AUTHOR=Chen Zhigeng , Huang Manxia , Lyu Jianbo , Qi Xin , He Fengtai , Li Xiang 

TITLE=Machine learning for predicting breast-conserving surgery candidates after neoadjuvant chemotherapy based on DCE-MRI

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1174843

DOI=10.3389/fonc.2023.1174843

ISSN=2234-943X

ABSTRACT=This research is aims to investigate the machine learning method for predicting breastconserving surgery (BCS) candidates, from the patients who accepted neoadjuvant chemotherapy (NAC), by using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) obtained before and after NAC.The research retrospectively included 75 patients who underwent NAC and breast surgery. First, 3390 features were comprehensively exacted from pre-and post-NAC DCE-MRI. Then, the patients were divided into two groups: type 1 was the patients with pathologic complete response (pCR), single lesion shrinkage; type 2 was major residual lesion with satellite foci, multifocal residual, stable disease (SD), and progressive disease (PD). The logistic regression (LR) was used for building prediction models to identify the two groups. The prediction performance was assessed by using the area under the curve (AUC), accuracy, sensitivity, and specificity. Results: The radiomics features were significantly related to the shrinkage of breast cancer after NAC. The Ccombination model achieved an AUC of 0.82, and the pre-NAC model was 0.64, the post-NAC model was 0.70, and the pre-post-NAC model was 0.80. In the combination model, 15 features, including 9 wavelet-based features, 4 Laplacian-of-Gauss (LoG) features, and 2 original features, were filtered out. Among these selected features, 4 features from pre-NAC DCE-MRI, 6 features from post-NAC DCE-MRI, and 4 features were pre-post-NAC features.The model combined with pre-and post-NAC DCE-MRI can effectively predict the candidates to undergo BCS and hereby provide AI-based decision support for clinicians with ensured safety. High-order (LoG and wavelet-based) features play an important role in our machine learning model. The features from pre-NAC DCE-MRI have better predictive performance.