A prediction model training method for predicting relative efficacy of EC and T in NAC

Abstract

The application discloses a prediction model training method for predicting the relative curative effects of EC and T in NAC, and the method comprises the following steps: acquiring MRI image data and clinical pathological data of a subject to construct a data set, and constructing an imageomics model based on DCE and ADC sequences through image segmentation, feature extraction, qualitative relative curative effect feature, model construction and verification, and a mixed model combined with clinical pathological factors is used to predict the relative curative effects of EC and T treatment. The application can effectively predict the relative curative effects of EC (epirubicin plus cyclophosphamide) and T (taxane drugs) treatment schemes in the middle stage of NAC, so as to timely adjust the treatment scheme and improve the treatment effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of neoadjuvant chemotherapy, in particular to a prediction model training method for predicting the relative efficacy of EC and T in NAC. BACKGROUND

[0002] Neoadjuvant chemotherapy (NAC) has become an important part of breast cancer treatment, and its main purpose is to reduce tumor volume and improve the success rate of breast-conserving surgery. However, there are significant individual differences in patients' responses to the same treatment regimen during NAC, which is not only reflected among different patients, but also reflected in different treatment stages of the same patient. At the same time, complete tumor volume change rate data for evaluating treatment effect can only be obtained after NAC treatment is completed, but at this time the treatment regimen cannot be adjusted. Therefore, accurately predicting the relative efficacy of different treatment regimens in NAC is of great significance for timely adjusting treatment regimens and optimizing personalized treatment strategies.

[0003] In recent years, with the continuous development of imaging technology, especially the widespread application of magnetic resonance imaging (MRI) technology, more accurate tools have been provided for evaluating tumor response during NAC. MRI not only can observe the size and volume changes of tumors, but also can provide detailed information of tumors and their microenvironment through multi-parameter imaging technology. The radiomics method based on MRI can reveal the correlation between medical images and pathophysiological information by extracting quantitative information from medical images, providing a new way for optimizing clinical diagnosis and treatment strategies.

[0004] Currently, research on predicting the relative efficacy of different treatment regimens in NAC using radiomics is gradually deepening. Researchers have attempted to predict the relative efficacy of EC (epirubicin plus cyclophosphamide) and T (taxane) treatment regimens in the middle of NAC by constructing machine learning models based on MRI radiomics, in order to timely adjust the treatment regimen and improve the treatment effect. SUMMARY

[0005] This section aims to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0006] Therefore, the purpose of the present application is to provide a prediction model training method for predicting the relative efficacy of EC and T in NAC based on MRI radiomics and clinical pathology, in order to solve the problem that the existing NAC treatment regimen cannot predict the relative efficacy, so as to timely adjust the treatment method and affect the treatment effect.

[0007] To solve the above technical problems, according to one aspect of the present application, the present application provides the following technical solutions:

[0008] A prediction model training method for predicting the relative efficacy of EC and T in NAC, the steps are as follows:

[0009] S1, screening clinical data: obtaining image data and clinical pathological data of the subjects to construct a data set;

[0010] S2, image segmentation: selecting the second contrast phase image of the DCE sequence to segment in 6 dynamic phases, manually delineating the region of interest containing the entire MRI visible tumor on the axial plane layer by layer, then expanding the intratumoral region three-dimensionally isotropically by 5mm to obtain the peritumoral region;

[0011] S3, feature extraction: extracting image features from the image, standardizing the intensity distribution of the image before feature extraction, then extracting relevant features from each region of interest of the DCE sequence and the apparent diffusion coefficient map, and calculating the difference between the image feature values of the pre-NAC and mid-NAC images as the Delta image feature to capture the longitudinal changes of the tumor;

[0012] S4, qualitative relative efficacy feature: using the tumor volume change rate to represent the relative efficacy of the first 4 EC treatment cycles and the subsequent 4 T treatment cycles;

[0013] S5, constructing an image feature model: the model construction is divided into two parts: the first part evaluates the performance of the model using multiple sequences and multiple regions of interest through ablation experiments, and the second part integrates various image features to construct the Original model, the Delta model and the Fusion model to predict the relative efficacy of EC and T treatment using multiple sequences and multiple regions of interest;

[0014] S6, constructing a hybrid model: the output of the Original, Delta and Fusion image feature models of each patient is used as an image feature, and a hybrid model is established by combining the image feature with important clinical pathological factors using a logistic regression method;

[0015] S7, statistical analysis and evaluation: evaluating the performance of the image feature model and the hybrid model.

[0016] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S1, the image data is MRI image data, which includes image data of three key points of NAC-pre, NAC-mid and NAC-post, and the image data includes at least one positioning sequence, an axial TIRM sequence, DWI, a fat-suppressed DCE sequence and a sagittal fat-suppressed T2-weighted imaging.

[0017] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S1, the clinical pathological data includes gender, age, menstrual status, clinical T stage and clinical N stage, ER status, PR status, HER2 status and Ki67.

[0018] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S2, the 3D Slicer software is used to manually delineate the region of interest containing the entire MRI visible tumor layer by layer on the axial plane.

[0019] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S2, the SimpleITK package in Python is used to expand the intratumoral region three-dimensionally isotropically by 5mm.

[0020] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S3, PyRadiomics is used to extract radiomics features from the image.

[0021] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S4, the relative therapeutic effect is divided into two types: the first type is to determine the therapeutic effect of 4 EC treatment cycles by calculating the relative net change of tumor volume from NAC-pre to NAC-mid; the second type is to calculate the relative net change from NAC-mid to NAC-post to determine the therapeutic effect of 4 T treatment cycles, and then the relative therapeutic effect of treatment is evaluated by comparing these net change values.

[0022] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effect of EC and T in NAC, in step S5, multiple sequences and multiple regions of interest mainly use two MRI sequences: DCE and ADC, and two multiple regions of interest: intratumoral region and peritumoral region.

[0023] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effects of EC and T in NAC, in step S5, the feature selection comprises the following steps:

[0024] Features with variance greater than 1.0 are selected by variance thresholding;

[0025] Mann-Whitney U test is used to select features related to the relative therapeutic effect of NAC treatment;

[0026] Random forest model is used to sort the feature importance, and the top 100 most important features are selected;

[0027] Least absolute shrinkage and selection operator regression and 10-fold cross-validation are used to select features with non-zero coefficients.

[0028] As a preferred scheme of the prediction model training method for predicting the relative therapeutic effects of EC and T in NAC, in step S5, the Original model combines the imaging features of NAC-early and NAC-medium, the Delta model includes the Delta imaging features from NAC-early to NAC-medium, and the Fusion model combines all the features.

[0029] Compared with the prior art, the present application has the beneficial effects that: the present application constructs an imaging model based on DCE and ADC sequences based on MRI imaging, and a hybrid model combining clinical and pathological factors, finally obtains the FusionOriginal and Delta model of the change imaging features, and the Delta+clinicopath hybrid model combining clinical and pathological factors, and the relative therapeutic effects of EC and T in NAC are best in predicting EC (epirubicin plus cyclophosphamide) and T (paclitaxel) treatment plan. The method can effectively predict the relative therapeutic effects of EC (epirubicin plus cyclophosphamide) and T (paclitaxel) treatment plan in NAC, so as to timely adjust the treatment plan and improve the treatment effect. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical scheme of the embodiments of the present application, the present application will be described in detail below in combination with the drawings and detailed embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor. Among them:

[0031] Figure 1 The model training flowchart of the prediction model training method for predicting the relative therapeutic effects of EC and T in NAC based on MRI imaging and clinical pathology provided by the embodiments of the present application;

[0032] Figure 2 A patient recruitment flowchart provided for an embodiment of the present application;

[0033] Figure 3 A ROC curve graph of the Delta model in the training set provided for an embodiment of the present application;

[0034] Figure 4 A ROC curve of the Delta model in the test set provided for an embodiment of the present application;

[0035] Figure 5 A calibration curve of the Delta model in the training set provided for an embodiment of the present application;

[0036] Figure 6 A calibration curve of the Delta model in the test set provided for an embodiment of the present application;

[0037] Figure 7 A DCA curve of the Delta model in the training set provided for an embodiment of the present application;

[0038] Figure 8 A DCA curve of the Delta model in the test set provided for an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the above objectives, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0040] The present application provides a prediction model training method for predicting the relative efficacy of EC and T in NAC based on MRI imaging and clinical pathology, which solves the problem that the existing NAC treatment plan cannot predict the relative efficacy, so as to timely adjust the treatment method and affect the treatment effect, such as Figure 1 The specific scheme is as shown below:

[0041] Step 1: Screen patients containing complete MRI images and clinical and pathological data. As shown in the recruitment flowchart Figure 2 , female patients diagnosed with breast cancer from April 2016 to March 2023 were included, and the qualified standards were: (1) invasive breast cancer; (2) using T scheme after completing all EC cycles; (3) MRI data obtained before NAC treatment (NAC-Pre), before the fifth cycle (NAC-Mid) and before surgery (NAC-Post); (4) complete clinical and pathological data. The exclusion criteria are: (1) incomplete or non-standard NAC treatment; (2) insufficient MRI quality or lack of MRI data; (3) previous chemotherapy or targeted therapy; (4) distant metastatic lesions.

[0042] Clinical data of each patient before NAC treatment were collected, including gender, age, menopausal status, clinical T stage and clinical N stage. Patients were divided into training set and test set in the ratio of 7:3.

[0043] A total of 303 patients were excluded from this study, of which 292 had insufficient MRI data and 11 had insufficient image quality. Therefore, 190 patients were included in this study. Table 1 summarizes the baseline characteristics of all patients. In the training set, the proportion of patients with EC superior to T was 48.1% (64:133), and in the test set, it was 47.4% (27:57). In both groups of patients, there was no significant difference in the clinicopathological characteristics of patients with EC superior to T and T superior to EC (all p values > 0.05).

[0044] Table 1 - Baseline characteristics of patients

[0045]

[0046]

[0047] The clinicopathological and treatment strategies were as follows: all patients received 4 cycles of EC (epirubicin, 90 mg / m 2 , day 1, every 14 or 21 days; cyclophosphamide, 600 mg / m 2 , day 1, every 14 or 21 days), followed by 4 cycles of taxane treatment. Taxane treatment included albumin-paclitaxel (Abraxane, 260 mg / m 2 , day 1, every 14 days), solvent-based paclitaxel (paclitaxel, 175 mg / m 2 , day 1, every 14 days), and docetaxel, 75 mg / m 2, day 1, every 21 days) or liposomal (paclitaxel liposomal, 175 mg / m2, day 1). For patients with epidermal growth factor receptor 2 (HER2)+, trastuzumab (Herceptin, 6 mg / kg every 21 days, 8 mg / kg loading dose) was added starting at cycle 5. Prior to each NAC cycle, patients were monitored for complete blood count, liver function, renal function, and electrocardiogram. If a patient developed febrile neutropenia, grade 4 neutropenia, grade 4 thrombocytopenia, or grade 4 non-hematologic toxicity (excluding nausea, vomiting, and fatigue), the dose of the NAC regimen was reduced. Estrogen receptor (ER), progesterone receptor (PR), HER2, and Ki67 status were detected using immunohistochemistry (IHC). Tumors were classified as ER / PR positive if they showed 1% of nuclei stained cells. HER2 status was assessed as negative (HER2-), IHC score of 0 and 1+, positive (HER2+), IHC score of 3+. For tumors with an IHC score of 2+, HER2 gene amplification was detected using fluorescence in situ hybridization (FISH). Expression of Ki-67 was assessed using a cutoff index of 30%; expression below 30% was considered low, and ≥ 30% was considered high.

[0048] The MRI image acquisition procedure and protocol were as follows: MRI scans were performed in prone position using 1.5T or 3.0T magnetic resonance at three key points: pre-NAC, mid-NAC, and post-NAC. The protocol included at least one localizer sequence, axial TIRM sequence, DWI, fat-suppressed DCE sequence, and sagittal fat-suppressed T2-weighted imaging (T2WI). The dose of gadolinium-DTPA was 0.1 mmol / kg with an infusion rate of 3 mL / s, followed by a 20 mL saline flush. ADC maps were generated from DWI images using two b values.

[0049] 1.5T magnetic resonance imaging protocol included: (1) axial TIRM (TR / TE: 3450 / 61 ms; field of view: 340 mm x 340 mm; matrix: 320 x 320; flip angle: 80°; slice thickness: 4 mm); (2) axial DWI (b values: 50 / 800 s / mm 2 ; TR / TE: 5200 / 65 ms; field of view: 323 mm x 161 mm; slice thickness: 5 mm); (3) DCE sequence (TR / TE: 4.23 / 1.57 ms; field of view: 340 mm x 340 mm; matrix: 448 x 448; slice thickness: 1 mm; flip angle: 10°; pixel resolution: 0.8 x 0.8 x 1.0 mm 3; temporal resolution: 1 min), repeated 5 times before and after administration; (4) sagittal fat-suppressed T2WI (TR / TE: 2910 / 72 ms; field of view: 200 mm x 200 mm; matrix: 320 x 320; flip angle: 80°; slice thickness: 4 mm). Total scan time was 16 min 55 s.

[0050] 3.0T MRI, imaging protocol included: (1) axial TIRM (TR / TE: 5320 / 57 ms; field of view: 340 mm x 340 mm; matrix: 384 x 384; flip angle: 150°; slice thickness: 4 mm); (2) axial DWI (b values: 50 / 800 s / mm2; TR / TE: 7500 / 64 ms; field of view: 350 mm x 163 mm; matrix: 180 x 84; flip angle: 180°; slice thickness: 5 mm); (3) DCE sequence (TR / TE / TE: 3.90 / 1.66 ms; field of view: 360 mm x 360 mm; matrix: 320 x 320; flip angle: 10°; slice thickness: 1.5 mm; pixel resolution: 1.1 x 1.1 x 1.5 mm3; temporal resolution: 1 min), repeated 5 times before and after administration; (4) sagittal fat-suppressed T2WI (TR / TE: 4000 / 70 ms; field of view: 250 mm x 250 mm; matrix: 320 x 320; flip angle: 150°; slice thickness: 4 mm). Total scan time was 18 min 6 s. 2 ; temporal resolution: 1 min), repeated 5 times before and after administration; (4) sagittal fat-suppressed T2WI (TR / TE: 2910 / 72 ms; field of view: 200 mm x 200 mm; matrix: 320 x 320; flip angle: 80°; slice thickness: 4 mm). Total scan time was 16 min 55 s.

[0051] Step 2, MRI image segmentation and feature extraction: The second contrast- post phase image of the DCE sequence was selected for segmentation in 6 dynamic phases. The region of interest (ROI) containing the entire MRI visible tumor was manually delineated layer by layer on axial planes using 3D Slicer software. The intratumoral region was then expanded 5 mm in three dimensions isotropically using the SimpleITK package in Python 3.9.13 to obtain the peritumoral region. One radiologist with 5 years of experience in breast imaging segmented all cases. To ensure inter-observer consistency, another radiologist with 5 years of experience re-segmented 50 randomly selected cases. Both radiologists were blinded to clinical and histopathological data.

[0052] Radiomics features were extracted from the Original and Filtered images using PyRadiomics. The intensity distribution of the images was standardized before feature extraction. A total of 1218 features were extracted from each ROI of the DCE sequence and 1132 features from the ADC map. Considering the two ROIs (intratumoral and peritumoral regions), the imaging data from the pre-NAC and on-NAC images of each patient contributed a total of 9400 features from the DCE and ADC images. To capture the longitudinal changes of the tumor, we calculated the delta radiomics features as the difference between the radiomics feature values of the pre-NAC and on-NAC images. Therefore, a total of 14,100 radiomics features were generated for each patient.

[0053] Step 3, Integration of specific MRI sequences and ROI features, preliminary construction of multiple radiomics models: Radiomics models were mainly constructed using two MRI sequences (DCE and ADC) and two ROIs (intratumoral and peritumoral regions). To assess the inter-observer reproducibility, we calculated the intra-class correlation coefficient (ICC) for each radiomics feature. Only features with satisfactory inter-observer reproducibility, defined as ICC > 0.80, were retained in the model.

[0054] For each radiomics model, the following feature selection process was performed in four steps: (1) features with variance greater than 1.0 were selected by variance threshold; (2) features associated with relative efficacy of NAC treatment were selected using Mann-Whitney U test; (3) the top 100 most important features were selected using random forest model for feature importance ranking; (4) features with non-zero coefficients were selected using least absolute shrinkage and selection operator (LASSO) regression and 10-fold cross-validation.

[0055] The effectiveness of multi-sequence and multi-ROI models was verified by ablation experiments. Nine combined models were constructed by integrating the features of selected MRI sequences or ROIs. Each model included radiomics features from the pre-NAC and on-NAC MRI images.

[0056] Step 4, Optimal model selection based on performance index and image information for deepening construction: Table 2 provides the performance index of each radiomics model in the training set and test set. In the training set, the DCE+ADC-tumor+peri model produced the best prediction, with an AUC of 0.864 (95% CI: 0.792-0.911). In the test set, the model using the DCE+ADC sequence showed relatively high and more stable performance, with an AUC of 0.663 (95% CI: 0.549-0.767) for the DCE+ADC-tumor model, 0.598 (95% CI: 0.502-0.690) for the DCE+ADC-peri model, and 0.585 (95% CI: 0.480-0.677) for the DCE+ADC-tumor+peri model. To obtain more information, the DCE+ADC-tumor+peri model, which is a multi-sequence and multi-region model, was selected for further radiomics model construction.

[0057] Table 2 - Radiomics first construction - performance index table

[0058]

[0059]

[0060] Step 5, Deepening construction of radiomics models (Original, Delta, Fusion): Three models were constructed using the combination of multi-sequence and multi-ROI features in different aspects of radiomics. The Original model combined the radiomics features of NAC-early and NAC-mid. The Delta model included the Delta radiomics features from the change from NAC-early to NAC-mid. The Fusion model combined all the above features.

[0061] The radiomics model used the XGBoost algorithm based on the selected radiomics features. The grid search method and five-fold cross-validation were used to determine the optimal hyperparameters of the model. 4-folds (80% of patients) were used to train the model, and the rest (20% of patients) were used to select the optimal hyperparameters.

[0062] Table 3 summarizes the AUC, accuracy, sensitivity, specificity, PPV, and NPV for each radiomics model in the training and test sets. The Delta model had an AUC of 0.887 (95% CI: 0.816–0.930) on the training set and 0.757 (95% CI: 0.683–0.817) on the test set. Compared to the Delta model, the aggregation model did not show improved performance, with an AUC of 0.887 (95% CI: 0.822–0.931) on the training set and 0.749 (95% CI: 0.644–0.837) on the test set.

[0063] Table 3 - Performance Indicators for Deepening Radiomics Construction (Original, Delta, Fusion)

[0064]

[0065]

[0066] The ROC curves of the Delta model in the training and test sets are as follows: Figure 3 and Figure 4 As shown. The calibration curve of the Delta model demonstrates good agreement between the predictions and observations of the two groups. Figure 5 and Figure 6 The Hosmer-Lemeshow fit test results were not significant (p = 0.343 on the training set and p = 0.464 on the test set), indicating that there was no deviation from the perfect fit.

[0067] For the Delta model, decision curve analysis (DCA) shows that when the threshold probability is between 0 and 0.95 ( Figure 7 ) and the tests focused on values ​​between 0 and 0.71. Figure 8 When the net clinical benefit is greater than 0, the benefit is greater than 0.

[0068] Step 6: Constructing a hybrid model (radiomics + clinicopathology) and clinicopathology model: The outputs of the Original, Delta, and Fusion radiomics models for each patient were used as radiomics features. To assess the correlation between clinicopathology variables and the relative efficacy of NAC, we employed univariate logistic regression analysis. Important variables were selected using backward stepwise regression based on the Akaike Information Criterion (AIC).

[0069] A logistic regression method was used to combine radiomic features with important clinicopathological factors to build a hybrid model. In addition, a clinicopathological model containing only clinicopathological variables was also built for comparison. To select the optimal hyperparameters of the model, a grid search method combined with five-fold cross-validation was used. Four folds (80% of patients) were used to train the model, and the remaining fold (20% of patients) was used to select the best hyperparameters. The predictive ability of the model was visually demonstrated by ROC curve, calibration curve, and decision curve.

[0070] Univariate analysis showed that the Original radiomic score (OR, 2756.20 [95% CI: 273.82-39542.24]; P < 0.001), Delta radiomic score (OR, 469.92 [95% CI: 84.13-3442.97]; P < 0.001), and Fusion radiomic score (OR, 743.78 [95% CI: 118.71-6508.12]; P < 0.001) were significantly associated with the relative efficacy of NAC in the training set, as shown in Table 5. Multivariate analysis showed that, in addition to the Original radiomic score, menstrual status, clinical T stage, and HER2 status were identified as independent predictors of the relative efficacy of NAC. For the Delta radiomic score, clinical T stage was identified as an independent predictor of the relative efficacy of NAC (Table 6). Similarly, clinical T stage also independently predicted the relative efficacy of NAC on the Fusion radiomic score.

[0071] A hybrid model was constructed using these variables to predict the relative efficacy of NAC. Table 4 provides a summary of the AUC, accuracy, sensitivity, specificity, PPV, and NPV for each hybrid model in the training and test sets. After incorporating important clinicopathological features into the radiomic model, the performance of the Delta+clinicopath model continued to outperform the Original+clinicopath and Fusion+clinicopath models, with an AUC of 0.893 (95% CI: 0.877-0.899) in the training set and 0.780 (95% CI: 0.747-0.799) in the training set. Figure 3 and Figure 4 The ROC curves of the most effective hybrid model (Delta+clinicopath) and the clinicopathological model only are shown.

[0072] Table 4 - Hybrid model construction (radiomic + clinicopath, single clinicopath)

[0073]

[0074] Table 5 - Clinicopathological features

[0075]

[0076]

[0077] Table 6 - Delta independent predictors

[0078]

[0079] For the optimal mixed model, calibration curve analysis showed reasonable agreement between predicted probabilities of relative effectiveness of NAC and actual outcomes in both the training and test sets. The Hosmer-Lemeshow test was not significant for both sets of results, with p-values of 0.592 for the training set and 0.295 for the test set, indicating no significant deviation from a perfect model fit. DCA showed that the optimal mixed model provided significant clinical net benefit at all threshold probabilities in the training set (as shown in Figure 7 ) and between 0 and 0.74 in the test set (as shown in Figure 8 ).

[0080] Step 7, Evaluate individual model performance metrics: Chi-square tests were used to compare clinical pathologic features of different groups or cohorts of patients using categorical variables. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC). Sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) were calculated for both the training and test sets. The 95% confidence intervals (CIs) for each metric were calculated using the bootstrap method of 1000 intervals. The optimal cutoff value for the radiation group score in the training set was determined by maximizing the Youden index, and these fixed cutoff values were then applied to the test set. The DeLong test was used to compare AUCs between models. All statistical tests were two-sided, with P-values < 0.05 considered statistically significant. All statistical analyses were performed using R 4.2.3 or Python 3.9.13.

[0081] Compared with the radiomics model, the corresponding hybrid model showed a certain degree of AUC improvement, especially the Delta+clinicopath model, with an AUC of 0.893 (95% CI: 0.877-0.899) in the training set and 0.780 (95% CI: 0.747-0.799) in the test set. In addition, the Delta radiomics model and the Delta+clinicopath model showed good calibration and superior clinical utility in DCA and calibration curve analysis. These results confirmed that the multi-sequence, multi-region MRI radiomics model is a reliable and reproducible relative efficacy tool for predicting four cycles of T treatment after four cycles of EC treatment. In summary, using the prediction model training method for predicting the relative efficacy of EC and T in NAC based on MRI radiomics and clinicopathology provided in the embodiment has the following technical effects: the prediction model training method for predicting the relative efficacy of EC and T in NAC based on MRI radiomics and clinicopathology provided in the embodiment: the imaging model based on DCE and ADC sequences and the hybrid model combined with clinicopathological factors are constructed, and finally the Delta model fusing the original and change radiomics features and the Delta+clinicopath hybrid model combined with the clinicopathological factors are obtained, which perform best in predicting the relative efficacy of EC and T treatment; this method can effectively predict the relative efficacy of EC (epirubicin plus cyclophosphamide) and T (taxane drugs) treatment in NAC, so as to timely adjust the treatment plan and improve the treatment effect.

[0082] It should be understood that the above-mentioned technical means can be implemented by hardware, software, or the integration of the two. Therefore, the method and device of the present application, as well as its constituent elements or aspects, can be built into a physical medium, such as a floppy disk, CD-ROM, hard disk, or any other machine-readable storage medium carrying program code (i.e. a series of instructions). When this program is loaded into a device such as a computer and executed, the device becomes a tool for implementing the present application.

[0083] Although the present application has been described above with reference to the embodiments, various modifications can be made thereto and equivalents can be substituted for elements thereof without departing from the scope of the present application. In particular, features of the disclosed embodiments can be combined together in any manner, provided that there is no structural conflict. The combinations of these features are not exhaustively described in the specification only for the purpose of omitting the length and saving resources. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A prediction model training method for predicting the relative efficacy of EC and T in NAC, characterized by, The steps are as follows: S1, screening clinical data: obtaining image data and clinical pathological data of the subjects to construct a data set; S2, image segmentation: selecting the second contrast phase image of the DCE sequence to segment in 6 dynamic phases, manually delineating the region of interest containing the entire MRI visible tumor on the axial plane layer by layer, then expanding the intratumoral region three-dimensionally isotropically by 5mm to obtain the peritumoral region; S3, feature extraction: extracting imaging features from the image, standardizing the intensity distribution of the image before feature extraction, then extracting relevant features from each region of interest of the DCE sequence and the apparent diffusion coefficient map, and calculating the difference between the imaging feature values of the NAC-pre and NAC-mid images as the Delta imaging feature to capture the longitudinal changes of the tumor; S4, qualitative relative efficacy features: using tumor volume change rate to represent the relative efficacy of the first 4 EC treatment cycles and the subsequent 4 T treatment cycles; S5, constructing an imaging model: the model construction is divided into two parts: the first part evaluates the performance of the model using multiple sequences and multiple regions of interest through ablation experiments, and the second part integrates various imaging features to construct the Original model, the Delta model and the Fusion model to predict the relative efficacy of EC and T treatment using multiple sequences and multiple regions of interest; Multiple sequences and multiple regions of interest mainly use two MRI sequences: DCE and ADC and two multiple regions of interest: intratumoral region and peritumoral region, the feature selection steps are as follows: select features with variance greater than 1.0 by variance threshold; select features related to the relative efficacy of NAC treatment by Mann-Whitney U test; select the top 100 most important features by random forest model; select features with non-zero coefficients by least absolute shrinkage and selection operator regression and 10-fold cross-validation; S6, constructing a hybrid model: the output of the Original, Delta and Fusion imaging model of each patient is used as an imaging feature, and a hybrid model is established by combining the imaging feature with important clinical and pathological factors using a logistic regression method; S7, statistical analysis and evaluation: evaluate the performance of the imaging model and the hybrid model. 2.The method of claim 1, wherein the method is characterized by, In step S1, the image data is MRI image data, including image data at three key points of NAC-pre, NAC-mid and NAC-post, which includes at least one positioning sequence, axial TIRM sequence, DWI, fat-suppressed DCE sequence and sagittal fat-suppressed T2 weighted imaging. 3.The method of claim 1, wherein the method is characterized by, In step S1, the clinical pathological data includes gender, age, menstrual status, clinical T stage and clinical N stage, ER status, PR status, HER2 status and Ki67. 4.The method of claim 1, wherein, In step S2, the 3D Slicer software is used to manually delineate the region of interest containing the entire MRI visible tumor on the axial plane layer by layer.

5. The method of claim 1, wherein the method is used to predict the relative efficacy of EC and T in NAC. In step S2, the SimpleITK package in Python is used to expand the intratumoral region three-dimensionally isotropically by 5mm.

6. The method of claim 1, wherein the method is used to train a predictive model for predicting the relative efficacy of EC and T in NAC. In step S3, radiomics features are extracted from the images using PyRadiomics.

7. The method of claim 1, wherein the method is used to train a predictive model for predicting the relative efficacy of EC and T in NAC. In step S4, the relative efficacy is divided into two types: the first type is to determine the efficacy of the 4 EC treatment cycles by calculating the relative net change of tumor volume from NAC-pre to NAC-mid; the second type is to calculate the relative net change from NAC-mid to NAC-post to determine the efficacy of the 4 T treatment cycles, and then the relative efficacy of the treatment is evaluated by comparing these net change values. 8.The method of claim 1, wherein the method is characterized by, In step S5, the Original model combines the radiomics features of NAC-pre and NAC-mid, the Delta model includes the Delta radiomics features from NAC-pre to NAC-mid, and the Fusion model combines all the above features.

Images