CN113571193B - Construction method and device of lymph node metastasis prediction model based on multi-view learning image histology fusion - Google Patents
Construction method and device of lymph node metastasis prediction model based on multi-view learning image histology fusion Download PDFInfo
- Publication number
- CN113571193B CN113571193B CN202110702168.7A CN202110702168A CN113571193B CN 113571193 B CN113571193 B CN 113571193B CN 202110702168 A CN202110702168 A CN 202110702168A CN 113571193 B CN113571193 B CN 113571193B
- Authority
- CN
- China
- Prior art keywords
- histology
- lymph node
- image
- feature
- tumor
- 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.)
- Active
Links
- 208000007433 Lymphatic Metastasis Diseases 0.000 title claims abstract description 55
- 230000004927 fusion Effects 0.000 title claims abstract description 26
- 238000010276 construction Methods 0.000 title claims description 5
- 210000001165 lymph node Anatomy 0.000 claims abstract description 87
- 206010028980 Neoplasm Diseases 0.000 claims abstract description 79
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000012360 testing method Methods 0.000 claims abstract description 15
- 238000010219 correlation analysis Methods 0.000 claims abstract description 9
- 238000012216 screening Methods 0.000 claims abstract description 8
- 238000007477 logistic regression Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 206010027476 Metastases Diseases 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000009401 metastasis Effects 0.000 claims description 3
- 238000011002 quantification Methods 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 abstract description 2
- 238000003384 imaging method Methods 0.000 description 12
- 201000011510 cancer Diseases 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 238000007500 overflow downdraw method Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 238000010187 selection method Methods 0.000 description 3
- 208000005718 Stomach Neoplasms Diseases 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 238000013079 data visualisation Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 206010017758 gastric cancer Diseases 0.000 description 2
- 201000011549 stomach cancer Diseases 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 208000035965 Postoperative Complications Diseases 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000000090 biomarker Substances 0.000 description 1
- 238000013170 computed tomography imaging Methods 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000007170 pathology Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 210000002966 serum Anatomy 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method and a device for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion, wherein the method comprises the following steps: acquiring image data and preprocessing to obtain tumor image histology characteristics and lymph node image histology characteristics; performing correlation analysis and feature screening based on a supervised feature selection strategy of a test time budget sequentially aiming at tumor image histology features and lymph node image histology features to obtain tumor image histology feature samples and lymph node image histology feature samples; performing unsupervised learning from multi-view subspaces to public spaces on the two samples by adopting an unsupervised multi-view partial least square method so as to map lymph node image histology features and tumor image histology features from subspaces of respective views to the public spaces; the lymph node image histology characteristics and the tumor image histology characteristics in the public space are taken as input, and the label of whether the lymph node is metastasized is taken as output, so that a logistic regression classifier is trained to obtain a lymph node metastasis prediction model.
Description
Technical Field
The invention belongs to the technical field of multi-view learning, and particularly relates to a method and a device for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion.
Background
For cancer patients, lymph node metastasis determines the extent of their lymph node cleansing and is also one of the major independent prognostic factors. The method can accurately predict the lymph node state of a cancer patient before operation, and has important significance for avoiding excessive treatment and reducing postoperative complications. Several studies have shown that preoperative CT imaging helps to achieve individualized predictions of lymph node status in cancer patients, but these studies often utilize tumor imaging histology features or incorporate small numbers of clinical pathology features (e.g., lymph node status in CT reports, serum biomarkers, TNM staging, etc.). In addition to these features, previous studies have found that lymph node imaging histology has predictive ability to distinguish between lymph node metastasis in cancer patients. By fusing the primary tumor and lymph node imaging histology features, the lymph node status of the patient can be better predicted. These two image histology features can be considered as two views, which can complement each other, and redundancy may also exist. Simply combining the two together does not fully describe the lymph node status information of a cancer patient, thereby limiting the ability to accurately predict lymph node metastasis prior to surgery.
The multi-view learning technology based on artificial intelligence, which utilizes data collected from multiple views to overcome the limitation of single-view analysis, is widely focused in recent years, and is increasingly applied to the fields of medical image processing and analysis, such as a multi-view learning and depth supervision self-encoder-based medical image classification method and device disclosed in patent application publication No. CN112488102A, and a multi-mode parametric model optimization fusion method based on image histology features disclosed in patent application publication No. CN 111462116A.
However, studies of multi-view learning image histology fusion methods for cancer patients are not clear. There is a need for a multi-view learning fusion method that is effective for tumor imaging histology and lymph node imaging histology for the accurate pre-operative prediction of lymph node metastasis in cancer patients.
Disclosure of Invention
In view of the above, the present invention aims to provide a method and an apparatus for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion, which map multi-view features to a common space through the acquisition of tumor image histology samples and lymph node image histology samples and the design of a multi-view learning fusion algorithm, and construct the lymph node metastasis prediction model by using the features of the common space so as to improve the prediction performance of the lymph node metastasis prediction model.
In a first aspect, an embodiment provides a method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion, including the following steps:
acquiring image data, carrying out region of interest sketching, region of interest difference value and image histology feature extraction on the image data, and obtaining tumor image histology features and lymph node image histology features;
performing correlation analysis and feature screening based on a supervised feature selection strategy of a test time budget sequentially aiming at the tumor image histology features to obtain the tumor image histology features with the least number and discrimination capability as tumor image histology feature samples;
performing correlation analysis and feature screening based on a supervised feature selection strategy of test time budget sequentially on the lymph node image histology features to obtain lymph node image histology features with minimum quantity and discrimination capability as lymph node image histology feature samples;
performing unsupervised study from multi-view subspaces to public spaces on the lymph node image histology feature samples and the tumor image histology feature samples by adopting an unsupervised multi-view partial least square method so as to map the lymph node image histology features and the tumor image histology features from subspaces of respective views to the public spaces;
the lymph node image histology characteristics and the tumor image histology characteristics in the public space are taken as input, and the label of whether the lymph node is metastasized is taken as output, so that a logistic regression classifier is trained to obtain a lymph node metastasis prediction model.
In a second aspect, an embodiment provides a lymph node metastasis prediction model based on multi-view learning image histology fusion, where the lymph node metastasis prediction model is constructed by the construction method in the first aspect.
In a third aspect, an embodiment provides a device for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion according to the first aspect when executing the computer program.
The method and the device for constructing the lymph node metastasis prediction model based on the multi-view learning image histology fusion have the advantages that at least:
the multi-view learning technology is used for fusing primary tumor image histology and lymph node image histology, the multi-step feature selection method has prediction capability for distinguishing patients with or without lymph node metastasis, and the supervised feature selection algorithm based on the test time budget is a supervised feature selection method with global targets, and different numbers of features are selected by using proper budgets to meet the requirement of image histology samples as prediction factors. The unsupervised multi-view partial least square method can maximize the cross covariance, and achieve a better data visualization effect while learning the orthogonal projection matrix to train the prediction model.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion according to an embodiment;
FIG. 2 is a representation of tumor image histology and lymph node image histology feature numbers retained after a Pearson's correlation coefficient analysis, as provided by an embodiment;
FIG. 3 is a representation of tumor image histology tag and lymph node image histology tag feature numbers after multi-step feature selection, as provided by one embodiment;
FIG. 4 is a schematic view showing the visual effect of a lymph node metastasis prediction model according to an embodiment;
FIG. 5 is a graph of performance of a lymph node metastasis prediction model in a training set and a validation set, as provided by an embodiment.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the detailed description is presented by way of example only and is not intended to limit the scope of the invention.
Lymph node metastasis is accompanied by the development of primary tumors and is a complex, continuous process. The primary tumor image histology is only relied on, and the lymph node image histology is not considered, so that the lymph node metastasis patients and the lymph node non-metastasis patients cannot be comprehensively distinguished. Based on the above, the embodiment of the invention provides a lymph node metastasis prediction model which comprehensively considers the characteristics of lymph node imaging histology and tumor imaging histology to realize the prediction of lymph node metastasis.
Fig. 1 is a flowchart of a method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion according to an embodiment. As shown in fig. 1, the method for constructing a lymph node metastasis prediction model provided in the embodiment includes the following steps:
and step 1, acquiring image data and dividing the image data into a training set and a verification set.
In an embodiment, a nanoarray standard of patient data is formulated, and image data and clinical data of the patient are collected retrospectively. And carrying out random experiments on the included image data and clinical data, dividing a training set and a verification set, and ensuring that the proportion of positive and negative samples of the training set and the verification set is kept consistent.
And 2, preprocessing the image data, including region of interest sketching, region of interest difference value and image histology feature extraction.
In an embodiment, when the image data is delineated into a region of interest, multiple imaging physicians manually delineate the outline of the primary tumor and the outline of the lymph node of each patient on the image as a tumor region of interest and a lymph node region of interest, respectively. If the physician's opinion is inconsistent, another physician participates in the discussion until a group consensus is reached. After obtaining the tumor region of interest and the lymph node region of interest, interpolation processing of the region of interest data is further carried out, and then, an internal development Python open source data packet is adopted to respectively conduct feature extraction on the tumor region of interest and the lymph node region of interest so as to obtain tumor image histology features and lymph node image histology features.
And 3, performing feature selection on the tumor image histology features and the lymph node image histology features extracted in the step 2, wherein the feature selection comprises correlation analysis and feature selection based on a supervised feature selection strategy of a test time budget.
In the embodiment, the pearson correlation coefficient analysis is adopted to respectively perform correlation analysis on the tumor image histology characteristics and the lymph node image histology characteristics so as to screen and remove redundant characteristics and reduce characteristic dimension. The specific process is as follows: for the tumor image histology characteristics, the pearson correlation coefficient of the tumor image histology characteristic pair is calculated, and the larger the pearson correlation coefficient is, the higher the correlation of the tumor image histology characteristic pair is, the characteristics with the pearson correlation coefficient larger than a certain threshold are regarded as redundant characteristics and are eliminated. The same operation is performed on the lymph node image histology characteristics, namely, the Pearson correlation coefficient of the lymph node image histology characteristic pair is calculated, the larger the Pearson correlation coefficient is, the higher the correlation of the lymph node image histology characteristic pair is, the characteristics with the Pearson correlation coefficient larger than a certain threshold are regarded as redundant characteristics, and the characteristics are eliminated. The tumor image histology and lymph node image histology feature numbers remaining after pearson correlation coefficient analysis are shown in fig. 2.
And removing redundant features according to correlation analysis, and performing feature screening on the remaining tumor image histology features and lymph node image histology features by adopting a supervised feature selection strategy based on test time budget so as to obtain tumor image histology features and lymph node image histology features which are the least in number and have identification capability as tumor image histology feature samples and lymph node image histology feature samples respectively.
The linear predictor is learned by selecting feature groups with explicit budget constraints by introducing binary index variables based on a supervised feature selection strategy for testing the time budget, thereby increasing the total cost if the cost of each group is available. The supervised feature selection policy based on the test time budget is described as:
wherein ,representing a training set of n patients and d features x i Is the feature vector of the ith patient, with the ith feature characterized as x i,s Wherein the feature vector here comprises tumor image histology feature vector, lymph node image histology feature vector, y i Values 1 and 0 represent lymph node metastasis and lymph node non-metastasis, respectively, θ is an index vector, values 0 or 1 represent unselected or selected, respectively, w and b are coefficients and bias of the linear predictor, f i Is a linear predictor pair x i L is defined at f i and yi The loss function on each feature vector is regarded as a group of cost 1, i.e. c s =1, the total budget B is the expected number of features to be selected;
in the training set, repeated random experiments are carried out on a plurality of preset super parameters C, the optimal super parameters C are searched, the expected feature number B is further determined, and then the tumor image histology features and the lymph node image histology features which are obtained through screening and have the minimum number and the identification capability are respectively used as tumor image histology feature samples and lymph node image histology feature samples. The tumor image histology tag and lymph node image histology tag feature numbers after the multi-step feature selection are shown in fig. 3. In an embodiment, the preset value of the hyper-parameter C includes 0.01,0.1,1,10,100.
And 4, performing spatial mapping by using feature samples of feature screening by using an unsupervised multi-view partial least square method and establishing a lymph node metastasis prediction model.
Tumor imaging and lymph node imaging are two different views of the lymph node classification prediction problem, which can complement each other and can also have redundancy. Simply combining the two together does not fully describe lymph node metastasis information, thereby limiting the ability to accurately predict the lymph node status of gastric cancer. The multi-view learning technology overcomes the limitation of single-view analysis by utilizing the data collected from different views, provides an effective noninvasive auxiliary diagnosis strategy for imaging doctors, and has potential to be applied to other clinical tasks and expanded to different patients.
In an embodiment, an unsupervised multi-view partial least squares method is used to perform multi-view subspace-to-common-space unsupervised learning on the lymph node image histology feature samples and the tumor image histology feature samples to map the lymph node image histology features and the tumor image histology features from the subspaces of the respective views to the common space.
The unsupervised multi-view partial least squares method proposed by the embodiments learns a function for modeling tumor image histology samples and lymph node image histology samples. On the premise of not lacking generality, X is set LN X for lymph node view with lymph node image histology tag TU For a tumor view with a tumor image histology signature, each column represents one patient of the training set, described as follows using the unsupervised multi-view partial least squares method:
wherein ,XLN and XTU N columns of (c) have an average value of zero, P LN ,P TU Representing projection matrices from lymph node feature numbers and tumor feature numbers to potential common space dimension k, respectively, I k ∈R k×k Is a unit matrix, and an optimal projection matrix P is obtained by solving LN and PTU 。
The unsupervised multi-view partial least squares method can guarantee orthogonality constraint, and meanwhile, the covariance in the public space is maximized by means of a mature numerical linear algebra technology. The existing method often encounters the situation of unstable values, and the orthogonality of the view-specific projection matrix cannot be ensured. Orthogonal projection not only has good metric retention characteristics, but also provides a natural representation for data visualization similar to principal component analysis.
In an embodiment, the projection matrix P is optimized LN and PTU View X of lymph node LN And tumor view X TU Mapping from subspaces of respective views to a common space:
wherein ,zLN 、z TU The lymph node image histology and the tumor image histology in the public space are respectively represented.
Although the two samples are located in different feature spaces, the two projection points are located in the same space.
When the lymph node metastasis prediction model is constructed, the lymph node image histology characteristics and the tumor image histology characteristics in the public space are taken as input, the label of whether the lymph node metastasizes is taken as output, and a logistic regression classifier is trained to obtain the lymph node metastasis prediction model. As for classification performance, the fusion of two projection points (i.e., z=, z LN ；z TU (-) generally exhibits better performance, i.e. a mosaic matrix of lymph node image histology features and tumor image histology features is input into the public space, and whether lymph nodes metastasize or not is predicted according to the mosaic matrix, and the visualization effect is shown in fig. 4.
When visualizing features projected to a common space, an average of two projection points is calculated (i.e., z= (z) LN +z TU ) 2) to enable visualization of the feature.
And 5, evaluating the performance of the final prediction model in the verification set.
In the embodiment, the lymph node metastasis prediction model is verified by utilizing the lymph node image histology characteristic sample and the tumor image histology characteristic sample in the verification set, and five performance indexes of the area under the working characteristic curve of the test subject, the accuracy, the recall rate and the F1 value are adopted for quantification during verification. The performance results of the training set and the validation set are shown in fig. 5.
The lymph node metastasis prediction model which does not meet the requirements is trained again until the requirements are met.
In the technical scheme, the performance of the lymph node metastasis prediction model in the verification set (the area under the working characteristic curve of the test subject: 0.8660) is superior to that of the prediction model using only tumor image histology (the area under the working characteristic curve of the test subject: 0.7582) and the prediction model using only lymph node image histology (the area under the working characteristic curve of the test subject: 0.8431), so that the multi-view fusion method can ensure better quantitative indexes than the single-view method.
According to the construction method of the lymph node metastasis prediction model based on multi-view learning image histology fusion, the two-step feature selection method is adopted to obtain primary tumor image histology and lymph node image histology prediction factors, then the multi-view learning technology is utilized to fuse tumor and lymph node image information, the limitation of single-view analysis is overcome, a noninvasive gastric cancer lymph node metastasis prediction auxiliary diagnosis strategy is provided for imaging doctors, and the method has potential to be applied to solutions of other clinical tasks and meets clinical requirements of different diseases.
The embodiment also provides a device for constructing the lymph node metastasis prediction model based on the multi-view learning image histology fusion, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the method for constructing the lymph node metastasis prediction model based on the multi-view learning image histology fusion is realized when the processor executes the computer program.
In practical applications, the computer memory may be a volatile memory at the near end, such as a RAM, or a nonvolatile memory, such as a ROM, a FLASH, a floppy disk, a mechanical hard disk, or a remote storage cloud. The computer processor may be a Central Processing Unit (CPU), a Microprocessor (MPU), a Digital Signal Processor (DSP), or a Field Programmable Gate Array (FPGA), that is, the steps of the method for constructing the lymph node metastasis prediction model based on multi-view learning image histology fusion may be implemented by these processors.
The foregoing detailed description of the preferred embodiments and advantages of the invention will be appreciated that the foregoing description is merely illustrative of the presently preferred embodiments of the invention, and that no changes, additions, substitutions and equivalents of those embodiments are intended to be included within the scope of the invention.
Claims (6)
1. The method for constructing the lymph node metastasis prediction model based on multi-view learning image histology fusion is characterized by comprising the following steps of:
acquiring image data, carrying out region of interest sketching, region of interest difference value and image histology feature extraction on the image data, and obtaining tumor image histology features and lymph node image histology features;
performing correlation analysis and feature screening based on a supervised feature selection strategy of a test time budget sequentially aiming at the tumor image histology features to obtain the tumor image histology features with the least number and discrimination capability as tumor image histology feature samples;
performing correlation analysis and feature screening based on a supervised feature selection strategy of test time budget sequentially on the lymph node image histology features to obtain lymph node image histology features with minimum quantity and discrimination capability as lymph node image histology feature samples;
performing unsupervised study from multi-view subspaces to public spaces on the lymph node image histology feature samples and the tumor image histology feature samples by adopting an unsupervised multi-view partial least square method so as to map the lymph node image histology features and the tumor image histology features from subspaces of respective views to the public spaces;
training a logistic regression classifier by taking the lymph node image histology characteristics and the tumor image histology characteristics of the public space as input and taking the label of whether the lymph node is metastasized as output so as to obtain a lymph node metastasis prediction model;
wherein the supervised feature selection policy based on the test time budget is described as:
wherein ,representing a training set of n patients and d features x i Is the feature vector of the ith patient, with the ith feature characterized as x i,s Wherein the feature vector here comprises tumor image histology feature vector, lymph node image histology feature vector, y i Values 1 and 0 represent lymph node metastasis and lymph node non-metastasis, respectively, θ is an index vector, values 0 or 1 represent unselected or selected, respectively, w and b are coefficients and bias of the linear predictor, f i Is a linear predictor pair x i L is defined at f i and yi The loss function on each feature vector is regarded as a group of cost 1, i.e. c s =1, the total budget B is the expected number of features to be selected;
in a training set, carrying out repeated random experiments on a plurality of preset super parameters C, searching for the optimal super parameters C, further determining an expected feature number B, and further determining that the screened tumor image histology features and lymph node image histology features which are the least in number and have identification capability are respectively used as tumor image histology feature samples and lymph node image histology feature samples;
the use of the unsupervised multi-view partial least squares method is described as:
wherein ,XLN 、X TU A lymph node view and a tumor view of a tumor image histology label respectively representing a lymph node image histology sample, X LN and XTU N columns of (c) have an average value of zero, P LN ,P TU Representing projection matrices from lymph node feature numbers and tumor feature numbers to potential common space dimension k, respectively, I k ∈R k×k Is a unit matrix, and an optimal projection matrix P is obtained by solving LN and PTU ；
Then, by the optimal projection matrix P LN and PTU View X of lymph node LN And tumor view X TU Mapping from subspaces of respective views to a common space:
wherein ,zLN 、z TU The lymph node image histology and the tumor image histology in the public space are respectively represented.
2. The method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion according to claim 1, wherein pearson correlation coefficient analysis is adopted to perform correlation analysis on tumor image histology features and lymph node image histology features respectively so as to screen and reject redundant features and reduce feature dimensions.
3. The method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion according to claim 1, wherein the preset super-parameter C value comprises 0.01,0.1,1,10,100.
4. The method for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion according to claim 1, further comprising:
the lymph node metastasis prediction model is verified by utilizing the lymph node image histology characteristic sample and the tumor image histology characteristic sample which are concentrated in verification, and five performance indexes of area, accuracy, recall rate and F1 value under the working characteristic curve of the subject are adopted for quantification during verification;
the lymph node metastasis prediction model which does not meet the requirements is trained again until the requirements are met.
5. A lymph node metastasis prediction model based on multi-view learning image histology fusion, characterized in that the lymph node metastasis prediction model is constructed by the construction method of the lymph node metastasis prediction model based on multi-view learning image histology fusion according to any one of claims 1 to 4.
6. A device for constructing a lymph node metastasis prediction model based on multi-view learning image histology fusion, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor realizes the method for constructing the lymph node metastasis prediction model based on multi-view learning image histology fusion according to any one of claims 1 to 4 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110702168.7A CN113571193B (en) | 2021-06-24 | 2021-06-24 | Construction method and device of lymph node metastasis prediction model based on multi-view learning image histology fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110702168.7A CN113571193B (en) | 2021-06-24 | 2021-06-24 | Construction method and device of lymph node metastasis prediction model based on multi-view learning image histology fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113571193A CN113571193A (en) | 2021-10-29 |
| CN113571193B true CN113571193B (en) | 2023-09-05 |
Family
ID=78162597
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110702168.7A Active CN113571193B (en) | 2021-06-24 | 2021-06-24 | Construction method and device of lymph node metastasis prediction model based on multi-view learning image histology fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113571193B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114429787B (en) * | 2021-12-30 | 2023-04-18 | 北京百度网讯科技有限公司 | Omics data processing method and device, electronic device and storage medium |
| CN116805536A (en) * | 2023-08-22 | 2023-09-26 | 乐陵市人民医院 | Data processing method and system based on tumor case follow-up |
| CN117649408A (en) * | 2024-01-29 | 2024-03-05 | 天津博思特医疗科技有限责任公司 | Lung nodule recognition processing method based on lung CT image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305685A (en) * | 2017-12-25 | 2018-07-20 | 王莹豪 | Luminal Type B HER-2 negative breast cancer prognostic predictive models KLP-PI |
| CN110111895A (en) * | 2019-05-15 | 2019-08-09 | 天津医科大学肿瘤医院 | A kind of method for building up of nasopharyngeal carcinoma far-end transfer prediction model |
| CN110751215A (en) * | 2019-10-21 | 2020-02-04 | 腾讯科技（深圳）有限公司 | Image identification method, device, equipment, system and medium |
| CN112216395A (en) * | 2020-09-11 | 2021-01-12 | 中山大学孙逸仙纪念医院 | Axillary lymph node metastasis prediction model for breast cancer patient and construction method thereof |
-
2021
- 2021-06-24 CN CN202110702168.7A patent/CN113571193B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305685A (en) * | 2017-12-25 | 2018-07-20 | 王莹豪 | Luminal Type B HER-2 negative breast cancer prognostic predictive models KLP-PI |
| CN110111895A (en) * | 2019-05-15 | 2019-08-09 | 天津医科大学肿瘤医院 | A kind of method for building up of nasopharyngeal carcinoma far-end transfer prediction model |
| CN110751215A (en) * | 2019-10-21 | 2020-02-04 | 腾讯科技（深圳）有限公司 | Image identification method, device, equipment, system and medium |
| CN112216395A (en) * | 2020-09-11 | 2021-01-12 | 中山大学孙逸仙纪念医院 | Axillary lymph node metastasis prediction model for breast cancer patient and construction method thereof |
Non-Patent Citations (1)
| Title |
|---|
| Preoperative radiomic signature based on multiparametric magnetic resonance imaging for noninvasive evaluation of biological characteristics in rectal cancer;MENG, XIAOCHUN, XIA, WEI, XIE, PEIYI, et al.;《European radiolog》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113571193A (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Park et al. | Comparison of machine and deep learning for the classification of cervical cancer based on cervicography images | |
| Fuhrman et al. | A review of explainable and interpretable AI with applications in COVID‐19 imaging | |
| CN113571193B (en) | Construction method and device of lymph node metastasis prediction model based on multi-view learning image histology fusion | |
| Alghamdi et al. | Deep learning approaches for detecting COVID-19 from chest X-ray images: A survey | |
| Graziani et al. | Concept attribution: Explaining CNN decisions to physicians | |
| US20220237788A1 (en) | Multiple instance learner for tissue image classification | |
| Shamrat et al. | High-precision multiclass classification of lung disease through customized MobileNetV2 from chest X-ray images | |
| Pellicer-Valero et al. | Deep learning for fully automatic detection, segmentation, and Gleason grade estimation of prostate cancer in multiparametric magnetic resonance images | |
| Vyas et al. | Generative Adversarial Networks for Anomaly Detection in Medical Images | |
| US20220058839A1 (en) | Translation of images of stained biological material | |
| Oliveira et al. | CAD systems for colorectal cancer from WSI are still not ready for clinical acceptance | |
| CN111814868A (en) | Model based on image omics feature selection, construction method and application | |
| Ahmad et al. | Optimized lung nodule prediction model for lung cancer using contour features extraction | |
| Tian et al. | Radiomics and Its Clinical Application: Artificial Intelligence and Medical Big Data | |
| Mei et al. | Exploiting partial common information microstructure for multi-modal brain tumor segmentation | |
| Wang et al. | SAC-Net: Enhancing Spatiotemporal Aggregation in Cervical Histological Image Classification via Label-Efficient Weakly Supervised Learning | |
| Kumar et al. | E-fuzzy feature fusion and thresholding for morphology segmentation of brain MRI modalities | |
| Sampath et al. | 3D brain image‐based Alzheimer's disease detection techniques using fish swarm optimizer's deep convolution Siamese neural network | |
| Vu et al. | An improved mammography malignancy model with self-supervised learning | |
| Park et al. | Classification of cervical cancer using deep learning and machine learning approach | |
| Padmavathi et al. | Effective deep learning based segmentation and classification in wireless capsule endoscopy images | |
| Jun et al. | Medical data science in rhinology: Background and implications for clinicians | |
| Liu et al. | Multi‐scale discriminative network for prostate cancer lesion segmentation in multiparametric MR images | |
| Guo et al. | Core dataset extraction from unlabeled medical big data for lesion localization | |
| Yang et al. | Leveraging auxiliary information from emr for weakly supervised pulmonary nodule detection |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |