Meeting Banner
Abstract #1763

Prediction of Neoadjuvant Chemotherapy Response of Breast Cancer with Changes of MR Perfusion and Diffusion Characteristics in Early Chemotherapy by Using Neural Network Algorithm

Huiyan Shao1, Li Guo2, Xiaoying Wang2, Rui Li1, Chun Yuan, 13, Huijun Chen1

1Tsinghua University, Beijing, China; 2Peking University First Hospital, Beijing, China; 3University of Washington, Seattle, WA, United States

Neoadjuvant Chemotherapy (NAC), locally enforced before the surgery, plays a significant role in the multimodality therapy of breast cancer. To predict whether a patient will have a complete response after the early NAC in-vivo is critical for the therapy plan. Diffusion-weighted MRI (DW-MRI) and dynamic contrast-enhanced MRI (DCE-MRI) have been reported great potential to predict NAC responder. However, most studies used the diffusion or perfusion imaging alone for prediction. In this study, we employed neural network and logistic regression to predict the final treatment response by using the changes of DW-MRI and DCE-MRI parameters after the early NAC cycles.

Abstract PDF Poster PDF

Keywords

according acquisition alone among another apparent approval array automatic axial better bias bilateral biopsy bolus bottom breast calibrate cancer cell cells characteristics chemotherapy china chose clinging coil combine combining complete completely contrast critical curve cycle cycles defined density determined differentiate diffusion done drawn dynamic early eliminate enforced enhanced enhancement entire ethics every exam examination example excite exponential extract facilitate final firstly function gradients great histological histology in vivo included independent informed injected injection institutional intensities intensity introduced kinetics lesions locally logistic manually mastectomy materials matrix medical microscopically model models monitoring mono multivariate muscle nearly network neural operation outcome park patient patients percentage perfusion personalized physical plays population positioning predication predict prediction process protocol radiology recruited reduction reflects regression respectively responder responders response ribs role samples segmentation select selected semi sensitivity sensitizing significantly slice slices spacing specificity specimen studies subsequent surgery system table therapy thresholding together treatment tumor typical underwent utilize wash wherein yuan