Construction and validation of a pretreatment DCE-MRI-based radiomics prediction model for axillary pathologic response after neoadjuvant therapy in node-positive breast cancer

Introduction

Neoadjuvant therapy (NAT) has increasingly been used in breast cancer, as it may downstage disease and increase the likelihood of breast and axillary conservation (1). Approximately 40–60% of patients with initial axillary lymph node (ALN) metastasis achieve an axillary pathologic complete response (pCR) after NAT (2,3). Nevertheless, many patients with initially ALN-positive breast cancer still undergo axillary lymph node dissection (ALND) after NAT to determine residual nodal burden (4), despite the risk of complications such as lymphedema or arm paresthesia. Sentinel lymph node biopsy (SLNB) has been considered an alternative strategy for patients who convert to clinically node-negative status after NAT (5). However, its use in initially node-positive patients remains a concern because prospective trials reported false-negative rates of 12.6–14.2% (2,6), raising concern that missed nodal metastases may compromise staging accuracy and subsequent treatment decisions. Therefore, the optimal selection of candidates for axillary de-escalation after NAT remains controversial, and accurate prediction of ALN response is clinically important.

Imaging-based assessment of axillary response after NAT has been extensively investigated, but it remains insufficiently accurate for reliable axillary treatment de-escalation (7). Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is the most accurate modality for evaluating the tumor response (8), but post-NAT MRI achieved an overall accuracy of only 58%, with low sensitivity for residual nodal disease (9). Against this background, increasing attention has been directed toward pretreatment imaging biomarkers. By characterizing the tumor before therapy-induced morphologic distortion occurs, baseline MRI may provide biologically relevant information on treatment sensitivity (10-12). Recent studies have shown promising results for predicting axillary response using pretreatment MRI (13,14). In particular, radiomics models based on pretreatment MRI have demonstrated encouraging performance for predicting axillary positive-node response after NAT (14). However, previous studies have relied on conventional imaging descriptors or single-phase radiomics, which may not fully capture dynamic heterogeneity throughout the entire contrast-enhancement process.

To address these challenges, we hypothesized that the phase-to-phase variability of radiomics features across the full pretreatment DCE-MRI could better reflect treatment-relevant spatiotemporal heterogeneity. Therefore, we aimed to develop and validate a radiomics-based model incorporating intratumoral and peritumoral feature variability across multiple DCE-MRI phases for predicting axillary pCR after NAT in initially ALN-positive breast cancer. We present this article in accordance with the TRIPOD reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0122/rc).

Methods

Patients and clinical data collection

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital (No. Ek2021235), and individual consent for this retrospective analysis was waived. Between January 2017 and December 2019, a total of 175 female patients with initially ALN-positive breast cancers treated at Tianjin Medical University Cancer Institute and Hospital were enrolled. The inclusion criteria were as follows: (I) pathologically confirmed breast cancer with ALN metastasis; (II) completion of at least four cycles of NAT followed by ALND; (III) pretreatment DCE-MRI performed within 1 month before NAT with sufficient image quality; and (IV) complete clinicopathologic information. All 175 patients were randomly divided into a primary cohort (n=140) and an internal validation cohort (n=35) at a ratio of 8:2 (Figure 1).

Figure 1 Flowchart of the study population. ALN, axillary lymph node; ALND, ALN dissection; MRI, magnetic resonance imaging; pCR, pathologic complete response.

Initial clinical T stage (cT) and clinical N stage (cN) according to the 8th TNM staging system of the American Joint Committee on Cancer (AJCC) were included in this study.

ALN response to NAT and histopathologic assessment

Axillary pCR was defined as the absence of micrometastasis and macrometastasis in ALNs on postoperative histopathology (ypN0). Isolated tumor cells were classified as ypN0 (i+) and were therefore considered non-pCR. Histologic type, lymphovascular invasion (LVI), estrogen receptor (ER) status, progesterone receptor (PR) status, Human epidermal growth factor receptor 2 (HER2) status, and Ki67 expression were recorded. ER and PR positivity were defined as nuclear staining in at least 1% of tumor cells on immunohistochemistry (IHC). HER2 positivity was defined as an IHC score of 3+. Cases with an equivocal HER2 result (IHC score of 2+) underwent confirmation by fluorescence in situ hybridization.

MRI acquisition and image evaluation

All MRI examinations were performed using a 3.0 T scanner (Discovery MR750, GE Healthcare). Detailed MRI acquisition parameters are provided in the Appendix 1.

All MRI examinations were independently reviewed by two radiologists (Y.L. and H.L., with 8 and 18 years of experience in breast MRI, respectively) who were blinded to the response to NAT. Disagreements were resolved by consensus. The following MRI features were assessed according to the 2013 Breast Imaging Reporting and Data System (BI-RADS^®) MRI lexicon of the American College of Radiology: largest tumor diameter, necrosis, peritumoral edema, maximum diameter of the abnormal ALN, and perinodal infiltration. An abnormal ALN was defined as having a cortical thickness greater than 3.5 mm or loss of the fatty hilum (15).

MRI preprocessing and tumor segmentation

Image preprocessing included N4 bias field correction, resampling voxels of 1×1×1 mm using B-spline interpolation, and standardization using z score normalization. The region of interest (ROI) covering the entire tumor was manually drawn on the first postcontrast images with ITK-SNAP software (http://www.itksnap.org) by a radiologist (Y.L., with 8 years of breast MRI experience). The peritumoral ROI was generated by isotropic 3-mm dilation using standard image morphological dilation operations (Figure 2). The tumor and peritumoral segmentations were then duplicated to the precontrast and remaining four postcontrast images.

Figure 2 Radiomics analysis workflow. HER2: 0= negative, 1= positive; perinodal infiltration: 0= absent, 1= present. ALN, axillary lymph node; AUC, area under the curve; cN stage, clinical node stage; DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; LASSO, least absolute shrinkage and selection operator; pCR, pathologic complete response; R-score, radiomics signature.

Radiomics feature analysis

For each phase of DCE images, a total of 2,990 radiomics features (1,502 from the tumor and 1,488 from the peritumoral region), including shape, first-order, and texture features (Table S1), were extracted from the origin and nine other types of derived images using the PyRadiomics package in Python (version 3.7, https://www.python.org). Spatiotemporal radiomics features were then calculated as the variance of each radiomics feature across the six DCE phases (Figure 2). These radiomics features were standardized by the z score method and selected in three steps in the primary cohort. First, the Mann-Whitney U test was used to select the features that were significantly different between patients with ALN pCR and non-pCR, with a significance threshold of P<0.05. Second, Spearman correlation analysis was used to remove highly correlated features, and features with a correlation coefficient greater than 0.9 were excluded. Third, least absolute shrinkage and selection operator (LASSO) was used to select the optimized features. The radiomics signature (R-score) was developed based on the selected features. The radiomics workflow is shown in Figure 2.

Development and validation of the prediction model

The clinical, histopathologic and MRI variables were first evaluated by univariate analyses. The factors with a P value of <0.05, the R-score, and eligible clinical predictive variables based on the experience of oncologists were included in backward stepwise multivariable logistic regression analysis with Akaike’s information criterion (AIC) to develop a clinical model. In addition, a final model combining the R-score and other predictive factors was established. The receiver operating characteristic (ROC) curve of each model was generated, and the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated to evaluate the discrimination performance. Difference in AUC between models was compared by DeLong’s test. Calibration was assessed with the calibration curve and the calibration slope. Decision curve analysis (DCA) was performed to evaluate the clinical net benefit of the models.

Statistical analysis

Statistical analyses were performed with R software (version 4.0.5). Continuous variables were compared using the two-sample t test or Mann-Whitney U test, as appropriate, and categorical variables were compared using the chi-square test or Fisher’s exact test. A two-tailed P value of <0.05 was considered significant.

Results

Patient characteristics

Among the 175 patients with initially ALN-positive breast cancer included in this study, 58 patients achieved ALN pCR after NAT. The baseline clinical, histopathologic and MRI characteristics of the patients in the primary cohort and validation cohorts are shown in Table 1. No significant differences were observed between the two cohorts.

In the primary cohort, patients who achieved ALN pCR were more likely ER-negative, PR-negative, and HER2-positive, and to have higher Ki67 levels than those who did not achieve ALN pCR. In addition, the absence of perinodal infiltration was significantly associated with ALN pCR. A comparison of the clinical, histopathologic, and MRI features between patients with and without ALN pCR in the primary cohort is presented in Table 2.

Radiomics feature selection and radiomics signature development

In the primary cohort, 111 features were retained after Mann-Whitney U test and Spearman correlation analysis from the initial set of 2,990 radiomics features. LASSO regression was then used for further feature selection, and 27 radiomics features were ultimately identified to construct the radiomics signature for predicting ALN pCR (Figure 3). The radiomics feature selection process is shown in Figure S1, and the details of the R-score calculation are available in the Appendix 2. The radiomics model achieved AUCs of 0.84 [95% confidence interval (CI): 0.76–0.91] and 0.82 (95% CI: 0.68–0.96) in the primary and validation cohorts, respectively (Figure 4).

Figure 3 Radiomics feature selection and radiomics signature construction using the LASSO regression model. (A) Tuning parameter log (λ) selection in the LASSO model used 10-fold cross-validation via minimum criteria. The vertical black dashed line represents the lowest mean-square error. (B) The LASSO coefficient profiles plot. The vertical black dashed line was drawn, and the optimal λ resulted in twenty-seven non-zero features. (C) The twenty-seven most optimal radiomics features to build the radiomics signature. LASSO, least absolute shrinkage and selection operator.

Figure 4 The ROC curve of clinical, radiomic and combined models in the primary cohort (A) and validation cohorts (B). ROC, receiver operating characteristic.

Development and validation of the clinical and combined models

Multivariate logistic regression analysis identified R-score, perinodal infiltration, cN stage, ER status, and HER2 status as independent predictors of ALN pCR (Table 3). A combined nomogram was then constructed based on these predictors (Figure 5). The combined model achieved AUCs of 0.90 (95% CI: 0.84–0.96) and 0.87 (95% CI: 0.75–0.99) in the primary and validation cohorts, respectively, indicating good predictive performance (Figure 4). In the validation cohort, the combined model achieved a sensitivity of 78.6%, specificity of 85.7%, PPV of 78.6%, NPV of 85.7%, and accuracy of 82.9%. The diagnosis performance of the clinical and radiomics models in the primary and validation cohorts is also summarized in Table 4.

Figure 5 Development and validation of the combined model. (A) Nomogram for predicting pathologic complete response of axillary lymph node after neoadjuvant therapy. (B) Decision curve analysis for the clinical, radiomics, and combined model. The x-axis shows the probability threshold, as the y-axis measures the net benefit of patients. The largest area under the decision curve of the combined model (yellow line) indicates the highest net benefit than the clinical (green line) and radiomics model (blue line). Calibration curves of the three models in the primary (C) and validation cohort (D). ALN, axillary lymph node; cN stage, clinical node stage; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; pCR, pathologic complete response; R-score, radiomics signature.

The clinical model, which included perinodal infiltration, cN stage, ER status, and HER2 status, yielded AUCs of 0.80 (95% CI: 0.72–0.87) and 0.74 (95% CI: 0.57–0.91) in the primary and validation cohorts, respectively (Figure 4, Table 4). According to DeLong’s test, the AUCs of the combined model were significantly higher than those of the clinical model in the primary cohort (P=0.003) and validation cohort (P =0.046), whereas the AUC of the combined model was significantly higher than that of the radiomics model only in the primary cohort (P=0.01).

The calibration curves of the clinical model, radiomics model, and combined nomogram for predicting ALN pCR in the primary and validation cohorts are shown in Figure 5C,5D. The bootstrap-corrected calibration slope of the combined model was 0.87 (95% CI: 0.48–1.33), indicating acceptable calibration. Decision curve analysis showed that the combined model provided greater net benefit than the clinical and radiomics models (Figure 5B). Figure 6 presents examples of patients with ALN pCR and without ALN pCR.

Figure 6 MR images and 3D visualization of two women with axillary pCR (top) and without axillary pCR (bottom). Images from left to right are respectively: showed DCE image (A,F), ROI of tumor (B,G), ROI of peritumoral region (C,H), 3D ROI visualization of tumor (D,I), and 3D ROI visualization of peritumoral region (E,J) of patient with axillary pCR. 3D, three-dimensional; ALN, axillary lymph node; DCE, dynamic contrast-enhanced; MR, magnetic resonance; pCR, pathologic complete response; ROI, region of interest.

Discussion

In this retrospective study, we developed and internally validated radiomics-based models integrating spatiotemporal information of pretreatment DCE-MRI for predicting of ALN pCR after NAT in initially ALN-positive breast cancer. The combined model, which incorporated clinical factors and a radiomics signature derived from intratumoral and peritumoral feature variability, showed the best overall performance. Notably, because the positive outcome in our study was axillary pCR, the PPV most directly reflects the reliability of identifying patients who truly achieved axillary pCR; in this regard, the combined model yielded PPVs of 76.6% and 78.6% in the primary and validation cohorts, respectively, higher than the clinical model. These results suggest that the model may support preoperative risk stratification of axillary response, although external validation is still required before clinical application.

Previous studies have shown that the conventional DCE-MRI findings of breast tumors may help predict ALN response to NAT, supporting the concept that axillary metastases and primary breast tumors are biologically related (16,17). A previous study also suggested that radiomics signatures integrating features from the breast tumor and the axillary area could predict the ALN response (18). However, radiomics features derived from a single postcontrast image may overlook spatiotemporal heterogeneity throughout the entire enhancement process. In this study, we showed that pretreatment DCE-MRI-based spatiotemporal radiomic features could predict ALN pCR after NAT. The combined model achieved AUCs of 0.90 and 0.87 in the primary and validation cohorts, respectively, which were numerically higher than those reported in several previous studies (AUC range, 0.75–0.84) (16,19-21). Notably, PPV most directly reflects the reliability of identifying patients who truly achieved axillary pCR; the relatively high PPVs observed in both cohorts suggest that the model may be useful for preoperative risk stratification of axillary response.

Many studies have showed that peritumoral radiomic characteristics of breast cancer are associated with treatment outcome (14,22-24). Braman et al. (25) first demonstrated that combining tumoral and peritumoral radiomics features from DCE-MRI could predict the response to NAT. A recent study further showed the potential value of peritumoral textural feature analysis from DCE-MRI for predicting ALN metastasis (26). Consistent with these findings, our results suggest that integrating intratumoral and peritumoral features may improve prediction of ALN pCR, supporting a complementary role of these two regions. Notably, seven of the 10 most discriminative features in the radiomics signature were derived from the peritumoral region, suggesting that peritumoral intensity and texture heterogeneity may provide additional information relevant to treatment response. Together, these findings support the value of combining intratumoral and peritumoral information in radiomics-based prediction.

In this study, all patients underwent ALND after NAT, which minimized the possibility of residual metastases remaining in non-excised ALNs. We found that ER-negative status, HER2-positive status, and lower initial clinical N stage were associated with higher ALN pCR rates, consistent with previous studies (15,20). We also found that pre-NAT perinodal infiltration was associated with a lower ALN pCR rate. Although the underlying biological mechanism remains unclear, perinodal infiltration may reflect the invasiveness of lymph vessels (27) and influence the distribution of therapeutic agents.

The clinical model, which yielded AUCs of 0.80 and 0.74 in the primary and validation cohorts, respectively, showed only moderate predictive performance. This finding suggests that conventional MRI findings alone may be insufficient for accurate prediction and may also be influenced by observer experience. In addition, ER and HER2 status determined by core biopsy may be affected by intratumoral heterogeneity.

Some limitations of this study should be acknowledged. First, this was a single-center retrospective study and was therefore subject to unavoidable selection bias. Although feature reduction was performed before multivariable modeling, the limited sample size and lack of external validation may still have introduced model optimism. Larger external validation cohorts are needed in future studies to further validate the present model. Second, ALN response to NAT may differ across molecular subtypes, although all patients received standard NAT regimens. Third, quantitative radiomics features of ALNs were not included because of potential axillary imaging artifacts and the difficulty of achieving one-to-one correspondence between MRI-visible ALNs and pathologic findings. In future studies, clip placement after biopsy may help localize metastatic ALNs more precisely and allow further evaluation of ALN-based imaging features. Finally, only pretreatment DCE-MRI was included. MRI obtained during NAT may provide additional biological information and further improve prediction of pCR, although a pretreatment-only strategy may reduce the need for repeated MRI examinations.

Conclusions

In conclusion, pretreatment DCE-MRI-based radiomics features derived from intratumoral and peritumoral regions showed promising performance for predicting ALN response to NAT. The proposed nomogram may support pre-treatment risk stratification of axillary response in initially ALN-positive breast cancer, although further external validation is required before clinical application.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0122/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0122/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0122/prf

Funding: This work was supported by the National Natural Science Foundation of China (Nos. 82172025 and 82302302), the Science & Technology Development Fund of Tianjin Education Commission for Higher Education (No. 2024KJ189), and Tianjin Key Medical Discipline Construction Project (No. TJYXZDXK-3-003A).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0122/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital (No. Ek2021235), and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Korde LA, Somerfield MR, Carey LA, et al. Neoadjuvant Chemotherapy, Endocrine Therapy, and Targeted Therapy for Breast Cancer: ASCO Guideline. J Clin Oncol 2021;39:1485-505. [Crossref] [PubMed]
2. Boughey JC, Suman VJ, Mittendorf EA, et al. Sentinel lymph node surgery after neoadjuvant chemotherapy in patients with node-positive breast cancer: the ACOSOG Z1071 (Alliance) clinical trial. JAMA 2013;310:1455-61. [Crossref] [PubMed]
3. Samiei S, Simons JM, Engelen SME, et al. Axillary Pathologic Complete Response After Neoadjuvant Systemic Therapy by Breast Cancer Subtype in Patients With Initially Clinically Node-Positive Disease: A Systematic Review and Meta-analysis. JAMA Surg 2021;156:e210891. [Crossref] [PubMed]
4. Gasparri ML, de Boniface J, Poortmans P, et al. Axillary surgery after neoadjuvant therapy in initially node-positive breast cancer: international EUBREAST survey. Br J Surg 2022;109:857-63. [Crossref] [PubMed]
5. Heidinger M, Weber WP. Axillary Surgery for Breast Cancer in 2024. Cancers (Basel) 2024;16:1623. [Crossref] [PubMed]
6. Kuehn T, Bauerfeind I, Fehm T, et al. Sentinel-lymph-node biopsy in patients with breast cancer before and after neoadjuvant chemotherapy (SENTINA): a prospective, multicentre cohort study. Lancet Oncol 2013;14:609-18. [Crossref] [PubMed]
7. Samiei S, de Mooij CM, Lobbes MBI, et al. Diagnostic Performance of Noninvasive Imaging for Assessment of Axillary Response After Neoadjuvant Systemic Therapy in Clinically Node-positive Breast Cancer: A Systematic Review and Meta-analysis. Ann Surg 2021;273:694-700. [Crossref] [PubMed]
8. Sheikhbahaei S, Trahan TJ, Xiao J, et al. FDG-PET/CT and MRI for Evaluation of Pathologic Response to Neoadjuvant Chemotherapy in Patients With Breast Cancer: A Meta-Analysis of Diagnostic Accuracy Studies. Oncologist 2016;21:931-9. [Crossref] [PubMed]
9. Moo TA, Jochelson MS, Zabor EC, et al. Is Clinical Exam of the Axilla Sufficient to Select Node-Positive Patients Who Downstage After NAC for SLNB? A Comparison of the Accuracy of Clinical Exam Versus MRI. Ann Surg Oncol 2019;26:4238-43.
10. Li Y, Chen Y, Zhao R, et al. Development and validation of a nomogram based on pretreatment dynamic contrast-enhanced MRI for the prediction of pathologic response after neoadjuvant chemotherapy for triple-negative breast cancer. Eur Radiol 2022;32:1676-87. [Crossref] [PubMed]
11. Lo Gullo R, Eskreis-Winkler S, Morris EA, et al. Machine learning with multiparametric magnetic resonance imaging of the breast for early prediction of response to neoadjuvant chemotherapy. Breast 2020;49:115-22. [Crossref] [PubMed]
12. Granzier RWY, van Nijnatten TJA, Woodruff HC, et al. Exploring breast cancer response prediction to neoadjuvant systemic therapy using MRI-based radiomics: A systematic review. Eur J Radiol 2019;121:108736. [Crossref] [PubMed]
13. Malhaire C, Umay O, Cockenpot V, et al. Predicting axillary residual disease after neoadjuvant therapy in breast cancer using baseline MRI and ultrasound. Eur Radiol 2025;35:4896-909. [Crossref] [PubMed]
14. Lin Y, Wang J, Li M, et al. Prediction of breast cancer and axillary positive-node response to neoadjuvant chemotherapy based on multi-parametric magnetic resonance imaging radiomics models. Breast 2024;76:103737. [Crossref] [PubMed]
15. Kim R, Chang JM, Lee HB, et al. Predicting Axillary Response to Neoadjuvant Chemotherapy: Breast MRI and US in Patients with Node-Positive Breast Cancer. Radiology 2019;293:49-57. [Crossref] [PubMed]
16. Kantor O, Sipsy LM, Yao K, et al. A Predictive Model for Axillary Node Pathologic Complete Response after Neoadjuvant Chemotherapy for Breast Cancer. Ann Surg Oncol 2018;25:1304-11. [Crossref] [PubMed]
17. Weiss A, Campbell J, Ballman KV, et al. Factors Associated with Nodal Pathologic Complete Response Among Breast Cancer Patients Treated with Neoadjuvant Chemotherapy: Results of CALGB 40601 (HER2+) and 40603 (Triple-Negative) (Alliance). Ann Surg Oncol 2021;28:5960-71. [Crossref] [PubMed]
18. Gan L, Ma M, Liu Y, et al. A Clinical-Radiomics Model for Predicting Axillary Pathologic Complete Response in Breast Cancer With Axillary Lymph Node Metastases. Front Oncol 2021;11:786346. [Crossref] [PubMed]
19. Guo R, Su Y, Si J, et al. A nomogram for predicting axillary pathologic complete response in hormone receptor-positive breast cancer with cytologically proven axillary lymph node metastases. Cancer 2020;126:3819-29. [Crossref] [PubMed]
20. Vila J, Mittendorf EA, Farante G, et al. Nomograms for Predicting Axillary Response to Neoadjuvant Chemotherapy in Clinically Node-Positive Patients with Breast Cancer. Ann Surg Oncol 2016;23:3501-9. [Crossref] [PubMed]
21. Ouldamer L, Chas M, Arbion F, et al. Risk scoring system for predicting axillary response after neoadjuvant chemotherapy in initially node-positive women with breast cancer. Surg Oncol 2018;27:158-65. [Crossref] [PubMed]
22. Braman N, Prasanna P, Whitney J, et al. Association of Peritumoral Radiomics With Tumor Biology and Pathologic Response to Preoperative Targeted Therapy for HER2 (ERBB2)-Positive Breast Cancer. JAMA Netw Open 2019;2:e192561. [Crossref] [PubMed]
23. Niu S, Jiang W, Zhao N, et al. Intra- and peritumoral radiomics on assessment of breast cancer molecular subtypes based on mammography and MRI. J Cancer Res Clin Oncol 2022;148:97-106. [Crossref] [PubMed]
24. Xu H, Liu J, Chen Z, et al. Intratumoral and peritumoral radiomics based on dynamic contrast-enhanced MRI for preoperative prediction of intraductal component in invasive breast cancer. Eur Radiol 2022;32:4845-56. [Crossref] [PubMed]
25. Braman NM, Etesami M, Prasanna P, et al. Intratumoral and peritumoral radiomics for the pretreatment prediction of pathological complete response to neoadjuvant chemotherapy based on breast DCE-MRI. Breast Cancer Res 2017;19:57. [Crossref] [PubMed]
26. Zhan C, Hu Y, Wang X, et al. Prediction of Axillary Lymph Node Metastasis in Breast Cancer using Intra-peritumoral Textural Transition Analysis based on Dynamic Contrast-enhanced Magnetic Resonance Imaging. Acad Radiol 2022;29:S107-15. [Crossref] [PubMed]
27. Kim WH, Kim HJ, Park CS, et al. Axillary Nodal Burden Assessed with Pretreatment Breast MRI Is Associated with Failed Sentinel Lymph Node Identification after Neoadjuvant Chemotherapy for Breast Cancer. Radiology 2020;295:275-82. [Crossref] [PubMed]