Breast cancer axillary lymph node metastasis prediction method based on radiomics and domain adaptation

By introducing a domain adaptation algorithm into the prediction model of axillary lymph node metastasis in breast cancer, the problem of poor prediction performance of multi-center data sets is solved, and higher prediction accuracy and adaptability are achieved, and the reliability and generalization ability of the model are improved. , provides more accurate technical support for the diagnosis and treatment of breast cancer.

CN118941777AActive Publication Date: 2024-11-12NANTONG UNIV

Patent Information

Application Number
CN202411056341.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-11-12
Estimated Expiration
2044-08-02

AI Technical Summary

Technical Problem

The existing prediction model of axillary lymph node metastasis in breast cancer has poor prediction performance in multi-center data sets and cannot effectively adapt to the data distribution differences between different hospitals and equipment, resulting in low model reliability and generalization ability.

Method used

Using an imagingomics and domain adaptation method, the image features are distributed and aligned through the balanced distribution adaptation algorithm (BDA), and the domain adaptation-imageomics model is constructed to reduce the difference in the feature distribution between the source domain and the target domain, and Iterative refinement is performed on the training set to improve the prediction performance and generalization capabilities of the model.

Benefits of technology

It significantly improves the prediction performance and generalization ability of the model in multi-center data sets, can better adapt to the data distribution of different hospitals and equipment, improves the accuracy and reliability of the prediction of axillary lymph node metastasis in breast cancer, and provides technology for precision medicine. support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941777A_ABST
    Figure CN118941777A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of pathological digital image processing, in particular to a breast cancer axillary lymph node metastasis prediction method based on radiomics and domain adaptation. The method comprises the specific steps of S1, selecting and dividing multi-center data; s2, preprocessing the data in the S1; s3, performing feature extraction on the data preprocessed in the S2; s4, screening the features extracted in the step S3; s5, constructing a domain adaptation-radiomics model and verifying the domain adaptation-radiomics model; according to the method, a domain adaptation algorithm is introduced, the feature distribution difference between a source domain and a target domain is reduced, a real clinical practice environment is better reflected by incorporating multi-center image data, multi-sequence preoperative mammary gland magnetic resonance images are incorporated, and available information of radiomics is enriched; the breast cancer axillary lymph node metastasis prediction method provides guarantee for improving the performance and generalization of breast cancer axillary lymph node metastasis prediction by radiomics, and provides technical support for precise medical service.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of pathological digital image processing, and in particular to a method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation. Background Art

[0002] In the diagnosis and treatment of breast cancer, axillary lymph node metastasis (ALNM) is highly correlated with distant metastasis and recurrence of breast cancer and survival rate. Accurate detection and effective evaluation of ALNM are of great significance to the patient's prognosis and treatment plan selection. Radiomics can reveal valuable insights and quantitative information in imaging data that may not be recognized by the naked eye, helping medical professionals predict the status of axillary lymph nodes in breast cancer patients. Among the existing similar patents, CN113555115A discloses a method for constructing a lymph node metastasis prediction model for breast cancer patients based on radiomics. The method uses a random forest algorithm to screen image features, performs univariate analysis on clinical feature data to obtain key clinical features, and uses a support vector machine algorithm to establish a comprehensive prediction model for lymph node metastasis based on key image features and key clinical features. A prediction model based on the principle of structural risk minimization is established to help predict lymph nodes. CN116741390A also discloses a prediction model for axillary lymph node metastasis in breast cancer patients. After three-dimensional reconstruction of the two-dimensional lung enhanced CT film, the method selects more than five imaging features from all the imaging features of each axillary lymph node and uses them together to distinguish whether the breast cancer has axillary lymph node metastasis, and uses random forest to build a prediction model.

[0003] Although some patents have disclosed methods for using radiomics to intelligently predict the status of axillary lymph nodes in breast cancer, their technical solutions still have some shortcomings: 1. The radiomics model has not been verified using a multi-center data set, which cannot guarantee that the radiomics model has good predictive performance in unfamiliar centers with different distributions, and the reliability of the model is low; 2. The model that uses a multi-center data set to verify performance does not have good predictive capabilities in every center. When the distribution of the external validation set is very different from that of the training set, the model performs poorly.

[0004] Therefore, it is of great significance to develop a new radiomics- and domain-adaptation-based method for predicting axillary lymph node metastasis in breast cancer. Summary of the invention

[0005] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method that is different from the existing patents of the same type in view of the shortcomings of the prior art. From the perspective of adding domain adaptation algorithm to the traditional imaging omics process to improve the performance of imaging omics models in multi-center data sets, a method for predicting axillary lymph node metastasis of breast cancer based on imaging omics and domain adaptation is developed.

[0006] In order to solve the above technical problems, the present invention discloses a method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation, and the specific steps are as follows:

[0007] S1. Select multi-center data and divide them;

[0008] S2. Preprocess the data in S1;

[0009] S3. Extract features from the data preprocessed by S2;

[0010] S4. Filter the features extracted in S3;

[0011] S5. Build and validate the domain-adapted radiomics model.

[0012] The multi-center data are multi-parameter MRI images; the multi-parameter MRI images are 8 sequences of phases 0-5 of T2WI, nonfs-T1WI and DCE-MRI.

[0013] The specific steps of S2 are: segmenting the divided data and outlining the ROI; resampling and standardizing the outlined image;

[0014] Preferably, the resampled size is 1×1×1 mm;

[0015] Among them, in S3, the features in the feature extraction process are shape features, first-order features, texture features and high-order features obtained by performing various transformations on the original image; the texture features are grayscale co-occurrence matrix features, grayscale run matrix features, grayscale area size matrix features and grayscale dependence matrix features; the transformations used in the high-order features are square, square root, logarithm, exponential, gradient size, local binary pattern, wavelet filtering and Gaussian Laplace.

[0016] Among them, in S4, feature screening is performed through ANOVA, mRMR and LASSO regression.

[0017] The domain adaptation-radiology model is specifically a model constructed by adding the domain adaptation algorithm to the model development step of the traditional radiomics process (image acquisition-image segmentation-radiology feature extraction-feature selection-model development-model performance evaluation). First, the key features selected in the training set and the corresponding same features in the external validation set are input into the balanced distribution adaptation (BDA) domain adaptation algorithm. Then, the BDA algorithm achieves the alignment of the two domains by minimizing the difference between the feature distributions of the source domain (training set) and the target domain (external validation set). The objective formula of the algorithm is as follows:

[0018]

[0019] sA T XHX T A=I.0≤μ≤1

[0020] Among them, X represents x s and x t The input data matrix is ​​composed of A, A represents the transformation matrix, μ is the balance factor, M0 and M c is the MMD matrix, c∈{1,2,···,C} is the label category, λ is the regularization parameter, is the Frobenius norm. .I∈R (m+n)×(m+n) ' is the unit matrix, H is the center matrix, that is, H = I-(1 / n)1.

[0021] There are two terms in the formula: the adaptation of the marginal distribution and conditional distribution with a balancing factor (term 1) and the regularization term (term 2). And the formula has two constraints: the first constraint ensures that the data after the transformed distribution (A T X) should preserve the internal properties of the original data. The second constraint represents the range of the balancing factor μ.

[0022] The best way to adapt to the marginal distribution and conditional distribution of the two domains is to adjust the parameter μ in the above formula. If the difference in the marginal distribution of the two domains is considered more important, the value of μ is tended to 0. If the difference in the conditional distribution is considered more important, the value of μ is tended to 1. Then, the domain adaptation-radiology model is constructed on the training set using the classifier embedded in the BDA algorithm using the new distribution. Finally, the domain adaptation-radiology model predicts the label of the target domain, i.e., the external validation set, obtains the pseudo label, and iteratively refines it to finally generate the output label for performance verification.

[0023] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation of the present invention.

[0024] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for predicting axillary lymph node metastasis of breast cancer based on imaging genomics and domain adaptation of the present invention are implemented.

[0025] Beneficial effects: Compared with traditional imaging genomics models, the present invention introduces a domain adaptation algorithm, which can narrow the feature distribution differences between the source domain and the target domain, overcome the differences in magnetic resonance imaging data caused by different hospitals, different instruments and equipment, or different types of patients, thereby improving the model's prediction performance and generalization ability in multiple centers; the present invention incorporates imaging data from multiple centers, and different centers have different data distributions, which can better reflect the real clinical practice environment; the present invention incorporates a total of 8 sequences of preoperative breast magnetic resonance images of breast cancer patients, including T2WI, non-fsT1WI, and DCE-MRI 0-5, which can more comprehensively extract and analyze the multi-dimensional imaging genomics characteristics of tumors, enriching the information available for imaging genomics; the present invention provides a guarantee for improving the performance and generalization of imaging genomics in predicting axillary lymph node metastasis of breast cancer, and provides technical support for precision medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments, and the above and / or other advantages of the present invention will become more clear.

[0027] Figure 1 Flow chart of a method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation in an embodiment of the present invention;

[0028] Figure 2 It is the LASSO regression analysis cross validation curve in the embodiment of the present invention;

[0029] Figure 3 It is a LASSO coefficient curve diagram in an embodiment of the present invention;

[0030] Figure 4 is the ROC curve of all models in the training set in the embodiment of the present invention;

[0031] Figure 5 : is the ROC curve in the external validation set in the embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following is a further description in conjunction with the embodiments and drawings, but this should not limit the protection scope of the present invention.

[0033] Embodiment 1:

[0034] 1. Experimental data: Through retrospective analysis, preoperative multi-parametric MRI images of 556 patients who underwent breast MRI examinations in four centers (centers 1-4) from January 2014 to December 2023, and whose malignant lesions were confirmed by surgery or puncture pathology and had immunohistochemistry results were collected. Center 1 is the Second Affiliated Hospital of Nantong University, Center 2 is the Third Affiliated Hospital of Nantong University, Center 3 is Nantong Traditional Chinese Medicine Hospital, and Center 4 is the Affiliated Hospital of Nantong University. After screening according to the patient inclusion criteria (① patients who underwent MRI examination 2 weeks before surgery; ② MRI results showed breast masses in at least two consecutive layers; ③ positive cases of lymph node metastasis confirmed by lymph node dissection or ultrasonic fine needle aspiration biopsy, and negative cases confirmed by lymph node dissection surgery; ④ laboratory test results and clinical medical records were not missing) and exclusion criteria (① no enhancement of breast cancer lesions after enhancement; ② history of puncture or chemotherapy before breast MRI examination; ③ poor image quality), 506 patients remained, of which 365 patients were included in Center 1, and 5 patients were included in bilateral breast lesions. Center 1 had a total of 370 lesions, and Centers 2, 3 and 4 included 25, 11 and 105 patients, respectively. The multi-parameter MRI images include 8 sequences of T2-weighted imaging (T2WI), non-fat-suppressed T1-weighted imaging (nonfs-T1WI) and dynamic enhanced MRI (DCE-MRI) 0-5 phases

[0035] 2. Dataset division: Center 1 was divided into a training set (n=370), and Centers 2, 3, and 4 were combined as an external validation set (n=141). Patients in the training set and external validation set were divided into a positive group and a negative group according to whether the axillary lymph nodes were metastatic. For lesions with positive axillary lymph node metastasis (ALNM), they were confirmed by lymph node dissection or ultrasound-guided fine needle aspiration pathological examination. For lesions with negative ALNM, they were confirmed pathologically only by axillary lymph node dissection. There were 169 lesions in the ALNM-positive group of the training set, 201 lesions in the ALNM-negative group, and the ALNM-positive rate was 45.7%. There were 70 lesions in the ALNM-positive group of the external validation set, and 71 lesions in the ALNM-negative group, and the ALNM-positive rate was 49.6%.

[0036] 3. Data preprocessing: ROI delineation: All patients’ T2WI, nonfs-T1WI and DCE-MRI 0-5 phase 8 sequence images were imported into the Deepwise research platform (https: / / keyan.deepwise.com / , v2.6.0) in DICOM format to semi-automatically delineate the region of interest (ROI), that is, the lesion area was automatically segmented first, and then manually fine-tuned by two imaging experts with 15 and 8 years of MRI reading experience respectively. When delineating the ROI, factors such as cystic changes, necrosis and calcification inside the tumor will not be avoided;

[0037] Resampling: T2WI, nonfs-T1WI and DCE-MRI 0-5 phase 8 sequence images were resampled to the same size of 1×1×1mm;

[0038] Standardization processing: The window width and window position of all images of 8 sequences of T2WI, nonfs-T1WI and DCE-MRI phases 0-5 were standardized.

[0039] 4. Feature extraction: On the Deepwise research platform (https: / / keyan.deepwise.com / , v2.6.0), a discretization operation was performed to extract 1648 imaging features from the ROI of each lesion and each MR sequence (8 sequences including T2WI, non-fsT1WI, and DCE-MRI phases 0-5) in the training set and the external validation set.

[0040] The 1648 radiomics features include 14 shape features, 18 first-order features, 68 texture features, and 1548 high-order features extracted by various transformations of the original image, among which the texture features include 22 gray-level co-occurrence matrix (GLCM) features, 16 gray-level run-length matrix (GLRLM) features, 16 gray-level region size matrix (GLSZM) features, and 14 gray-level dependency matrix (GLDM) features; the transformations used by the high-order features include square (86), square root (86), logarithm (86), exponential (86), gradient size (86) and local binary pattern (86), and also include wavelet filtering (688) and Gaussian Laplace (344). 1648 features are extracted from each sequence, so a total of 13184 features are extracted from 8 sequences.

[0041] 5. Feature screening: First, ANOVA was used to screen features, with the variance threshold set to 1e5, and 2926 features were screened out. Then, mRMR was used to select features that maximized the correlation with the target category and had minimal redundant information between each other. The threshold was set to 130, and 130 features were screened out. Finally, LASSO regression was used to select all features with non-zero coefficients, and finally 19 radiomics features were screened out for subsequent analysis. Figure 2 The cross-validation curve of LASSO regression analysis of radiomics features is shown. The gray dotted line in the figure represents the optimal value of λ. Figure 3 This is the LASSO coefficient curve of the 19 radiomics features. Table 1 lists the names and coefficient values ​​of the 19 features;

[0042] Table 1 Screened radiomics features and coefficients

[0043] Feature Name coefficient T2WI_index_gldm_grayscale non-uniformity 0.05082014 T2WI_square root_glszm_grayscale variance -0.1003547 T2WI_l_log_glrlm_grayscale variance 0.07035194 nonfsT1WI_raw_glszm_region_size_nonuniformity 0.02143709 nonfsT1WI_log-σ-3-0-mm-3D_gldm_large dependency high grayscale emphasis -0.02576701 nonfsT1WI_Wavelet_HLL_glcm_Cluster Shadow 0.008178519 DCE-0_Wavelet-LLL_gldm_Large Dependency High Grayscale Emphasis -0.000731806 DCE-1_index_glrlm_trip non-uniformity 0.005392295 DCE-1_log_first_order_energy -0.09972658 DCE-1_Log_First_Order_Total_Energy -7.22545E-16 DCE-2_log_gldm_dependency_nonuniformity -0.04126104 DCE-3_Original_Shape_Surface Area 0.06134264 DCE-3_index_glszm_region size non-uniformity 0.00252967 DCE-3_index_glrlm_trip unevenness 0.08836839 DCE-4_square_root_glszm_grayscale variance -0.03344872 DCE-4_log-σ-2-0-mm-3D_gldm_large-dependency-high-grayscale-emphasis -0.005703682 DCE-4_Wavelet-LLL_gldm_Large Dependency High Grayscale Emphasis 0.08913544 DCE-5_log-σ-5-0-mm-3D_glszm_area size inhomogeneity 0.05786319 DCE-5_Wavelet-LHL_glszm_Large Area High Grayscale Emphasis -0.03250647

[0044] Wherein: T2WI, T2-weighted imaging; gldm, gray-level dependency matrix; glszm, gray-level region size matrix; glrlm, gray-level run-length matrix; nonfsT1WI, non-fat-suppressed T1-weighted imaging; log, Laplace of Gaussian; glcm, gray-level co-occurrence matrix; DCE, dynamic contrast enhancement.

[0045] 6. Domain adaptation-radiology model construction and external validation: When constructing the domain adaptation-radiology model, the key features selected in the training set in S5 and the corresponding same features in the external validation set are first input into the balanced distribution adaptation (BDA) domain adaptation algorithm. Then the BDA algorithm achieves the alignment of the two domains by minimizing the difference between the feature distributions of the source domain (training set) and the target domain (external validation set). The objective formula of the algorithm is as follows:

[0046]

[0047] sA T XHX T A=I.0≤μ≤1

[0048] Among them, X represents x s and x t The input data matrix is ​​composed of A, A represents the transformation matrix, μ is the balance factor, M0 and M c is the MMD matrix, c∈{1,2,···,C} is the label category, λ is the regularization parameter, is the Frobenius norm. I∈R (m+n)×(m+n) ' is the unit matrix, H is the center matrix, that is, H = I-(1 / n)1.

[0049] There are two terms in the formula: the adaptation of the marginal distribution and conditional distribution with a balancing factor (term 1) and the regularization term (term 2). And the formula has two constraints: the first constraint ensures that the data after the transformed distribution (A T X) should preserve the internal properties of the original data. The second constraint represents the range of the balancing factor μ.

[0050] The best way to adapt to the marginal distribution and conditional distribution of the two domains is to adjust the parameter μ of the above formula. If the difference in the marginal distribution of the two domains is considered to be more important, the value of μ is tended to 0. If the difference in the conditional distribution is considered to be more important, the value of μ is tended to 1. The new distribution is then used to construct the domain adaptation-radiology model on the training set using the classifier embedded in the BDA algorithm. Finally, the domain adaptation-radiology model predicts the label of the target domain, i.e., the external validation set, obtains a pseudo-label, and iteratively refines it to finally generate an output label for performance verification. In addition, in order to investigate the superiority of the domain adaptation algorithm, this embodiment constructs six traditional radiomics models for performance comparison. The following six models were developed on the training set: support vector machine (SVM), k nearest neighbor (KNN), logistic regression (LR), decision tree (DT), extreme gradient boosting (XGBoost), naive Bayes (NB), with the selected key features as input and lymph node metastasis status as output;

[0051] All trained models predicted the axillary lymph node status of each patient in the external validation set. The receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC) were used to evaluate the performance of each model in the training set and the external validation set, and all models were compared. Figure 3 The AUC results (95% confidence interval) and the p-values ​​obtained by the Delong test are shown in Table 2.

[0052] Table 2 AUC results of domain adaptation-radiology model and traditional radiomics model for predicting ALNM

[0053]

[0054]

[0055] Where: p-values ​​represent the comparative analysis of the BDA domain adaptation-radiology model with other traditional radiomics models in the training cohort and external validation cohort. AUC: area under the curve; BDA: balanced distribution adaptation; SVM: support vector machine; KNN: k-nearest neighbor; LR: logistic regression; DT: decision tree; XGBoost: extreme gradient boosting; NB: naive Bayes.

[0056] The results showed that the BDA domain adaptation-radiomics model had the best prediction performance, with an AUC of 0.978 in the training set. Among the six traditional radiomics models, DT performed best in the training set, with an AUC of only 0.866. On the other hand, the BDA domain adaptation-radiomics model had an AUC of 0.796 in the external validation set, which was higher than all traditional radiomics models. Among the traditional radiomics models, XGBoost had the highest AUC, which only reached 0.698. This shows that the domain adaptation-radiomics model can improve the low prediction performance and generalization ability of traditional radiomics models in external centers, which is conducive to the promotion and application of this method in various clinical centers.

[0057] Embodiment 2:

[0058] The computer-readable storage medium of this embodiment stores a computer program thereon, and when the program is executed by a processor, the steps in the method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation of Example 1 are implemented.

[0059] The computer-readable storage medium of this embodiment may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal; the computer-readable storage medium of this embodiment may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card, a secure digital card, a flash memory card, etc. equipped on the terminal; further, the computer-readable storage medium may also include both an internal storage unit of the terminal and an external storage device.

[0060] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or is to be output.

[0061] Embodiment 3:

[0062] The computer device of this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for predicting axillary lymph node metastasis of breast cancer based on imaging genomics and domain adaptation of Example 1 are implemented.

[0063] In this embodiment, the processor may be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, readily available programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0064] Those skilled in the art will appreciate that the disclosed content of the embodiments may be provided as methods, systems, or computer program products. Therefore, the present solution may adopt the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Moreover, the present solution may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program codes.

[0065] The present solution is described with reference to the flowchart and / or block diagram of the method and computer program product according to the embodiment of the present solution. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions; these computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0066] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0068] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0069] The present invention provides a method for predicting axillary lymph node metastasis of breast cancer based on imaging omics and domain adaptation. There are many methods and approaches to implement the technical solution. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention. All components not specified in this embodiment can be implemented using existing technologies.

Claims

1. A method for predicting axillary lymph node metastasis in breast cancer based on radiomics and domain adaptation. It is characterized in that The specific steps are: S1. Select multi-center data and divide them; S2. Preprocess the data in S1; S3. Extract features from the data preprocessed by S2; S4. Filter the features extracted in S3; S5. Build and validate the domain-adapted radiomics model.

2. The prediction method according to claim 1, It is characterized in that The multi-center data are multi-parameter MRI images; the multi-parameter MRI images are 8 sequences of phases 0-5 of T2WI, nonfs-T1WI and DCE-MRI.

3. The prediction method according to claim 1, It is characterized in that The specific steps of S2 are: segmenting the divided data and outlining the ROI; and resampling and standardizing the outlined image.

4. The prediction method according to claim 1, It is characterized in that In S3, the features in the feature extraction process are shape features, first-order features, texture features and high-order features obtained by performing various transformations on the original image; the texture features are grayscale co-occurrence matrix features, grayscale run-length matrix features, grayscale area size matrix features and grayscale dependence matrix features; the transformations used in the high-order features are square, square root, logarithm, exponential, gradient size, local binary pattern, wavelet filtering and Gaussian Laplace.

5. The method according to claim 1, It is characterized in that In S4, feature screening was performed by ANOVA, mRMR and LASSO regression.

6. The method according to claim 1, It is characterized in that The construction process of the domain adaptation-radiology model is as follows: input the screened key features and the corresponding same features in the external validation set, use the balanced distribution adaptation domain adaptation algorithm to adjust the feature distribution of the source domain and the target domain to achieve alignment of the two domains, and use the new distribution to use the classifier embedded in the balanced distribution adaptation domain adaptation algorithm to complete the construction of the domain adaptation-radiology model on the training set.

7. The method according to claim 6, It is characterized in that The objective formula of the balanced distribution adaptation domain adaptation algorithm is as follows: Among them, X represents x s and x t The input data matrix is ​​composed of A, A represents the transformation matrix, μ is the balance factor, M 0 and M c is the MMD matrix, c∈{1,2,···,C} is the label category, λ is the regularization parameter, is the Frobenius norm; I∈R (m +n)×(m+n)' is the unit matrix, H is the center matrix, that is, H = I-(1 / n)1.

8. A computer-readable storage medium having a computer program stored thereon, Features: When the program is executed by a processor, the steps in the method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation are implemented as described in any one of claims 1 to 7.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the steps in the method for predicting axillary lymph node metastasis of breast cancer based on radiomics and domain adaptation are implemented as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Photoacoustic pathological image processing method, storage medium and terminal equipment

    CN113299372A

  • Multiple instance learner for prognostic tissue pattern identification

    CN113454733A

  • Lymph node metastasis state classification prediction method and device based on radiomics

    CN117711615A

  • Predicting tumor prognoses based on a combination of radiomic and clinico-pathological features

    US20210110540A1

Cited By

  • Multi-center data aggregation method

    CN119479830A