, the liver fibrosis detection model training system based on Resnet provided in this embodiment includes: the system comprises a determining module, a processing module and a processing module, wherein the determining module is used for determining a training data set, the training data set comprises liver sample CT image data, serum index data and liver fibrosis degree data corresponding to the liver sample CT image data, and the number of each type of training sample in the training data set is consistent;
the construction module is used for constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain an initial liver fibrosis detection model; and the training module is used for training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model.
As shown in fig. 6, on the basis of the above embodiment, the embodiment of the present invention further provides a computer device, including a storage, a processor, and a computer program stored on the storage and executable on the processor, where the processor implements the method according to any one of the above embodiments when executing the computer program.
On the basis of the above embodiments, the present embodiment further provides a computer storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements a method as described in any of the above.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention, and are intended to be included within the scope of the appended claims and description.
Claims (10)
1. The liver fibrosis detection model training method based on Res2net is characterized by comprising the following steps:
determining a training data set, wherein the training data set comprises liver sample CT image data, and serum index data and liver fibrosis degree data corresponding to the liver sample CT image data, and the number of each type of training samples in the training data set is consistent;
constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain a hepatic fibrosis detection model;
and training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model.
2. The method of claim 1, further comprising preprocessing liver sample image data, the preprocessing comprising:
extracting liver areas from all CT images by adopting a liver segmentation algorithm;
correcting the segmented liver region, and reducing the noise influence of the image acquisition process on the image;
and cutting the liver image through the 3D connected domain, reducing irrelevant areas, and then scaling the liver image to a preset specification.
3. The method of claim 2, wherein the preprocessing further comprises:
manually labeling the fibrotic degree livers of the liver images;
and determining the proportion of the liver images corresponding to each degree, and carrying out data enhancement processing on the liver images according to the proportion.
4. The method of claim 1, wherein improving the Res2net model comprises: the channel attention extraction module and the spatial attention module are added after the Res2net model is added to the first layer and the second layer, the global average pooling layer is added before the third layer, and the output of the original image and the first spatial attention module is led to the third layer.
5. The method of claim 4, wherein the channel attention module comprises a full pooling layer, an average pooling layer, two layers of sensors and an activation function, wherein the features processed by the first layer are respectively input to the full pooling layer and the average pooling layer, the outputs of the full pooling layer and the average pooling layer are simultaneously input to the two layers of sensors, and the outputs of the two layers of sensors are added and multiplied by the features processed by the first layer through the activation function.
6. The method of claim 4, wherein the spatial attention module comprises, the spatial attention module comprising an average pooling and max pooling layer, a convolution layer, and an activation function.
7. The method of claim 1, wherein training the liver fibrosis detection model using the training dataset comprises:
determining training parameters, wherein determining the training parameters comprises using random gradient descent as an optimizer during training, setting initial power to 0.9, learning rate to 0.001, and weight attenuation coefficient to 0.0001;
and obtaining a target liver fibrosis detection model based on the training parameters and the training data set for the improved Res2net model until the model error is smaller than a preset value.
8. A Res2net based liver fibrosis detection model training system comprising:
the system comprises a determining module, a processing module and a processing module, wherein the determining module is used for determining a training data set, the training data set comprises liver sample CT image data, serum index data and liver fibrosis degree data corresponding to the liver sample CT image data, and the number of each type of training sample in the training data set is consistent;
the construction module is used for constructing a Res2net model, improving the Res2net model, constructing an index weighted model, and adding a global average pooling layer and a convolution layer after the improved Res2net model and the index weighted model to obtain a liver fibrosis detection model;
and the training module is used for training the liver fibrosis detection model by using the training data set to obtain a target liver fibrosis detection model.
9. A computer storage medium having stored thereon a computer program which, when executed by a processor, implements the method of any of claims 1-7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any of claims 1-7 when the computer program is executed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311286872.4A CN117333736A (en) | 2023-10-07 | 2023-10-07 | Res2 net-based liver fibrosis detection model training method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311286872.4A CN117333736A (en) | 2023-10-07 | 2023-10-07 | Res2 net-based liver fibrosis detection model training method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117333736A true CN117333736A (en) | 2024-01-02 |
Family
ID=89294805
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311286872.4A Pending CN117333736A (en) | 2023-10-07 | 2023-10-07 | Res2 net-based liver fibrosis detection model training method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117333736A (en) |
-
2023
- 2023-10-07 CN CN202311286872.4A patent/CN117333736A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111951344B (en) | Magnetic resonance image reconstruction method based on cascade parallel convolution network | |
CN112258530A (en) | Neural network-based computer-aided lung nodule automatic segmentation method | |
CN114581662B (en) | Brain tumor image segmentation method, system, device and storage medium | |
CN113516659A (en) | Medical image automatic segmentation method based on deep learning | |
WO2022213654A1 (en) | Ultrasonic image segmentation method and apparatus, terminal device, and storage medium | |
CN113034507A (en) | CCTA image-based coronary artery three-dimensional segmentation method | |
CN117078692B (en) | Medical ultrasonic image segmentation method and system based on self-adaptive feature fusion | |
CN109003280B (en) | Method for segmenting intima in blood vessel by two-channel intravascular ultrasonic image | |
CN114882996B (en) | Hepatocellular carcinoma CK19 and MVI prediction method based on multitask learning | |
CN114511581B (en) | Multi-task multi-resolution collaborative esophageal cancer lesion segmentation method and device | |
CN113256670A (en) | Image processing method and device, and network model training method and device | |
CN109215035B (en) | Brain MRI hippocampus three-dimensional segmentation method based on deep learning | |
CN117409002A (en) | Visual identification detection system for wounds and detection method thereof | |
CN112767403A (en) | Medical image segmentation model training method, medical image segmentation method and device | |
CN111681297A (en) | Image reconstruction method, computer device, and storage medium | |
CN116883341A (en) | Liver tumor CT image automatic segmentation method based on deep learning | |
CN114693671B (en) | Lung nodule semi-automatic segmentation method, device, equipment and medium based on deep learning | |
CN116503506A (en) | Image reconstruction method, system, device and storage medium | |
CN117333736A (en) | Res2 net-based liver fibrosis detection model training method and system | |
CN116188479A (en) | Hip joint image segmentation method and system based on deep learning | |
CN116138877A (en) | Target positioning method, target positioning device, electronic equipment and storage medium | |
CN115375787A (en) | Artifact correction method, computer device and readable storage medium | |
CN113379770B (en) | Construction method of nasopharyngeal carcinoma MR image segmentation network, image segmentation method and device | |
CN114937044A (en) | Lightweight image segmentation method and device and storage medium | |
CN114821049A (en) | Method, device and equipment for segmenting image 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 |