Deep Learning-Based Segmentation of Locally Advanced Breast Cancer on MRI in Relation to Residual Cancer Burden: A Multi-Institutional Cohort Study.

BACKGROUND: While several methods have been proposed for automated assessment of breast-cancer response to neoadjuvant chemotherapy on breast MRI, limited information is available about their performance across multiple institutions. PURPOSE: To assess the value and robustness of deep learning-derived volumes of locally advanced breast cancer (LABC) on MRI to infer the presence of residual disease after neoadjuvant chemotherapy. STUDY TYPE: Retrospective. SUBJECTS: Training cohort: 102 consecutive female patients with LABC scheduled for neoadjuvant chemotherapy (NAC) from a single institution (age: 25-73 years). Independent testing cohort: 55 consecutive female patients with LABC from four institutions (age: 25-72 years). FIELD STRENGTH/SEQUENCE: Training cohort: single vendor 1.5 T or 3.0 T. Testing cohort: multivendor 3.0 T. Gradient echo dynamic contrast-enhanced sequences. ASSESSMENT: A convolutional neural network (nnU-Net) was trained to segment LABC. Based on resulting tumor volumes, an extremely randomized tree model was trained to assess residual cancer burden (RCB)-0/I vs. RCB-II/III. An independent model was developed using functional tumor volume (FTV). Models were tested on an independent testing cohort and response assessment performance and robustness across multiple institutions were assessed. STATISTICAL TESTS: The receiver operating characteristic (ROC) was used to calculate the area under the ROC curve (AUC). DeLong's method was used to compare AUCs. Correlations were calculated using Pearson's method. P values <0.05 were considered significant. RESULTS: Automated segmentation resulted in a median (interquartile range [IQR]) Dice score of 0.87 (0.62-0.93), with similar volumetric measurements (R = 0.95, P < 0.05). Automated volumetric measurements were significantly correlated with FTV (R = 0.80). Tumor volume-derived from deep learning of DCE-MRI was associated with RCB, yielding an AUC of 0.76 to discriminate between RCB-0/I and RCB-II/III, performing similar to the FTV-based model (AUC = 0.77, P = 0.66). Performance was comparable across institutions (IQR AUC: 0.71-0.84). DATA CONCLUSION: Deep learning-based segmentation estimates changes in tumor load on DCE-MRI that are associated with RCB after NAC and is robust against variations between institutions. EVIDENCE LEVEL: 2. TECHNICAL EFFICACY: Stage 4.

Improving prediction of response to neoadjuvant treatment in patients with breast cancer by combining liquid biopsies with multiparametric MRI: protocol of the LIMA study - a multicentre prospective observational cohort study.

INTRODUCTION: The response to neoadjuvant chemotherapy (NAC) in breast cancer has important prognostic implications. Dynamic prediction of tumour regression by NAC may allow for adaption of the treatment plan before completion, or even before the start of treatment. Such predictions may help prevent overtreatment and related toxicity and correct for undertreatment with ineffective regimens. Current imaging methods are not able to fully predict the efficacy of NAC. To successfully improve response prediction, tumour biology and heterogeneity as well as treatment-induced changes have to be considered. In the LIMA study, multiparametric MRI will be combined with liquid biopsies. In addition to conventional clinical and pathological information, these methods may give complementary information at multiple time points during treatment. AIM: To combine multiparametric MRI and liquid biopsies in patients with breast cancer to predict residual cancer burden (RCB) after NAC, in adjunct to standard clinico-pathological information. Predictions will be made before the start of NAC, approximately halfway during treatment and after completion of NAC. METHODS: In this multicentre prospective observational study we aim to enrol 100 patients. Multiparametric MRI will be performed prior to NAC, approximately halfway and after completion of NAC. Liquid biopsies will be obtained immediately prior to every cycle of chemotherapy and after completion of NAC. The primary endpoint is RCB in the surgical resection specimen following NAC. Collected data will primarily be analysed using multivariable techniques such as penalised regression techniques. ETHICS AND DISSEMINATION: Medical Research Ethics Committee Utrecht has approved this study (NL67308.041.19). Informed consent will be obtained from each participant. All data are anonymised before publication. The findings of this study will be submitted to international peer-reviewed journals. TRIAL REGISTRATION NUMBER: NCT04223492.

Prognostic Value of Residual Disease after Neoadjuvant Therapy in HER2-Positive Breast Cancer Evaluated by Residual Cancer Burden, Neoadjuvant Response Index, and Neo-Bioscore.

PURPOSE: In breast cancer, pathologic complete response (pCR) to neoadjuvant systemic therapy (NST) is associated with favorable long-term outcome. Trastuzumab emtansine as additional adjuvant therapy improves recurrence-free survival of patients with HER2-positive breast cancer without pCR, but it is uncertain whether all patients without pCR need additional therapy. We evaluated the prognostic value of residual disease after trastuzumab-based NST in patients with HER2-positive breast cancer using Residual Cancer Burden (RCB), Neoadjuvant Response Index (NRI), and Neo-Bioscore. EXPERIMENTAL DESIGN: We included patients with stage II or III HER2-positive breast cancer treated with trastuzumab-based NST and surgery at The Netherlands Cancer Institute between 2004 and 2016. RCB, NRI, and Neo-Bioscore were determined. Primary endpoint was 5-year recurrence-free interval (RFI). A 3% difference compared with the pCR group was considered acceptable as noninferiority margin on the 5-year RFI estimate, based on a proportional hazards model, and its lower 95% confidence boundary. RESULTS: A total of 283 women were included. Median follow-up was 67 months (interquartile range 44-100). A total of 157 patients (56%) with pCR (breast and axilla) had a 5-year RFI of 92% (95% CI, 88-97); patients without pCR had a 5-year RFI of 80% (95% CI, 72-88). Patients with an RCB = 1 (N = 40, 15%), an NRI score between 0.75 and 0.99 (N = 30, 11%), or a Neo-Bioscore of 0 to 1 (without pCR; N = 28, 11%) have a 5-year RFI that falls within a predefined noninferiority margin of 3% compared with patients with pCR. CONCLUSIONS: The RCB, NRI, and Neo-Bioscore can identify patients with HER2-positive breast cancer with minimal residual disease (i.e., RCB = 1, NRI ≥ 0.75, or Neo-Bioscore = 0-1) after NST who have similar 5-year RFI compared with patients with pCR.