Prediction method, device and equipment of pCR based on DCE-MRI

Through a DCE-MRI-based method, combining imagingomics and clinical features, the pCR category of breast cancer patients was extracted and predicted, and the problems of low accuracy and lag after neoadjuvant chemotherapy in the prior art were solved, and pCR prediction was improved accuracy and timeliness.

CN119008006AActive Publication Date: 2024-11-22DONGGUAN PEOPLES HOSPITAL +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411091149.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-11-22
Estimated Expiration
2044-08-09

AI Technical Summary

Technical Problem

The prior art has low accuracy and lag in the evaluation of pathological responses after neoadjuvant chemotherapy in breast cancer patients, which affects the timeliness and accuracy of pCR prediction.

Method used

Using a prediction method based on DCE-MRI, by obtaining DCE-MRI sets and auxiliary information, including SER information and clinical information, it is input into the trained pCR prediction model, extracting image features and target features, combining imaging omics and clinical features Perform pCR category prediction.

Benefits of technology

It improves the accuracy of pCR prediction and the timeliness of neoadjuvant chemotherapy response evaluation, and enhances the predictive ability of breast cancer treatment effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119008006A_ABST
    Figure CN119008006A_ABST
Patent Text Reader

Abstract

The invention discloses a DCE-MRI-based pCR prediction method, device and equipment, and the method comprises the steps: obtaining a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, inputting the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determining image features corresponding to the DCE-MRI set through the pCR prediction model, and outputting the image features corresponding to the DCE-MRI set. And determining a target feature based on the image feature and the auxiliary information, and predicting a pCR category based on the target feature. According to the invention, the SER information is used as the supplementary information of the clinical information, and the auxiliary information including the clinical information and the SER information is combined with the DCE-MRI set, so that the pCR prediction model can learn the image features carried by the DCE-MRI set and the clinical features carried by the auxiliary information, and the pCR prediction is carried out in combination with the image features and the clinical features, thereby improving the accuracy of the pCR prediction, and improving the accuracy of the pCR prediction. And thus, the accuracy of NAC response evaluation can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of biomedical engineering technology, and in particular to a method, device and equipment for predicting pCR based on DCE-MRI. Background Art

[0002] Breast cancer is one of the most common malignant tumors with a high incidence rate. Surgical resection is usually the first choice for breast cancer treatment. However, for patients with locally advanced disease, the extensiveness of the tumor or the invasion of adjacent tissues makes breast-conserving surgery difficult to perform. Therefore, neoadjuvant chemotherapy (NAC) for breast cancer patients has become the standard choice for the treatment of locally advanced breast cancer, aiming to improve the tumor resection rate and the possibility of breast-conserving surgery by reducing the clinical stage of the tumor. On the other hand, pCR, that is, the absence of residual invasive cancer cells or carcinoma in situ in the breast, has been confirmed to be the best prognostic indicator after NAC treatment. However, only about 30% of patients can achieve pCR, and a small number of patients will still progress during NAC. In addition, since the evaluation of the pathological response status needs to be completed through postoperative pathological examination, this process has a certain lag, which in turn affects the accuracy and timeliness of pCR prediction.

[0003] Therefore the prior art still needs to be improved and enhanced. Summary of the invention

[0004] The technical problem to be solved by the present application is to provide a method, device and equipment for predicting pCR based on DCE-MRI in view of the deficiencies of the existing technology.

[0005] In order to solve the above technical problems, the first aspect of the present application provides a method for predicting pCR based on DCE-MRI, wherein the method for predicting pCR based on DCE-MRI specifically includes:

[0006] Acquire a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, wherein the auxiliary information includes SER information and clinical information, and the DCE-MRI in the DCE-MRI set carries a breast cancer lesion area;

[0007] The DCE-MRI set and the auxiliary information are input into a trained pCR prediction model, the image features corresponding to the DCE-MRI set are determined by the pCR prediction model, target features are determined based on the image features and the auxiliary information, and the pCR category is predicted based on the target features.

[0008] The method for predicting pCR based on DCE-MRI, wherein the DCE-MRI set includes a first DCE-MRI before NAC and at least a second DCE-MRI in NAC.

[0009] In the method for predicting pCR based on DCE-MRI, the process of obtaining the SER information specifically includes:

[0010] Extracting the breast background region of interest of the first DCE-MRI and the breast background region of interest of the second DCE-MRI;

[0011] Locating a target lesion region with the largest signal intensity of the first DCE-MRI and a target lesion region with the largest signal intensity of the second DCE-MRI;

[0012] Based on the breast background region of interest and the target lesion region of the first DCE-MRI and the breast background region of interest and the target lesion region of the second DCE-MRI, the SER information corresponding to the DCE-MRI set is determined.

[0013] In the DCE-MRI-based pCR prediction method, the target lesion area positioning process specifically includes:

[0014] Extracting a first three-dimensional mask of a breast background region of interest and a second three-dimensional mask of a breast cancer lesion region;

[0015] constructing a slidable region adapted to the breast background region of interest based on the first three-dimensional mask;

[0016] The slidable area is slid within the second three-dimensional mask, and the signal strength of the slidable area at each sliding position is calculated to obtain a target lesion area.

[0017] The method for predicting pCR based on DCE-MRI, wherein determining the image features corresponding to the DCE-MRI set by the pCR prediction model specifically includes:

[0018] extracting radiomic features of the DCE-MRI set;

[0019] Extracting slice features of the DCE-MRI set;

[0020] Feature screening is performed on the imaging omics features and the extracted slice features respectively to obtain image features.

[0021] The pCR prediction method based on DCE-MRI, wherein the pCR prediction model includes an image feature extraction module, a splicing module and a classifier, the image feature extraction module is connected to the splicing module, and the splicing module is connected to the classifier; the image feature extraction module includes an imaging genomics feature extraction unit, a slice feature extraction unit and a feature screening module, the imaging genomics feature extraction unit and the slice feature extraction unit are in parallel and are both connected to the feature screening unit; the input items of the splicing module include image features determined by the image feature extraction module and auxiliary information.

[0022] The method for predicting pCR based on DCE-MRI, wherein the step of determining the target feature based on the image feature and the auxiliary information specifically includes:

[0023] Obtaining an information feature vector corresponding to the auxiliary information;

[0024] The information feature vector is concatenated with the image feature to obtain the target feature.

[0025] The second aspect of the present application provides a device for predicting pCR based on DCE-MRI, wherein the device for predicting pCR based on DCE-MRI specifically comprises:

[0026] An acquisition module, used for acquiring a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, wherein the auxiliary information includes SER information and clinical information, and the DCE-MRI set carries a breast cancer lesion area;

[0027] A control module is used to input the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determine the image features corresponding to the DCE-MRI set through the pCR prediction model, determine the target features based on the image features and the auxiliary information, and predict the pCR category based on the target features.

[0028] A third aspect of the present application provides a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in any of the DCE-MRI-based pCR prediction methods described above.

[0029] A fourth aspect of the present application provides a terminal device, comprising: a processor and a memory;

[0030] The memory stores a computer-readable program executable by the processor;

[0031] When the processor executes the computer-readable program, the processor implements the steps in any of the above-described methods for predicting pCR based on DCE-MRI.

[0032] Beneficial effects: Compared with the prior art, the present application provides a method, device and apparatus for predicting pCR based on DCE-MRI, the method comprising obtaining a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, inputting the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determining the image features corresponding to the DCE-MRI set through the pCR prediction model, determining the target features based on the image features and the auxiliary information, and predicting the pCR category based on the target features. The present application uses SER information as supplementary information for clinical information, and combines auxiliary information including clinical information and SER information with the DCE-MRI set, so that the pCR prediction model can learn the image features carried by the DCE-MRI set and the clinical features carried by the auxiliary information, and combine the image features and clinical features to perform pCR prediction, thereby improving the accuracy of pCR prediction, and further improving the accuracy of NAC response assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 Flow chart of the method for predicting pCR based on DCE-MRI provided in an embodiment of the present application.

[0035] Figure 2 This is an example diagram of the SER information acquisition process.

[0036] Figure 3 A principle flow chart of a specific implementation of the DCE-MRI-based pCR prediction method provided in an embodiment of the present application.

[0037] Figure 4 This is a schematic diagram of the structure of a device for predicting pCR based on DCE-MRI provided in an embodiment of the present application.

[0038] Figure 5 A schematic diagram of the structure of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0039] The embodiments of the present application provide a method, device and apparatus for predicting pCR based on DCE-MRI. In order to make the purpose, technical solution and effect of the present application clearer and more specific, the present application is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0040] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0041] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0042] It should be understood that the sequence numbers and sizes of the steps in this embodiment do not mean the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0043] After research, it was found that breast cancer is one of the most common and highly prevalent malignant tumors in the world. Surgical resection is usually the preferred treatment for breast cancer. However, for patients with locally advanced disease, the extensiveness of the tumor or the invasion of adjacent tissues makes breast-conserving surgery difficult to perform. Therefore, neoadjuvant chemotherapy (NAC) for breast cancer patients has become the standard choice for the treatment of locally advanced breast cancer, aiming to increase the tumor resection rate and the possibility of breast-conserving surgery by reducing the clinical stage of the tumor. On the other hand, pCR, that is, the absence of residual invasive cancer cells or carcinoma in situ in the breast, has been confirmed to be the best prognostic indicator after NAC treatment. However, only about 30% of patients can achieve pCR, and a small number of patients will still progress during NAC. In addition, since the evaluation of the pathological response status needs to be completed through postoperative pathological examination, this process has a certain lag.

[0044] In recent years, some studies have shown that breast imaging can predict pCR, and DCE-MRI can be used to evaluate the tumor response to NAC, and its effect is better than breast X-ray examination, ultrasound examination and other imaging methods. DCE-MRI is a non-invasive comprehensive examination that continuously monitors the distribution and washout process of contrast agents in tissues after injection, provides quantitative information about the vascular permeability and hemodynamics of tissues, and can show changes in tumor microcirculation and angiogenesis, thus helping to predict the final outcome after the completion of NAC for breast cancer.

[0045] At present, imaging omics is commonly used to evaluate NAC response based on DCE-MRI, that is, to extract massive quantitative features from accurately segmented medical images to reveal the deep connection between these features and potential pathophysiological processes, so as to significantly reflect the microscopic tissue structure of the tumor. However, due to the high heterogeneity and complexity of the internal structure of malignant tumors, the kinetic characteristics and texture characteristics of drug metabolism in tumors will change significantly during the treatment process, resulting in the problem of low accuracy in evaluating NAC response based on imaging omics.

[0046] In order to solve the above problems, in an embodiment of the present application, a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set are obtained, the DCE-MRI set and the auxiliary information are input into a trained pCR prediction model, the image features corresponding to the DCE-MRI set are determined by the pCR prediction model, and the target features are determined based on the image features and the auxiliary information, and the pCR category is predicted based on the target features. The present application uses SER information as supplementary information for clinical information, and combines auxiliary information including clinical information and SER information with the DCE-MRI set, so that the pCR prediction model can learn the image features carried by the DCE-MRI set and the clinical features carried by the auxiliary information, and combine the image features and clinical features to perform pCR prediction, thereby improving the accuracy of pCR prediction, and thus improving the accuracy of NAC response assessment.

[0047] The application content is further explained below through the description of embodiments in conjunction with the accompanying drawings.

[0048] This embodiment provides a method for predicting pCR based on DCE-MRI, such as Figure 1 As shown, the method includes:

[0049] S10. Acquire a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set.

[0050] Specifically, the DCE-MRI (dynamic enhanced magnetic resonance image) in the DCE-MRI set is obtained by performing a breast MRI examination on a breast cancer patient, and each DCE-MRI in the DCE-MRI set carries a breast cancer lesion area. The auxiliary information is used for the clinical characteristics of the breast cancer patient corresponding to the DCE-MRI set, wherein the auxiliary information includes SER (Signal Enhancement Rate, signal intensity ratio of breast cancer lesions to breast background substance) information and clinical information. The SER information is a quantitative index, which is the ratio of the signal intensity of the selected tumor area to the signal intensity of the circular region of interest of the breast background. The SER information can reflect the absorption and accumulation of contrast agents in the tumor area. The clinical information may include one or more of the tumor proliferation index Ki67, progesterone receptor (PR), estrogen receptor (ER), and human epidermal growth factor receptor 2 (HER2). Of course, in actual applications, clinical information may also include patient information, wherein the patient information may include age of first onset, menstrual status, fertility status, clinical stage, NAC treatment regimen, chemotherapy cycle, and immunohistochemistry results.

[0051] The DCE-MRI set may be a DCE-MRI before NAC, or a DCE-MRI during NAC, or a DCE-MRI before NAC and a DCE-MRI during NAC. In an embodiment of the present application, the DCE-MRI set includes a DCE-MRI set before NAC and at least one DCE-MRI during NAC, that is, the DCE-MRI set includes a first DCE-MRI before NAC and at least one second DCE-MRI during NAC, for example, the DCE-MRI set includes a first DCE-MRI within two weeks before NAC and a second DCE-MRI in the early stage of NAC, or the DCE-MRI set includes a first DCE-MRI within two weeks before NAC, a second DCE-MRI in the early stage of NAC, and a second DCE-MRI in the middle stage of NAC, etc. The embodiment of the present application obtains a first DCE-MRI before NAC and a second DCE-MRI in at least one NAC, and then obtains image features through the first DCE-MRI before NAC and the second DCE-MRI in at least one NAC, so that the image features can learn that the kinetic characteristics and texture characteristics of drug metabolism inside the tumor change significantly during the treatment process, thereby improving the accuracy of pCR prediction.

[0052] NAC (Neoadjuvant Chemotherapy, neoadjuvant chemotherapy for breast cancer patients), pCR means no residual invasive cancer cells or carcinoma in situ in the breast, pCR is the prognostic indicator of NAC, and the prognosis of NAC can be predicted by predicting pCR. Among them, the pCR category is divided based on the prognostic indicators of NAC, which includes pCR and non-pCR. Non-pCR means that the prognostic indicators of NAC belong to level 1-4, and pCR means that the prognostic indicators of NAC are level 5. The prognostic indicators of NAC are graded according to Miller-Payne (MP), among which level 1 means that the invasive cancer cells have not changed or only individual cancer cells have changed, and the number of cancer cells has not decreased overall; level 2 means that the invasive cancer cells have decreased slightly, but the total number is still high, and the reduction of cancer cells does not exceed 30%; level 3 means that the invasive cancer cells have decreased between 30% and 90%; level 4 means that the invasive cancer cells have decreased significantly by more than 90%, and only scattered small clusters of cancer cells or single cancer cells remain; level 5 means that there are no invasive cancer cells in the original tumor bed, but ductal carcinoma in situ may exist.

[0053] It should be noted that since the first DCE-MRI and the second DCE-MRI are acquired at different times, there may be differences in the resolution of the first DCE-MRI and the second DCE-MRI, which in turn leads to different voxel spacings between the first DCE-MRI and the second DCE-MRI. For example, the voxel spacing of the first DCE-MRI is [0.7, 0.7, 2.91], and the voxel spacing of the second DCE-MRI is [0.7, 0.7, 2.93]. Different voxel spacings will affect the manifestation of the tumor in the image, and thus affect the calculation of its related features. In order to obtain the first DCE-MRI and the second DCE-MRI, the first DCE-MRI and the second DCE-MRI can be resampled to ensure that the features extracted from all images are consistent in spatial scale, ensuring the comparability between features, thereby enhancing the generalization ability of the machine learning classification model established subsequently.

[0054] Further, the DCE-MRI set includes the first DCE-MRI before NAC and the second DCE-MRI in at least one NAC, and accordingly, the SER information includes the SER information of the first DCE-MRI and the SER information of the second DCE-MRI. Therefore, when acquiring the SER information, the SER information of the first DCE-MRI and the second DCE-MRI will be acquired in sequence. In addition, since the SER information is used to represent the signal intensity ratio between the breast cancer lesion and the breast background substance, the signal intensity of the breast background substance (i.e., the breast background region of interest) and the maximum signal intensity of the breast cancer lesion will be extracted when acquiring the SER information, and then the SER information will be calculated based on the two acquired signal intensities. Based on this, the SER information acquisition process specifically includes:

[0055] Extracting the breast background region of interest of the first DCE-MRI and the breast background region of interest of the second DCE-MRI;

[0056] Locating a target lesion region with the largest signal intensity of the first DCE-MRI and a target lesion region with the largest signal intensity of the second DCE-MRI;

[0057] Based on the breast background region of interest and the target lesion region of the first DCE-MRI and the breast background region of interest and the target lesion region of the second DCE-MRI, the SER information corresponding to the DCE-MRI set is determined.

[0058] Specifically, the breast background regions of interest of the first DCE-MRI and the second DCE-MRI are both circular regions of interest, which can be determined by deep learning, or by traditional edge recognition methods, or by manual annotation. In a typical implementation, the breast background regions of interest and the breast cancer lesion regions are extracted using the Nibabel library. Specifically, after acquiring the first DCE-MRI and the second DCE-MRI, the first DCE-MRI and the second DCE-MRI are loaded using the Nibabel library, and the breast background regions of interest and the breast cancer lesion regions in the first DCE-MRI and the second DCE-MRI are extracted respectively, and then the target lesion region with the largest signal intensity is selected in the breast cancer lesion region, and finally the SER information is calculated based on the signal intensity of the breast background regions of interest and the signal intensity of the target lesion region, wherein the SER information of the first DCE-MRI is calculated based on the breast background regions of interest and the target lesion region of the first DCE-MRI, and the SER information of the second DCE-MRI is based on the breast background regions of interest and the target lesion region of the second DCE-MRI. The present application quantifies the signal enhancement degree of breast tumors by calculating the SER information which is used to reflect the ratio of the signal intensity of the breast background region of interest and the maximum signal intensity in the breast cancer lesion area. In this way, the pCR prediction model can learn the signal enhancement degree of breast tumors through the SER information, and thus can predict the pCR more accurately.

[0059] It should be noted that when locating the target lesion area with the largest signal strength, it can be located by manual annotation, or by deep learning, or by sliding the breast background region of interest. In one implementation of the embodiment of the present application, the target lesion area is located by sliding the breast background region of interest. Accordingly, the target lesion area positioning process specifically includes:

[0060] Extracting a first three-dimensional mask of a breast background region of interest and a second three-dimensional mask of a breast cancer lesion region;

[0061] constructing a slidable region adapted to the breast background region of interest based on the first three-dimensional mask;

[0062] The slidable area is slid within the second three-dimensional mask, and the signal strength of the slidable area at each sliding position is calculated to obtain a target lesion area.

[0063] Specifically, Figure 2As shown, the slidable area is a three-dimensional area, and the shape and volume of the slidable area are consistent with the shape and volume of the first three-dimensional mask. That is to say, when constructing the slidable area, the shape and volume of the breast background region of interest will be obtained first, and then a slidable three-dimensional matrix will be constructed based on the acquired shape and volume to obtain a slidable area that is compatible with the breast background region of interest. Wherein, the shape and volume of the breast background region of interest can be determined by calculating the maximum and minimum values ​​of the first three-dimensional mask in three dimensions of x, y, and z. The embodiment of the present application constructs a slidable area that is compatible with the breast background region of interest, and then determines the target lesion area by sliding the slidable area, so that the target lesion area can be quickly determined, and the shape and volume of the target lesion area can be made consistent with the shape and volume of the breast background region of interest, thereby improving the accuracy of the calculated SER information.

[0064] S20, inputting the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determining the image features corresponding to the DCE-MRI set through the pCR prediction model, determining the target features based on the image features and the auxiliary information, and predicting the pCR category based on the target features.

[0065] Specifically, the pCR prediction model is trained and used to predict the pathological complete response of neoadjuvant chemotherapy for breast cancer, that is, it is used to predict the pCR category in the process of neoadjuvant chemotherapy for breast cancer based on the DCE-MRI set and auxiliary information to predict the pathological complete response of neoadjuvant chemotherapy for breast cancer. Among them, the pCR category includes non-pCR and pCR. The NAC grade can be predicted by the pCR category, and then the prognosis of NAC can be predicted. The image features are extracted by the pCR prediction model based on the DCE-MRI set, and the target features are obtained based on the image features and auxiliary information. It carries the feature information included in the image features and the feature information included in the auxiliary information, enriching the feature information that can be learned by the pCR prediction model and improving the prediction accuracy of the pCR prediction model.

[0066] Exemplarily, the pCR prediction model may include an image feature extraction module, a splicing module, and a classifier, wherein the image feature extraction module is used to extract image features corresponding to the DCE-MRI set, the splicing module is used to determine target features based on the image features and the auxiliary information, and the classifier is used to predict the pCR category based on the target features. That is, after the DCE-MRI set and the auxiliary information are input into the trained pCR prediction model, the image feature extraction module is used to extract features from the DCE-MRI set to obtain image features. After the image features and the auxiliary information are input into the splicing module, the image features and the auxiliary information are spliced ​​into the target features through the splicing module; the target features are input into the classifier, and the pCR category is output through the classifier.

[0067] It should be noted that, since the DCE-MRI set may include two or more DCE-MRIs, the embodiment of the present application inputs each DCE-MRI in the DCE-MRI set into the pCR prediction model, and extracts features from each DCE-MRI in turn through the image feature extraction module to obtain the image features corresponding to each DCE-MRI, and then splices the image features of each DCE-MRI as image features, or, the image feature set of each DCE-MRI is used as image features, etc. Of course, in actual applications, before the DCE-MRI set is input into the pCR prediction model, the DCE-MRI set may be spliced ​​in the channel direction to obtain a spliced ​​DCE-MRI, and then the spliced ​​DCE-MRI may be input into the pCR prediction model, and the image feature extraction module may be used to extract features from the spliced ​​DCE-MRI to obtain image features.

[0068] In one implementation, the image features include radiomics features and slice features, that is, the pCR prediction model extracts radiomics features and slice features from DCE-MRI. Accordingly, the image features corresponding to the DCE-MRI set determined by the pCR prediction model specifically include:

[0069] extracting radiomic features of the DCE-MRI set;

[0070] Extracting slice features of the DCE-MRI set;

[0071] Feature screening is performed on the imaging omics features and the extracted slice features respectively to obtain image features.

[0072] Specifically, the image feature extraction module is used to extract radiomics features and slice features, and to screen the extracted radiomics features and slice features. Figure 3As shown, the image feature extraction module may include an imaging genomics feature extraction unit, a slice feature extraction unit, and a feature screening module. The imaging genomics feature extraction unit and the slice feature extraction unit are in parallel and are both connected to the feature screening unit. The imaging genomics feature extraction unit is used to extract imaging genomics features, the slice feature extraction unit is used to extract slice features, and the feature screening unit is used to screen imaging genomics features and slice features. That is, Figure 3 As shown, the pCR prediction model includes an image feature extraction module, a splicing module and a classifier, the image feature extraction module is connected to the splicing module, and the splicing module is connected to the classifier; the image feature extraction module includes an imaging genomics feature extraction unit, a slice feature extraction unit and a feature screening module, the imaging genomics feature extraction unit and the slice feature extraction unit are parallel and are both connected to the feature screening unit; the input items of the splicing module include image features and auxiliary information determined by the image feature extraction module. The embodiment of the present application can learn more feature information by extracting the imaging genomics features and slice features of DCE-MRI, thereby improving the prediction accuracy of the pCR prediction model.

[0073] Radiomic features can include intensity features, shape features, texture features, and transformation features. Radiomic features can be extracted by radiomics, for example, a python-based radiomics toolkit can automatically extract radiomic features. In addition, radiomic features extracted by radiomics generally contain a large number of redundant features, which may lead to overfitting and "dimensionality disaster". For this reason, after the radiomic features are extracted, they are screened to retain the features most relevant to pCR prediction.

[0074] In one implementation, the radiomics features can be screened by using a coarse-to-fine feature screening method. Specifically, the F_class if feature selection method in the scikit-learn toolkit can be used to perform a rough screening of the radiomics features to obtain the radiomics features after the rough screening (for example, retaining the 10 most statistically significant radiomics features, etc.), wherein the F_class feature selection method is to evaluate the importance of the feature by calculating the F statistic between each radiomics feature and the pCR and non-category for rough screening. Secondly, the Least Absolute Shrinkage and Selection Operator (LASSO) is introduced to perform a fine screening on the radiomics features after the rough screening to obtain the radiomics features after the screening corresponding to DCE-MRI.

[0075] It should be noted that the DCE-MRI set includes a first DCE-MRI and at least one second DCE-MRI, so that when performing radiomics feature extraction, radiomics feature extraction and screening are performed on the first DCE-MRI and the second DCE-MRI respectively, and the number of features of the radiomics features corresponding to the first DCE-MRI may be the same as the number of features of the radiomics features corresponding to the second DCE-MRI, or may be different from the number of features of the radiomics features corresponding to the second DCE-MRI. In an embodiment of the present application, the number of features of the radiomics features corresponding to the first DCE-MRI is different from the number of features of the radiomics features corresponding to the second DCE-MRI, so that in addition to retaining the radiomics features most relevant to pCR prediction, the number of radiomics features can be minimized, thereby better avoiding overfitting and "dimensionality disaster". For example, the first DCE-MRI data retains 5 radiomics features, and the second DCE-MRI data retains 4 radiomics features.

[0076] The slice feature is obtained by learning a visual language model for DCE-MRI, and the sub-slice features of each slice in the DCE-MRI are extracted by the visual language model, and then all the sub-slice features are fused to obtain the slice feature. When the DCE-MRI set includes a first DCE-MRI and at least one second DCE-MRI, the slice feature is all sub-slice features of the first DCE-MRI and all sub-slice features of each second DCE-MRI. In one implementation, the visual language model can use the image encoder of the BiomedCLIP model, and the sub-slice features of each slice of the first DCE-MRI and the second DCE-MRI are extracted by the image encoder of the BiomedCLIP model, and then the sub-slice features of each slice of the first DCE-MRI and the second DCE-MRI are fused to obtain the slice feature. In addition, before extracting features through the image encoder of the BiomedCLIP model, the DCE-MRI can be cropped for ROI to focus on the breast cancer lesion area and reduce the interference of redundant information. The ROI cropping can be carried out by equidistant diffusion from the ROI center to obtain the ROI area, and the ROI area is used as the DCE-MRI for slice feature extraction. For example, the ROI center is used as the starting point to diffuse equidistantly to obtain a 224×224 pixel ROI area.

[0077] In one implementation, determining the target feature based on the image feature and the auxiliary information specifically includes:

[0078] Obtaining an information feature vector corresponding to the auxiliary information;

[0079] The information feature vector is concatenated with the image feature to obtain the target feature.

[0080] Specifically, the auxiliary information includes SER information and clinical information, so that when determining the target features based on the auxiliary information and image features, the SER information and the clinical information can be spliced ​​into an information feature vector, or the clinical information can be first screened to remove redundant information, and then the screened clinical information and the SER information can be spliced ​​into an information feature vector; or the clinical information can be first passed through a text feature extraction module, and the clinical information is extracted by the text feature extraction module to obtain clinical features, and then the clinical features and the SER information are combined to obtain an information feature vector. The information feature vector is spliced ​​with the image features to obtain the target features. Of course, in practical applications, the SER information, clinical information, and the radiomics features and slice features in the image features can also be directly spliced ​​to obtain the target features.

[0081] In one implementation, the predicted pCR category is obtained by classification based on the target feature, wherein the classification process is performed by a classifier, and the classifier can be an LR (Logistic Regression) classifier, which converts a linear combination of one or more independent variables (or features, explanatory variables) into a probability prediction by applying a logistic function (Sigmoid). Given the independent variables, the goal is to predict the probability of a binary response variable to predict the pCR category.

[0082] In order to further illustrate the prediction method of pCR based on DCE-MRI provided in the embodiment of the present application, the present application conducted a comparative test. In the comparative test, first, without combining clinical characteristics, three feature extraction methods of radiomics, BiomedCLIP, and radiomics combined with BiomedCLIP were used respectively, SER was introduced, a prediction model was constructed, and each model without introducing SER was compared. On the other hand, in combination with clinical characteristics, three feature extraction methods of radiomics, BiomedCLIP, and radiomics combined with BiomedCLIP were used respectively, and SER was introduced to construct a prediction model. Similarly, the prediction model constructed by this method was compared with the corresponding model without introducing SER. Such an experimental design idea is to explore whether SER can contribute to improving the accuracy of pCR prediction independently of clinical characteristics.

[0083] The final results are as follows: When analyzing the changes in model performance under different feature combinations based on multiple periods of NAC, it can be observed that the performance of the trained model on the test set did not improve significantly when only SER was added. However, when SER was combined with clinical features, the performance of each trained model on the test set improved. This confirms that introducing SER combined with clinical features into the model is effective and can help the model better capture the treatment response of breast cancer patients to NAC.

[0084] In summary, when using radiomics methods, the prediction models based on pre-NAC, early NAC and clinical information had AUCs of 0.641, 0.603, and 0.513 on the validation set, and AUCs of 0.506, 0.677, and 0.628 on the test set, respectively. The prediction model combining multiple periods and clinical characteristics had an AUC of 0.633 on the validation set and an AUC of 0.665 on the test set. When the BiomedCLIP method was combined with radiomics, the AUC of the model increased to 0.830 on the validation set and to 0.773 on the test set. Finally, the introduction of SER increased the AUC of the model to 0.842 on the validation set and to 0.781 on the test set.

[0085] In summary, this embodiment provides a prediction method for pCR based on the above-mentioned DCE-MRI, which includes obtaining a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, inputting the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determining the image features corresponding to the DCE-MRI set through the pCR prediction model, and determining the target features based on the image features and the auxiliary information, and predicting the pCR category based on the target features. The present application first introduces SER as an auxiliary feature, and uses it together with clinical information as auxiliary information. Secondly, effective image depth features are extracted through BiomedCLIP, and the features extracted by BiomedCLIP are combined with imaging genomics features as image features. Finally, the auxiliary information and image features are combined to predict the pCR category, so that the pCR prediction model can learn a variety of feature information, and can better predict pCR during breast cancer NAC.

[0086] Based on the above-mentioned DCE-MRI-based pCR prediction method, this embodiment provides a DCE-MRI-based pCR prediction device, such as Figure 4 As shown, the pCR prediction device based on DCE-MRI specifically includes:

[0087] An acquisition module 100 is used to acquire a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, wherein the auxiliary information includes SER information and clinical information, and the DCE-MRI set carries a breast cancer lesion area;

[0088] The control module 200 is used to input the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determine the image features corresponding to the DCE-MRI set through the pCR prediction model, determine the target features based on the image features and the auxiliary information, and predict the pCR category based on the target features.

[0089] Based on the above-mentioned DCE-MRI-based pCR prediction method, this embodiment provides a computer-readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the steps in the DCE-MRI-based pCR prediction method as described in the above-mentioned embodiment.

[0090] Based on the above-mentioned DCE-MRI-based pCR prediction method, the present application also provides a terminal device, such as Figure 5 As shown, it includes at least one processor 20, a display screen 21, and a memory 22, and may also include a communication interface 23 and a bus 24. The processor 20, the display screen 21, the memory 22, and the communication interface 23 may communicate with each other through the bus 24. The display screen 21 is configured to display a preset user guide interface in the initial setting mode. The communication interface 23 may transmit information. The processor 20 may call the logic instructions in the memory 22 to execute the method in the above embodiment.

[0091] In addition, the logic instructions in the memory 22 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0092] The memory 22 is a computer-readable storage medium that can be configured to store software programs, computer executable programs, such as program instructions or modules corresponding to the methods in the embodiments of the present disclosure. The processor 20 executes functional applications and data processing by running the software programs, instructions or modules stored in the memory 22, that is, implementing the methods in the above embodiments.

[0093] The memory 22 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory 22 may include a high-speed random access memory and may also include a non-volatile memory. For example, a variety of media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, may also be a transient storage medium.

[0094] In addition, the specific process of loading and executing the multiple instruction processors in the above-mentioned storage medium and the terminal device has been described in detail in the above-mentioned method, and will not be described one by one here.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. Prediction method of pCR based on DCE-MRI, It is characterized in that The method for predicting pCR based on DCE-MRI specifically includes: Acquire a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, wherein the auxiliary information includes SER information and clinical information, and the DCE-MRI in the DCE-MRI set carries a breast cancer lesion area; The DCE-MRI set and the auxiliary information are input into a trained pCR prediction model, the image features corresponding to the DCE-MRI set are determined by the pCR prediction model, target features are determined based on the image features and the auxiliary information, and the pCR category is predicted based on the target features.

2. The method for predicting pCR based on DCE-MRI according to claim 1, It is characterized in that The DCE-MRI set includes a first DCE-MRI before NAC and at least one second DCE-MRI in NAC.

3. The method for predicting pCR based on DCE-MRI according to claim 2, It is characterized in that The process of obtaining the SER information specifically includes: Extracting the breast background region of interest of the first DCE-MRI and the breast background region of interest of the second DCE-MRI; Locating a target lesion region with the largest signal intensity of the first DCE-MRI and a target lesion region with the largest signal intensity of the second DCE-MRI; Based on the breast background region of interest and the target lesion region of the first DCE-MRI and the breast background region of interest and the target lesion region of the second DCE-MRI, the SER information corresponding to the DCE-MRI set is determined.

4. The method for predicting pCR based on DCE-MRI according to claim 3, It is characterized in that The process of locating the target lesion area specifically includes: Extracting a first three-dimensional mask of a breast background region of interest and a second three-dimensional mask of a breast cancer lesion region; constructing a slidable region adapted to the breast background region of interest based on the first three-dimensional mask; The slidable area is slid within the second three-dimensional mask, and the signal strength of the slidable area at each sliding position is calculated to obtain a target lesion area.

5. The method for predicting pCR based on DCE-MRI according to claim 1, It is characterized in that Determining the image features corresponding to the DCE-MRI set by the pCR prediction model specifically includes: extracting radiomic features of the DCE-MRI set; Extracting slice features of the DCE-MRI set; Feature screening is performed on the imaging omics features and the extracted slice features respectively to obtain image features.

6. The method for predicting pCR based on DCE-MRI according to claim 1 or 5, It is characterized in that The pCR prediction model includes an image feature extraction module, a splicing module and a classifier, wherein the image feature extraction module is connected to the splicing module, and the splicing module is connected to the classifier; the image feature extraction module includes an imaging genomics feature extraction unit, a slice feature extraction unit and a feature screening module, wherein the imaging genomics feature extraction unit and the slice feature extraction unit are in parallel and are both connected to the feature screening unit; the input items of the splicing module include image features determined by the image feature extraction module and auxiliary information.

7. The method for predicting pCR based on DCE-MRI according to claim 1, It is characterized in that The determining of the target feature based on the image feature and the auxiliary information specifically includes: Obtaining an information feature vector corresponding to the auxiliary information; The information feature vector is concatenated with the image feature to obtain the target feature.

8. A prediction device for pCR based on DCE-MRI, It is characterized in that The DCE-MRI-based pCR prediction device specifically comprises: An acquisition module, used for acquiring a DCE-MRI set and auxiliary information corresponding to the DCE-MRI set, wherein the auxiliary information includes SER information and clinical information, and the DCE-MRI set carries a breast cancer lesion area; A control module is used to input the DCE-MRI set and the auxiliary information into a trained pCR prediction model, determine the image features corresponding to the DCE-MRI set through the pCR prediction model, determine the target features based on the image features and the auxiliary information, and predict the pCR category based on the target features.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the DCE-MRI-based pCR prediction method as described in any one of claims 1-7.

10. A terminal device, It is characterized in that include: Processor and memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the processor implements the steps in the method for predicting pCR based on DCE-MRI as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Breast cancer neoadjuvant chemotherapy curative effect prediction device based on multi-feature fusion

    CN114974575A

  • Method and system for cancer risk assessment using sound velocity and hardness of tissue

    CN115135230A

  • Breast cancer neoadjuvant chemotherapy pCR prediction method based on longitudinal DCE-MRI

    CN116504399A

  • Mammary gland benign and malignant lesion identification method based on mammary gland MRI dynamic graph parameter change

    CN117322866A