Meeting Banner
Abstract #1466

Multivariate Analysis of DCE-MRI for Early Prediction of Breast Tumor Response Using Machine Learning

Xia Li1, Subramani Mani1, Lori R. Arlinghaus1, A. Bapsi chakravarthy1, E. Brian Welch1, Thomas E. Yankeelov1

1Vanderbilt University Institute of Imaging Science, Nashville, TN, United States

Dynamic contrast enhanced MRI (DCE-MRI) can offer information related to tumor perfusion and permeability, vascular volume, extravascular extracellular volume fraction, and the intracellular water lifetime of a water molecule. There have been many efforts employing DCE-MRI as a surrogate biomarker for predicting the response of breast tumors to neoadjuvant chemotherapy. However, most studies perform univariate analysis on these parameters. In this study, we perform multivariate analysis using machine learning methods. The preliminary results demonstrate the feasibility of using DCE-MRI data and machine learning for predicting the response of breast tumors to a single cycle of neoadjuvant chemotherapy.

Abstract PDF Poster PDF

Keywords

accuracy acquisition active appears application appropriate approved assess assessing basis best better breast cancer chemotherapy classification clinical collected combination complete compute computed considered contrast cycle defined deviation differently dynamic early efforts employed employing enhanced enrolled example exchange extended extracellular fast feasibility fraction function functions future gradient improve include injected inputs institute intracellular invasive kernel kernels lead leading learning leave lifetime linear lists logistic machine many matrix model models molecule multivariate note offer optimizing particular pathologic patient patients perfusion permeability physiological precision predict predicting prediction predictors preliminary primary quadratic radial regime regression related residual respectively responders response scanner science selection sensitive sensitivity serial significantly site slice specificity spoiled stage studies support surgery surrogate table therapy third training trans treatment trial tumor tumors univariate variable vascular vector volume water yielded