Dataset Information

A machine learning model that classifies breast cancer pathologic complete response on MRI post-neoadjuvant chemotherapy.

ABSTRACT: BACKGROUND:For breast cancer patients undergoing neoadjuvant chemotherapy (NAC), pathologic complete response (pCR; no invasive or in situ) cannot be assessed non-invasively so all patients undergo surgery. The aim of our study was to develop and validate a radiomics classifier that classifies breast cancer pCR post-NAC on MRI prior to surgery. METHODS:This retrospective study included women treated with NAC for breast cancer from 2014 to 2016 with (1) pre- and post-NAC breast MRI and (2) post-NAC surgical pathology report assessing response. Automated radiomics analysis of pre- and post-NAC breast MRI involved image segmentation, radiomics feature extraction, feature pre-filtering, and classifier building through recursive feature elimination random forest (RFE-RF) machine learning. The RFE-RF classifier was trained with nested five-fold cross-validation using (a) radiomics only (model 1) and (b) radiomics and molecular subtype (model 2). Class imbalance was addressed using the synthetic minority oversampling technique. RESULTS:Two hundred seventy-three women with 278 invasive breast cancers were included; the training set consisted of 222 cancers (61 pCR, 161 no-pCR; mean age 51.8?years, SD 11.8), and the independent test set consisted of 56 cancers (13 pCR, 43 no-pCR; mean age 51.3?years, SD 11.8). There was no significant difference in pCR or molecular subtype between the training and test sets. Model 1 achieved a cross-validation AUROC of 0.72 (95% CI 0.64, 0.79) and a similarly accurate (P?=?0.1) AUROC of 0.83 (95% CI 0.71, 0.94) in both the training and test sets. Model 2 achieved a cross-validation AUROC of 0.80 (95% CI 0.72, 0.87) and a similar (P?=?0.9) AUROC of 0.78 (95% CI 0.62, 0.94) in both the training and test sets. CONCLUSIONS:This study validated a radiomics classifier combining radiomics with molecular subtypes that accurately classifies pCR on MRI post-NAC.

SUBMITTER: Sutton EJ

PROVIDER: S-EPMC7254668 | biostudies-literature | 2020 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A machine learning model that classifies breast cancer pathologic complete response on MRI post-neoadjuvant chemotherapy.

Sutton Elizabeth J EJ Onishi Natsuko N Fehr Duc A DA Dashevsky Brittany Z BZ Sadinski Meredith M Pinker Katja K Martinez Danny F DF Brogi Edi E Braunstein Lior L Razavi Pedram P El-Tamer Mahmoud M Sacchini Virgilio V Deasy Joseph O JO Morris Elizabeth A EA Veeraraghavan Harini H

Breast cancer research : BCR 20200528 1

<h4>Background</h4>For breast cancer patients undergoing neoadjuvant chemotherapy (NAC), pathologic complete response (pCR; no invasive or in situ) cannot be assessed non-invasively so all patients undergo surgery. The aim of our study was to develop and validate a radiomics classifier that classifies breast cancer pCR post-NAC on MRI prior to surgery.<h4>Methods</h4>This retrospective study included women treated with NAC for breast cancer from 2014 to 2016 with (1) pre- and post-NAC breast MRI ...[more]

PMID: 32466777

Similar Datasets

Project description:BackgroundTo investigate the association of axillary pathologic complete response (pCR) rate among breast cancer patients with pCR after neoadjuvant chemotherapy (NCT).MethodsThe retrospective clinical data of 1,903 patients who were treated with NCT between March, 2010 and December, 2018, were collected from one Chinese database and analyzed. The correlations between clinicopathological characteristics and breast pCR with axillary pCR were calculated by χ2 test. Binary logistic regression analysis was used for multivariate analysis. The relative risk of positive axillary nodes after NCT in patients with and without breast pCR was analyzed using a Cochran-Mantel-Haenszel (CMH) test stratified by initial N stage and tumor subtype.ResultsThe rate of axillary pCR was increased in the cases with initial cN0, Ki67 high expression, HR+HER2+/HR-HER2+/TN subtypes, and breast pCR. After NCT, the relative risk of nodal disease burden was 4.81 in patients without breast pCR compared with patients with breast pCR. The relative risk of positive nodal status in patients with cN0, cN1, cN2, and cN3 disease without vs. with breast pCR was 6.45, 4.88, 5.69 and 6.24, respectively. The relative risk of positive nodal status in patients with HR+HER2-, HR+HER2+, HR-HER2+, and TN disease was 4.02, 4.50, 3.82 and 4.18, respectively. Of cN0 patients with breast pCR, only 4 out of 44 (9%) with HER2-positive disease had 1 or 2 axillary lymph node metastases at final surgical pathology compared to 30 out of 98 (31%) of those without breast pCR. There was no evidence of positive nodal residue among all 21 patients (100%) with TN disease compared to 65% (36 of 55) of patients without breast pCR.ConclusionsNodal status is strongly correlated with breast pCR after NCT. Patients with initial cN0/1 TN/HER2 positive disease who achieve breast pCR at surgery have a low risk of nodal metastasis. These results suggest that the failure rate of missing positive lymph nodes among those patients was very low and that it is safe for such patients to undergo sentinel lymph node biopsy (SLNB) after NCT. This study also provides a theoretical basis for clinical trials focused on the avoidance of axillary surgery in selected patients.

Project description:BackgroundMagnetic resonance imaging (MRI) has been proposed to guide breast cancer surgery by measuring residual tumour after neoadjuvant chemotherapy. This study-level meta-analysis examines MRI's agreement with pathology, compares MRI with alternative tests and investigates consistency between different measures of agreement.MethodsA systematic literature search was undertaken. Mean differences (MDs) in tumour size between MRI or comparator tests and pathology were pooled by assuming a fixed effect. Limits of agreement (LOA) were estimated from a pooled variance by assuming equal variance of the differences across studies.ResultsData were extracted from 19 studies (958 patients). The pooled MD between MRI and pathology from six studies was 0.1 cm (95% LOA: -4.2 to 4.4 cm). Similar overestimation for MRI (MD: 0.1 cm) and ultrasound (US) (MD: 0.1 cm) was observed, with comparable LOA (two studies). Overestimation was lower for MRI (MD: 0.1 cm) than mammography (MD: 0.4 cm; two studies). Overestimation by MRI (MD: 0.1 cm) was smaller than underestimation by clinical examination (MD: -0.3 cm). The LOA for mammography and clinical examination were wider than that for MRI. Percentage agreement between MRI and pathology was greater than that of comparator tests (six studies). The range of Pearson's/Spearman's correlations was wide (0.21-0.92; 16 studies). Inconsistencies between MDs, percentage agreement and correlations were common.ConclusionMagnetic resonance imaging appears to slightly overestimate pathologic size, but measurement errors may be large enough to be clinically significant. Comparable performance by US was observed, but agreement with pathology was poorer for mammography and clinical examination. Percentage agreement can provide supplementary information to MDs and LOA, but Pearson's/Spearman's correlation does not provide evidence of agreement and should be avoided. Further comparisons of MRI and other tests using the recommended methods are warranted.

Dataset Information

A machine learning model that classifies breast cancer pathologic complete response on MRI post-neoadjuvant chemotherapy.

Publications

A machine learning model that classifies breast cancer pathologic complete response on MRI post-neoadjuvant chemotherapy.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure