A radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer

Abstract

The present invention provides a prediction model for the resection margin status of breast cancer conserving surgery based on radiomics, and belongs to the field of biotechnology. The construction method of the prediction model of the present invention comprises the following steps: (1) dividing breast cancer conserving surgery patients who meet the standards into a training set and a test set, and collecting breast MRI images of the breast cancer conserving surgery patients who meet the standards; (2) outlining the target lesions of the breast MRI on the T1 enhanced sequence of the MRI image, and extracting imaging features in the T1 enhanced sequence, T1 plain scan sequence and T2 plain scan sequence; (3) using the dimension reduction strategy of backward recursive feature elimination to reduce the dimension of the imaging features, and then using the XGBoost algorithm to construct a prediction model for the resection margin status of breast cancer conserving surgery based on radiomics. The prediction model of the present invention uses preoperative MRI images of patients undergoing breast conserving surgery and is established using the analysis method of radiomics. It can predict the resection margin status of the patient's breast conserving surgery and provide a corresponding reference for the formulation of the surgical plan.

Description

Technical Field

[0001] The present invention belongs to the field of biotechnology, and in particular relates to a radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer. Background Art

[0002] In my country, breast cancer is the cancer with the highest incidence and the sixth highest mortality rate among women. Globally, breast cancer accounts for about 24% of the number of new cancers in women worldwide each year, which is also the first among women. Among all new breast cancers, about 15.7% are clinical stage I, 44.9% are clinical stage II, and 18.7% are clinical stage III, all of which can be treated with surgery. Breast-conserving radical mastectomy (breast-conserving surgery) is one of the standard treatments for early breast cancer. Several large-scale clinical studies have shown that breast-conserving surgery plus standard radiotherapy has the same survival benefit as patients who undergo traditional mastectomy. Successful breast-conserving surgery must achieve negative surgical margins. Previous studies have shown that positive margins are significantly associated with an increase in local recurrence of breast cancer. Due to positive margins after breast-conserving surgery, about 20-30% of patients need to undergo secondary surgery, which may increase patient anxiety and increase surgical costs. The ASBS consensus proposes several methods to reduce the secondary surgery rate of breast-conserving surgery, including intraoperative freezing, fine needle aspiration, etc. However, none of these methods can predict the risk of positive margins before surgery. Accurately estimating the risk of positive margins in breast-conserving surgery before surgery is of reference significance for clinical decision-making. According to literature reports, several clinical prediction models have been developed to predict the probability of positive margins. However, most prediction models only contain clinical pathological variables, which are not strong predictors of margin status, and therefore have poor ability to predict margin status. In addition, the sample size of these studies is too small to draw convincing conclusions. Therefore, a reliable model for preoperative prediction of breast-conserving surgery margin status is needed.

[0003] Radiomics extracts high-throughput imaging features from medical images for quantitative research. In the field of breast cancer radiomics research, it has been applied to the diagnosis of breast cancer, prediction of axillary lymph node metastasis, prediction of treatment effect, and prediction of survival. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide a prediction model for the resection margin status of breast-conserving surgery for breast cancer based on radiomics.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a method for constructing a prediction model of the resection margin status of breast-conserving surgery for breast cancer based on radiomics, comprising the following steps:

[0006] (1) The breast cancer patients who met the criteria for breast-conserving surgery were divided into a training set and a test set, and breast MRI images of the breast cancer patients who met the criteria for breast-conserving surgery were collected;

[0007] (2) Outline the target lesions of breast MRI on the T1 enhanced sequence of MRI images, and extract imaging features on the T1 enhanced sequence, T1 plain scan sequence, and T2 plain scan sequence;

[0008] (3) The dimensionality reduction strategy of backward recursive feature elimination was used to reduce the dimensionality of imaging features, and then the XGBoost algorithm was used to construct a radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer.

[0009] The present invention uses preoperative MRI images of patients undergoing breast-conserving surgery and adopts the analysis method of radiomics to establish a prediction model to achieve the preoperative prediction of the resection margin status of breast-conserving surgery and provide corresponding reference for the formulation of surgical plan.

[0010] As a preferred implementation of the method for constructing a radiomics-based breast cancer breast-conserving surgery margin status prediction model of the present invention, it also includes step (5): testing the model using the test set to obtain the optimal parameters of the model.

[0011] As a preferred implementation of the method for constructing a radiomics-based breast cancer breast-conserving surgery margin status prediction model according to the present invention, after constructing the radiomics-based breast cancer breast-conserving surgery margin status prediction model, a validation set is used to verify the prediction results of the radiomics-based breast cancer breast-conserving surgery margin status prediction model.

[0012] As a preferred embodiment of the method for constructing a radiomics-based breast cancer breast-conserving surgery margin status prediction model of the present invention, the verification includes the area under the ROC curve AUC, sensitivity and specificity.

[0013] As a preferred embodiment of the method for constructing a radiomics-based breast cancer breast-conserving surgery margin status prediction model of the present invention, the target lesion includes the tumor parenchyma and the tumor margin in the range of 0 to 10 mm.

[0014] As a preferred implementation of the method for constructing a breast cancer conserving surgery margin status prediction model based on radiomics in the present invention, the imaging features include shape features (Shape Features); first-order features (First Order Features); grayscale co-occurrence matrix features (GLCM); grayscale size region matrix features (GLSZM); grayscale run length matrix features (GLRLM); adjacent grayscale hue difference matrix features (NGTDM); grayscale dependence matrix features (GLDM) and wavelet features (Wavelet-based Features).

[0015] As a preferred implementation of the method for constructing a prediction model for the resection margin status of breast cancer conserving surgery based on imaging genomics described in the present invention, the criteria include inclusion criteria and exclusion criteria; the inclusion criteria are pathologically diagnosed as invasive breast cancer, receiving breast conserving surgery for breast cancer (including patients who undergo breast resection surgery because they cannot achieve a negative resection margin after breast conserving surgery), and receiving a breast MRI examination within 1-2 weeks before surgery, and storing it in DICOM format, and the breast MRI includes a T1 enhanced sequence, a T1 plain scan sequence, and a T2 plain scan sequence; the exclusion criteria are receiving neoadjuvant chemotherapy, neoadjuvant endocrine therapy, preoperative anti-tumor therapy, bilateral breast cancer, multifocal / multi-center breast cancer, poor MRI imaging quality (such as motion artifacts, etc.), or the tumor has been excised and biopsied before MRI filming.

[0016] The present invention also provides a radiomics-based prediction model for the resection margin status of breast cancer conserving surgery constructed by the construction method.

[0017] The present invention also provides an application of the imaging genomics-based breast cancer breast-conserving surgery margin status prediction model in predicting the breast-conserving surgery margin status of breast cancer patients. The imaging genomics-based breast cancer breast-conserving surgery margin status prediction model provided by the present invention only needs to input the magnetic resonance imaging data of the breast cancer patient into the model to obtain the corresponding model score, and the breast-conserving surgery margin status of the breast cancer patient can be predicted according to the score, and the breast-conserving surgery strategy can be determined.

[0018] As a preferred implementation of the application of the present invention, when the model score of a breast cancer patient is less than 0.2 points, it is a low risk of positive margin; when the model score of a breast cancer patient is greater than 0.6 points, it is a high risk of positive margin.

[0019] The beneficial effects of the present invention are as follows: the present invention provides a prediction model for the resection margin status of breast cancer conserving surgery based on radiomics. The prediction model of the present invention uses preoperative MRI images of patients undergoing breast conserving surgery and is established using the analysis method of radiomics. The prediction model can predict the resection margin status of patients' breast conserving surgery and provide a corresponding reference for the formulation of surgical plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The figure shows the ROC curve of the radiomics-based breast cancer breast-conserving surgery margin status prediction model for the test set.

[0021] Figure 2 This is the ROC curve of the imaging omics-based breast cancer breast-conserving surgery margin status prediction model for validation set 1.

[0022] Figure 3 This is the ROC curve of the imaging omics-based breast cancer breast-conserving surgery margin status prediction model for validation set 2. DETAILED DESCRIPTION

[0023] In order to more concisely and clearly demonstrate the technical solution, purpose and advantages of the present invention, the technical solution of the present invention is described in detail below in conjunction with specific embodiments and drawings.

[0024] Example 1

[0025] The present embodiment provides a method for constructing a prediction model for the resection margin status of breast cancer conserving surgery based on radiomics, comprising the following steps:

[0026] (1) Baseline data of patients from multiple centers and their original breast MRI data were collected. A total of 576 patients met the inclusion and exclusion criteria, including 304 patients from Sun Yat-sen Memorial Hospital of Sun Yat-sen University, 131 patients from Sun Yat-sen University Cancer Center, and 141 patients from Tangshan People's Hospital. The inclusion and exclusion criteria include inclusion criteria and exclusion criteria. The inclusion criteria are (1) pathologically diagnosed with invasive breast cancer; (2) receiving breast-conserving surgery for breast cancer (including patients who underwent breast-conserving surgery but were unable to achieve negative margins and were switched to mastectomy); (3) receiving breast MRI within 1-2 weeks before surgery and stored in DICOM format. The breast MRI included T1 enhanced sequence, T1 plain scan sequence, and T2 plain scan sequence. Patients who met any of the following exclusion criteria were excluded: (1) those who had received preoperative anti-tumor treatment such as neoadjuvant chemotherapy or neoadjuvant endocrine therapy; (2) those with bilateral breast cancer; (3) those with multifocal / multicentric breast cancer; (4) those with poor MRI imaging quality (such as motion artifacts); and (5) those whose tumors had been excised and biopsied before MRI.

[0027] (2) The patients of Sun Yat-sen Memorial Hospital of Sun Yat-sen University were randomly divided into training set and test set according to the ratio of 7:3. The patients of Sun Yat-sen University Cancer Center were validation set 1, and the patients of Tangshan People's Hospital were validation set 2.

[0028] (3) Based on the magnetic resonance image data, 3D-Slicer software was used to delineate the target lesions of breast MRI on the T1 enhanced sequence and extract imaging features, where the target lesions include: tumor parenchyma and tumor margin 0-10mm range. The extracted imaging features include shape features; first order features; gray level co-occurrence matrix features (GLCM); gray level size region matrix features (GLSZM); gray level run length matrix features (GLRLM); neighboring gray level tone difference matrix features (NGTDM); gray level dependence matrix features (GLDM) and wavelet features (Wavelet-based Features).

[0029] (4) Statistical analysis was performed using R language. First, the dimension reduction strategy of backward recursive feature elimination was used for dimensionality reduction, and then the XGBoost algorithm was used to construct a radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer.

[0030] (5) The above test set was used to adjust the parameters of the radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer.

[0031] (6) The above validation sets 1 and 2 were used to validate the above imaging-based prediction model for breast cancer breast-conserving surgery margin status, draw the ROC curve, and calculate the corresponding AUC value.

[0032] The ROC curves of the training set, validation set 1, and validation set 2 of this embodiment are as follows: Figures 1 to 3 As shown in the figure, the AUCs of the training set, validation set 1, and validation set 2 were 0.79, 0.82, and 0.76, respectively, indicating that the radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer has good accuracy.

[0033] Example 2

[0034] The magnetic resonance imaging data of 576 patients were input into the imaging genomics-based breast cancer breast-conserving surgery margin prediction model of Example 1, and the model score was calculated. Statistical analysis shows that when the score is <0.2, the margin positive rate is <10%, and when the score is >0.6, the margin positive rate is >50%. Therefore, the patient's prediction score can be obtained according to the model to obtain the following breast-conserving surgery planning strategy: (1) When the score predicted by preoperative breast MRI is less than 0.2, the margin positive rate is very low, and the intraoperative freezing evaluation method can be omitted to reduce the operation time; (2) When the score predicted by preoperative breast MRI is greater than 0.6, the margin positive rate of conventional width is high, and the surgical margin width can be appropriately expanded to achieve negative margin. (3) When the score is between 0.2 and 0.6, the model's ability to predict the margin is insufficient, and the conventional breast-conserving surgery strategy can be used.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A method for constructing a prediction model for the resection margin status of breast-conserving surgery for breast cancer based on radiomics, characterized in that: The following steps are involved: (1) The breast cancer patients who met the criteria for breast-conserving surgery were divided into a training set and a test set, and breast MRI images of the breast cancer patients who met the criteria for breast-conserving surgery were collected; (2) Outline the target lesions of breast MRI on the T1 enhanced sequence of MRI images, and extract imaging features on the T1 enhanced sequence, T1 plain scan sequence, and T2 plain scan sequence; (3) The dimension reduction strategy of backward recursive feature elimination was used to reduce the dimension of imaging features, and then the XGBoost algorithm was used to build a radiomics-based prediction model for the resection margin status of breast-conserving surgery for breast cancer; The imaging features include shape features, first-order features, grayscale co-occurrence matrix features, grayscale size region matrix features, grayscale run length matrix features, adjacent grayscale tone difference matrix features, grayscale dependency matrix features and wavelet features; When the model score of breast cancer patients is less than 0.2 points, they are at low risk of positive margin; when the model score of breast cancer patients is greater than 0.6 points, they are at high risk of positive margin.

2. The method according to claim 1, characterized in that The method also includes step (5): using the test set to test the model to obtain the optimal parameters of the model.

3. The method according to claim 1, characterized in that After constructing a radiomics-based prediction model for breast cancer breast-conserving surgery resection margin status, a validation set was used to verify the prediction results of the radiomics-based prediction model for breast cancer breast-conserving surgery resection margin status.

4. The method according to claim 3, characterized in that The validation included area under the ROC curve AUC, sensitivity and specificity.

5. The method according to claim 1, characterized in that The target lesion includes the tumor substance and the tumor margin within a range of 0 to 10 mm.

6. The method according to claim 1, characterized in that The criteria include inclusion criteria and exclusion criteria; the inclusion criteria are pathological diagnosis of invasive breast cancer, breast-conserving surgery, and breast MRI examination within 1-2 weeks before surgery, and stored in DICOM format; the exclusion criteria are neoadjuvant chemotherapy, neoadjuvant endocrine therapy, preoperative anti-tumor treatment, bilateral breast cancer, multifocal / multicentric breast cancer, poor MRI imaging quality, or tumor resection biopsy before MRI.

7. A radiomics-based prediction model for the resection margin status of breast cancer conserving surgery constructed according to the method according to any one of claims 1 to 6.

8. Application of the radiomics-based breast cancer breast-conserving surgery margin status prediction model according to claim 7 in predicting the breast-conserving surgery margin status of breast cancer patients.

9. The use according to claim 8, characterized in that: When the model score of breast cancer patients is less than 0.2 points, they are at low risk of positive margin; when the model score of breast cancer patients is greater than 0.6 points, they are at high risk of positive margin.

Images