CN116030971A - Intelligent auxiliary system for classifying benign and malignant liver tumors - Google Patents
Intelligent auxiliary system for classifying benign and malignant liver tumors Download PDFInfo
- Publication number
- CN116030971A CN116030971A CN202310308198.9A CN202310308198A CN116030971A CN 116030971 A CN116030971 A CN 116030971A CN 202310308198 A CN202310308198 A CN 202310308198A CN 116030971 A CN116030971 A CN 116030971A
- Authority
- CN
- China
- Prior art keywords
- clinical information
- benign
- liver
- malignant
- ultrasonic image
- 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
- 206010073069 Hepatic cancer Diseases 0.000 title claims abstract description 15
- 206010028980 Neoplasm Diseases 0.000 claims abstract description 42
- 230000003902 lesion Effects 0.000 claims abstract description 40
- 210000004185 liver Anatomy 0.000 claims abstract description 35
- 230000003211 malignant effect Effects 0.000 claims abstract description 34
- 238000012545 processing Methods 0.000 claims abstract description 9
- 210000002966 serum Anatomy 0.000 claims abstract description 8
- 230000005540 biological transmission Effects 0.000 claims abstract description 7
- 238000012360 testing method Methods 0.000 claims abstract description 3
- 238000007781 pre-processing Methods 0.000 claims description 22
- 206010019695 Hepatic neoplasm Diseases 0.000 claims description 15
- 208000014018 liver neoplasm Diseases 0.000 claims description 15
- 238000000034 method Methods 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 10
- 230000003321 amplification Effects 0.000 claims description 9
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 9
- 238000002604 ultrasonography Methods 0.000 claims description 9
- 230000000007 visual effect Effects 0.000 claims description 7
- 238000012800 visualization Methods 0.000 claims description 5
- 238000007477 logistic regression Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 abstract description 6
- 201000011510 cancer Diseases 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 7
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000006073 displacement reaction Methods 0.000 description 5
- 101100194362 Schizosaccharomyces pombe (strain 972 / ATCC 24843) res1 gene Proteins 0.000 description 4
- 101100194363 Schizosaccharomyces pombe (strain 972 / ATCC 24843) res2 gene Proteins 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000005489 elastic deformation Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012821 model calculation Methods 0.000 description 2
- 238000009877 rendering Methods 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000002792 vascular Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The application relates to the technical field of medical image processing, in particular to an intelligent auxiliary system for classifying benign and malignant liver tumors, which comprises the following components: the data transmission module is used for acquiring clinical information of an inspector; wherein, the clinical information comprises basic information, serum test indexes and clinical electronic medical record data; the clinical information auxiliary identification module is used for taking clinical information as input of a pre-constructed classifier, and calculating benign probability values and malignant probability values of tumor lesions according to each feature and weight corresponding to each feature contained in the clinical information; the liver ultrasonic image and clinical information of the patient are put into the corresponding model for calculation, and a benign and malignant tumor classification result of the patient is obtained, and as the model and the characteristics of the clinical information and the ultrasonic image are different in two-time identification, the uncertainty of classification can be reduced as much as possible through the two-time identification, and the identification reliability is improved.
Description
Technical Field
The application relates to the technical field of medical image processing, in particular to an intelligent auxiliary system for classifying benign and malignant liver tumors.
Background
With the continuous progress of science and technology, ultrasonic equipment can obtain an ultrasonic image with obvious characteristics and higher definition. In clinical discrimination, the ultrasonic image obtained by ultrasonic examination of the liver can be used as the main discrimination basis, but the ultrasonic image discrimination of liver tumor depends largely on the professional experience and discrimination level of the ultrasonic examination doctor. The difference of the discrimination levels of the ultrasonic examination doctors in different areas and different levels of hospitals is large, and the discrimination of different doctors in the same case is often caused to be different. In addition, the identification of ultrasound images is time consuming, requiring a physician to spend a great deal of time and effort identifying liver tumor lesions to obtain image data, and manually marking important areas.
In recent years, with the continuous development of deep learning, automatic segmentation technology of liver medical images has made great progress, and a computer-aided identification system using a computer image processing technology as a core is an important direction of accurate medical development in the future. The computer-aided identification system can automatically and accurately detect, identify and divide liver tumors based on ultrasonic images, so that the objectivity and accuracy of identification are ensured, and the workload of medical staff is greatly reduced.
However, when the computer-aided identification system is used for identifying the ultrasonic image, the problems of inaccurate image detection and influence of other conditions still exist, and the identification reliability still has room for improvement.
Disclosure of Invention
In order to solve the problems, the application provides an intelligent auxiliary system for classifying benign and malignant liver tumors.
The application provides an intelligent auxiliary system for classifying benign and malignant liver tumors, which adopts the following technical scheme:
an intelligent assistance system for classification of benign and malignant liver tumors, comprising:
the data transmission module is used for acquiring clinical information of an inspector; wherein, the clinical information comprises basic information, serum test indexes and clinical electronic medical record data;
the clinical information auxiliary identification module is used for taking clinical information as input of a pre-constructed classifier, and calculating benign probability values and malignant probability values of tumor lesions according to each feature and weight corresponding to each feature contained in the clinical information;
the clinical information auxiliary authentication module includes:
loading each feature of the clinical information into each pre-trained sub-classifier, inputting the weight corresponding to each feature, and adding the scores obtained by each sub-classifier to obtain a final predicted value;
processing the predicted value by logistic regression to obtain benign and malignant probability distribution of the tumor lesion
The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>
Benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying +.>
。
By adopting the technical scheme, the clinical information of the patient is put into the corresponding model for calculation, the benign and malignant classification result of the tumor of the patient is obtained, and the clinical information is combined with the ultrasonic image result obtained by the existing computer-aided identification system, and as the model and the characteristics of the two times of identification of the clinical information and the ultrasonic image are different, the uncertainty of classification can be reduced as much as possible through the two times of identification, and the identification reliability is improved;
in addition, when the result of the ultrasonic image auxiliary identification module is the same as the result output by the clinical information auxiliary identification module, the case is identified as a malignant case or a benign case; if the identification is different, the suspected case is identified, and only secondary judgment is needed to be carried out on the suspected case, so that the workload is effectively reduced.
Optionally, the data transmission module is further used for acquiring a liver ultrasonic image of the inspector; further comprises:
and the ultrasonic image auxiliary discrimination module is used for taking the liver ultrasonic image as the input of a pre-constructed ultrasonic image feature extractor and outputting benign probability values and malignant probability values of liver tumor lesions.
Optionally, the ultrasound image auxiliary authentication module includes:
inputting a gray level image of a liver ultrasonic image with a standard size, and processing the gray level image of the liver ultrasonic image to obtain a characteristic output image of the liver ultrasonic image; wherein the characteristic output graph comprises tumor size, morphology, echo and presence or absence of halation;
the feature output graph is mapped into a one-dimensional feature vector in a non-linear mode and becomes a 64-dimensional feature vector;
based on the feature vector, outputting benign and malignant probability distribution of tumor lesions by using softmax function
Wherein->
Benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying
。
Optionally, the method further comprises:
and the data preprocessing module is used for screening the clinical information and performing size scaling and data amplification on the liver ultrasonic image.
Optionally, the data preprocessing module includes:
the scaling pretreatment unit is used for scaling the liver ultrasonic image to ensure that the length of the liver ultrasonic image is standard;
the data amplification preprocessing unit is used for changing the positions of the pixel points in the liver ultrasonic image while not changing the relative positions of human tissues in the liver ultrasonic image;
and the clinical information preprocessing unit is used for eliminating incomplete clinical information and extracting main reference characteristics in the clinical information.
Optionally, the clinical information auxiliary authentication module includes:
loading each feature of the clinical information into each pre-trained sub-classifier, inputting the weight corresponding to each feature, and adding the scores obtained by each sub-classifier to obtain a final predicted value;
processing the predicted value by logistic regression to obtain benign and malignant probability distribution of the tumor lesion
The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>
Benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying +.>
。
Optionally, the method further comprises:
and the image information visualization module is used for integrating the liver ultrasonic image, the clinical information and the benign probability value and the malignant probability value of the corresponding liver tumor lesion to perform visual display.
In summary, the present application includes at least one of the following beneficial technical effects:
1. the liver ultrasonic image and clinical information of the patient are put into the corresponding model for calculation, and a benign and malignant classification result of the tumor of the patient is obtained, and as the model and the characteristics of the clinical information and the ultrasonic image are different in two-time identification, the uncertainty of classification can be reduced as much as possible through the two-time identification, and the identification reliability is improved;
2. when the result of the ultrasonic image auxiliary identification module is the same as the result output by the clinical information auxiliary identification module, the case is identified as a malignant case or a benign case; if the identification is different, the suspected case is identified, and only secondary judgment is needed to be carried out on the suspected case, so that the workload is effectively reduced;
3. rendering the ultrasonic image of the liver, clinical information, and benign probability value and malignant probability value of corresponding liver tumor lesions into the image for visual display.
Drawings
Fig. 1 is a block diagram of a smart assistance system according to an embodiment of the present application.
Reference numerals illustrate: 1. a data transmission module; 2. a data preprocessing module; 3. an ultrasonic image auxiliary identification module; 4. a clinical information auxiliary identification module; 5. and an image information visualization module.
Detailed Description
The present application is described in further detail below in conjunction with fig. 1.
The embodiment of the application discloses an intelligent auxiliary system for classifying benign and malignant liver tumors, which effectively reduces uncertainty caused by a single algorithm and improves accuracy of results through mutual verification of clinical information and ultrasonic images.
As an embodiment of the intelligent auxiliary system, it comprises:
the data transmission module 1 is used for acquiring liver ultrasonic images and clinical information of the inspector; the liver ultrasonic image is an ultrasonic image containing liver lesions, and the clinical information comprises basic information, serum inspection indexes and clinical electronic medical record data;
the data preprocessing module 2 is used for screening the clinical information, and performing size scaling and data amplification on the liver ultrasonic image;
the ultrasonic image auxiliary identification module 3 is used for taking the liver ultrasonic image as the input of a pre-constructed ultrasonic image feature extractor and outputting benign probability values and malignant probability values of liver tumor lesions;
the clinical information auxiliary identification module 4 is used for taking clinical information as input of a pre-constructed classifier and calculating benign probability value and malignant probability value of the tumor lesion according to each feature and weight corresponding to each feature contained in the clinical information;
and the image information visualization module 5 is used for integrating the liver ultrasonic image, the clinical information and the benign probability value and the malignant probability value of the corresponding liver tumor lesion to perform visual display.
And establishing an ultrasonic image classification model and a classifier, immediately preprocessing data after uploading liver ultrasonic images and clinical information of a patient by a doctor, and then putting the preprocessed data into a corresponding model for calculation to obtain a benign and malignant tumor classification result of the patient. And integrating all the processed and analyzed information, and performing visual processing for viewing by doctors. The whole treatment process is very rapid, and the efficiency of classifying benign and malignant liver tumors is greatly improved.
Specifically, the data preprocessing module 2 includes a scaling preprocessing unit, a data amplification preprocessing unit, and a clinical information preprocessing unit.
It should be noted that, since the shapes and sizes of the lesion areas of different liver tumors are different, the general ultrasound image tumor segmentation map is manually outlined by a doctor, so that the size difference of the final image is very large, and the size of the ultrasound image needs to be scaled. The scaling preprocessing unit is used for scaling the ultrasonic image so that the final length of the ultrasonic image is standard. The method comprises the following steps: aiming at two directions of the dimension of the long side of the ultrasonic image, one side of the ultrasonic image is scaled to a standard size, and the other side of the ultrasonic image is scaled in an equal proportion; zero padding is carried out on the two directions of the dimension of the short side, so that the final length of the zero padded part is standard size, and the shape of the tumor lesion area is maintained unchanged.
Meanwhile, in order to increase the data volume and improve the accuracy of model calculation, data amplification is required. The data amplification preprocessing unit adopts a mode of horizontal overturning and elastic deformation, and does not change the relative position of human tissues in the ultrasonic image, the shape, the size, the position and the like of a tumor area while changing the position of a pixel point in the ultrasonic image.
The first mode is horizontal overturning, namely overturning by taking the vertical symmetry axis of the ultrasonic image as a central line. The second mode is elastic deformation, and specifically comprises the following steps: first creating a random displacement field to deform the ultrasound image, i.e
Wherein->
Generating a random number uniformly distributed between (-1, 1) for each pixel point, and representing the moving distance of the pixel (x, y) on the longitudinal axis of the horizontal axis; then use standard deviation +.>
Is a Gaussian function of (1) versus displacement field->
Convolving, and mixing the convolved displacement field with deformation control factor +.>
Multiplying to obtain an elastically deformed displacement field; and finally, applying the displacement field to the ultrasonic image after affine transformation to obtain data with enhanced elastic deformation. Wherein in MNIST experiments, the value yielding the best result is
。
The ultrasonic image is changed in the two modes, so that a training sample is enhanced, and the accuracy of model calculation is improved.
And the clinical information preprocessing unit is used for eliminating incomplete clinical information and extracting main reference characteristics in the clinical information. Specifically, the clinical information is screened piece by piece, and incomplete clinical information is removed, for example: the clinical information includes patient basic information (age, sex, etc.), ultrasonic examination, serum examination index, etc., and if any of the clinical information is incomplete, the clinical information is deleted. Meanwhile, main reference characteristics required for identification, such as basic information of a patient, ultrasonic examination, serum examination indexes and the like, are extracted and stored in corresponding characteristics.
After preprocessing the ultrasonic image by the scaling preprocessing unit and the data amplification preprocessing unit, analyzing and calculating the preprocessed ultrasonic image by a pre-constructed ultrasonic image feature extractor. The ultrasonic image feature extractor is constructed by a deep learning classification model, such as LeNet, alexNet. Since the definition of a tumor feature by man is typically a low-level feature such as edges, shapes, textures, corner points, etc. And the convolutional neural network obtained by combining deep learning with a large data set and training can perform higher-level feature extraction on each tumor lesion region.
The identification of benign and malignant lesions of liver tumors is a typical supervised learning problem, a classifier with good generalization performance is trained through training data and label information, namely, characteristic information of liver tumors is independently learned in a data-driven mode, and final classification is carried out by combining final weight parameters obtained through continuous iterative updating according to a random gradient algorithm. The whole calculation process is as follows:
the first stage: the gray level image of the tumor segmentation image of the ultrasonic image with standard size is input, the stage only comprises one convolution layer conv1, the convolution layer with compensation of s=2 is used for downsampling, and the ultrasonic image with the size (112,112,32) is output.
And a second stage: this stage contains two convolutional layers conv2, conv3 and one max pooling layer pool1. In order to make the local receptive field sizes of neurons in conv2 and conv3 output characteristic diagrams on an input layer be the same and enhance the nonlinearity of each neuron activation value in the conv3 output characteristic on input image data, a convolution layer with a convolution kernel space size of 1*1 is used for conv 3. The maximum pooling layer pool1 reduces the size of the output feature map to (56,56,128).
And a third stage: this stage consists of three residual blocks, in turn called res1, res2, res3. In convolutional neural networks, residual delivery routes for standard residual blocks
By two successive complex operations->
The composition, where bn represents batch regularization, acti represents an activation function, conv represents an inactive convolution operation, and X represents an input signature. The parameters of the two residual blocks of res1 and res2 are the same, while the two convolution layer parameters of res3 are different. The size of the final output feature map is (28,28,256).
Fourth stage: this stage contains three residual blocks res1, res2, res3 and one convolutional layer conv4. The feature map is input to the convolutional layer conv4 for calculation after the effect of three residual blocks. The size of the final feature map is 14,14,128.
Fifth stage: this stage contains two residual blocks res1, res2 and one convolutional layer conv5. The final output feature size is (7,7,64).
Sixth stage: in order to avoid explosive increase of network parameters and improve the generalization performance of the network, the final output characteristic diagram in the fifth stage is taken as an input characteristic diagram, and the characteristic output diagram with the size of (1,1,64) is obtained through the action of a global average pooling layer (GAP). Wherein the feature output map comprises a tumor lesion area.
Seventh stage: this phase includes a fully connected layer with a number of neurons of 64 and an output layer with a number of neurons of 2.
It should be noted that, the feature output map with the size (1,1,64) obtained in the sixth stage corresponds to a feature value concentrated in each channel, and the feature output map obtained by feature extraction is mapped into one-dimensional feature vectors in a nonlinear manner by the full-connection layer to become a 64-dimensional feature vector, that is, a 64-dimensional feature vector is generated based on the features of the tumor lesions contained in the feature output map. The characteristics of tumor lesions include tumor size, morphology, number and vascular relationship, echo and silent halation.
Based on the feature vector, the softmax function is utilized to output the probability of two classifications, namely, the benign and malignant probability distribution of the tumor lesions
Wherein->
Benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying +.>
。
Stage 1-7 mainly introduces the structure of the classification convolutional neural network, directly learns the original tumor ultrasonic image, automatically extracts the characteristics and performs layer-by-layer abstraction, and finally realizes the classification task. The method is simply summarized as a 64-dimensional feature extraction and final two-classification process of the liver tumor ultrasonic image.
In addition, after the clinical information preprocessing unit preprocesses the clinical information, the clinical information is analyzed and calculated through a classifier constructed in advance. The classifier is constructed by adopting an XGBoost algorithm, the XGBoost algorithm is realized based on Boosting, the Boosting algorithm concept is to continuously improve and promote the weak classifier, and the classifiers are integrated together to form a strong classifier. In short, the XGBoost algorithm can be said to be an integrated lifting algorithm, which integrates many basic models together to form a strong model. The basic model can be a classifier regression decision tree CART or a linear model.
And taking the preprocessed clinical information as input of a classifier, and calculating benign probability values and malignant probability values of the tumor lesions according to each feature and the weight corresponding to each feature contained in the clinical information. It should be noted that, the clinical information includes different information, such as basic patient information, ultrasound examination, serum examination indexes, etc., and the different information is classified and stored as corresponding features, different results corresponding to each feature are provided with different weights, and the weights corresponding to the results of each feature are set in advance by staff. If the result corresponding to the serum examination index is x or y, the weight corresponding to x is A, and the weight corresponding to y is B (A > B).
Loading each feature of clinical information into each pre-trained sub-classifier, inputting the adjusted optimal values of the n_ estimators, max _depth and the learning_rate parameters, adding the scores obtained by each sub-classifier to obtain a final predicted value, and finally obtaining the benign and malignant probability distribution of the tumor lesion through logistic regression
Wherein->
Benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying
。
After obtaining the benign and malignant probability value of the tumor lesion of the patient, integrating all information through an image information visualization module 5 for visual display. The method comprises the steps of rendering a liver ultrasonic image, clinical information, and a benign probability value and a malignant probability value of a corresponding liver tumor lesion into the image for visual display.
If the result of the ultrasound image auxiliary discrimination module 3 and the result of the clinical information auxiliary discrimination module 4 are the same, the case is discriminated as a malignant case or a benign case, and if the discrimination is not the same, the case is discriminated as a suspected case. The suspected cases need to be given to the doctor for confirmation, and the final identification result is made. Because the clinical information and the ultrasonic image are different in the model and the characteristics of the two-time identification, the uncertainty of classification can be reduced as much as possible through the two-time identification, and the identification reliability is improved.
The foregoing are all preferred embodiments of the present application, and are not intended to limit the scope of the present application in any way, therefore: all equivalent changes in structure, shape and principle of this application should be covered in the protection scope of this application.
Claims (6)
1. An intelligent assistance system for classification of benign and malignant liver tumors, comprising:
the data transmission module is used for acquiring clinical information of an inspector; wherein, the clinical information comprises basic information, serum test indexes and clinical electronic medical record data;
the clinical information auxiliary identification module is used for taking clinical information as input of a pre-constructed classifier, and calculating benign probability values and malignant probability values of tumor lesions according to each feature and weight corresponding to each feature contained in the clinical information;
the clinical information auxiliary authentication module includes:
loading each feature of the clinical information into each pre-trained sub-classifier, inputting the weight corresponding to each feature, and adding the scores obtained by each sub-classifier to obtain a final predicted value;
processing the predicted value by logistic regression to obtain benign and malignant probability distribution of the tumor lesion
The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>
Benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying
。
2. The intelligent assistance system for classification of benign and malignant liver tumors of claim 1, wherein said data transmission module is further adapted to acquire an ultrasound image of the liver of an inspector; further comprises:
and the ultrasonic image auxiliary discrimination module is used for taking the liver ultrasonic image as the input of a pre-constructed ultrasonic image feature extractor and outputting benign probability values and malignant probability values of liver tumor lesions.
3. An intelligent assistance system for classification of benign and malignant liver tumors as claimed in claim 2, wherein said ultrasound image assisted identification module comprises:
inputting a gray level image of a liver ultrasonic image with a standard size, and processing the gray level image of the liver ultrasonic image to obtain a characteristic output image of the liver ultrasonic image; wherein the characteristic output graph comprises tumor size, morphology, echo and presence or absence of halation;
the feature output graph is mapped into a one-dimensional feature vector in a non-linear mode and becomes a 64-dimensional feature vector;
based on the feature vector, outputting benign and malignant probability distribution of tumor lesions by using softmax function
Wherein, the method comprises the steps of, wherein,
benign probability value representing the tumor lesion, +.>
A malignant probability value representing the tumor lesion and satisfying +.>
。
4. An intelligent assistance system for classification of benign and malignant liver tumors as claimed in claim 2, further comprising:
and the data preprocessing module is used for screening the clinical information and performing size scaling and data amplification on the liver ultrasonic image.
5. An intelligent assistance system for classification of benign and malignant liver tumors as claimed in claim 4, wherein: the data preprocessing module comprises:
the scaling pretreatment unit is used for scaling the liver ultrasonic image to ensure that the length of the liver ultrasonic image is standard;
the data amplification preprocessing unit is used for changing the positions of the pixel points in the liver ultrasonic image while not changing the relative positions of human tissues in the liver ultrasonic image;
and the clinical information preprocessing unit is used for eliminating incomplete clinical information and extracting main reference characteristics in the clinical information.
6. An intelligent assistance system for classification of benign and malignant liver tumors as claimed in claim 2, further comprising:
and the image information visualization module is used for integrating the liver ultrasonic image, the clinical information and the benign probability value and the malignant probability value of the corresponding liver tumor lesion to perform visual display.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310308198.9A CN116030971A (en) | 2023-03-28 | 2023-03-28 | Intelligent auxiliary system for classifying benign and malignant liver tumors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310308198.9A CN116030971A (en) | 2023-03-28 | 2023-03-28 | Intelligent auxiliary system for classifying benign and malignant liver tumors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116030971A true CN116030971A (en) | 2023-04-28 |
Family
ID=86079839
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310308198.9A Pending CN116030971A (en) | 2023-03-28 | 2023-03-28 | Intelligent auxiliary system for classifying benign and malignant liver tumors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116030971A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113077434A (en) * | 2021-03-30 | 2021-07-06 | 零氪智慧医疗科技(天津)有限公司 | Method, device and storage medium for lung cancer identification based on multi-modal information |
| CN115394426A (en) * | 2022-07-26 | 2022-11-25 | 电子科技大学 | Juvenile IgA nephropathy prediction analysis system based on transfer learning |
-
2023
- 2023-03-28 CN CN202310308198.9A patent/CN116030971A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113077434A (en) * | 2021-03-30 | 2021-07-06 | 零氪智慧医疗科技(天津)有限公司 | Method, device and storage medium for lung cancer identification based on multi-modal information |
| CN115394426A (en) * | 2022-07-26 | 2022-11-25 | 电子科技大学 | Juvenile IgA nephropathy prediction analysis system based on transfer learning |
Non-Patent Citations (2)
| Title |
|---|
| 吴宇雳: "基于卷积网络的乳腺超声肿瘤识别与分割", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》, no. 6, pages 1 - 3 * |
| 张伟 著: "《智能风控：评分卡建模原理、方法与风控策略构建》", 机械工业出版社, pages: 125 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Saba | Automated lung nodule detection and classification based on multiple classifiers voting | |
| Beeravolu et al. | Preprocessing of breast cancer images to create datasets for deep-CNN | |
| Maharjan et al. | A novel enhanced softmax loss function for brain tumour detection using deep learning | |
| Soudani et al. | An image-based segmentation recommender using crowdsourcing and transfer learning for skin lesion extraction | |
| CN111161275B (en) | Method and device for segmenting target object in medical image and electronic equipment | |
| CN110310287B (en) | Automatic organ-at-risk delineation method, equipment and storage medium based on neural network | |
| Yousef et al. | A holistic overview of deep learning approach in medical imaging | |
| Yonekura et al. | Automatic disease stage classification of glioblastoma multiforme histopathological images using deep convolutional neural network | |
| Shen et al. | Mass image synthesis in mammogram with contextual information based on GANs | |
| Pourreza et al. | Brain tumor segmentation in MRI scans using deeply-supervised neural networks | |
| CN110110808B (en) | Method and device for performing target labeling on image and computer recording medium | |
| CN111640120A (en) | Pancreas CT automatic segmentation method based on significance dense connection expansion convolution network | |
| Shanthi et al. | Automated detection and classification of cervical cancer using pap smear microscopic images: a comprehensive review and future perspectives | |
| Teramoto et al. | Computer-aided classification of hepatocellular ballooning in liver biopsies from patients with NASH using persistent homology | |
| Saha Roy et al. | An automated liver tumour segmentation and classification model by deep learning based approaches | |
| Jena et al. | Morphological feature extraction and KNG‐CNN classification of CT images for early lung cancer detection | |
| Prusty et al. | A novel transfer learning technique for detecting breast cancer mammograms using VGG16 bottleneck feature | |
| Tasdemir et al. | ROI detection in mammogram images using wavelet-based Haralick and HOG features | |
| Liu et al. | Extracting lungs from CT images via deep convolutional neural network based segmentation and two-pass contour refinement | |
| Tsai et al. | Machine learning based common radiologist-level pneumonia detection on chest X-rays | |
| Kumar et al. | Skin cancer segmentation with the aid of multi-class dilated D-net (MD2N) framework | |
| Routray et al. | Ensemble Learning with Symbiotic Organism Search Optimization Algorithm for Breast Cancer Classification & Risk Identification of Other Organs on Histopathological Images | |
| Öztürk et al. | A novel polyp segmentation approach using U-net with saliency-like feature fusion | |
| CN116030971A (en) | Intelligent auxiliary system for classifying benign and malignant liver tumors | |
| Sharma et al. | Solving image processing critical problems using machine learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230428 |