Meeting Banner
Abstract #1732

Quantitative Background Parenchymal Enhancement Estimation on Breast DCE-MRI by Measuring Relative Voxel-Wise Enhancement

Shandong Wu1, Susan P. Weinstein2, Emily F. Conant2, Despina Kontos1

1Radiology, University of Pennsylvania, Philadelphia, PA, United States; 2Radiology, Hospital of the University of Pennsylvania, Philadelphia, PA, United States

Studies suggest that background parenchymal enhancement (BPE) in breast DCE-MRI is associated with an increased breast cancer risk for high-risk populations. Most previous work on BPE estimation is however, based on readers qualitative assessment. The purpose of our study is to develop a fully automated computational method for quantitative BPE estimation and optimize the related parameters to identify the BPE. We estimate BPE through identifying the enhancing voxels in the DCE-MRI images by measuring the relative voxel-wise enhancement. Our algorithm achieves high agreement with manual segmentation and could be used to standardize measures of BPE in clinical applications.

Abstract PDF Poster PDF

Keywords

absolute achieved advised agent agreement alignment although amount applications apply assessment atlas audience automated background belonging best bilateral breast cancer categories category chest clinical coarse coded coefficient color computational compute computed consensus consider considering context contours contrast correlation correlations criteria currently custom define definition density derived determine determined develop developed difficult enables enhance enhancement enhancing equal established estimation examining example experienced extraction full fully gold greater hence highest identified identify identifying injection intensity interactive kind king lead makes manual manually masks measures measuring morris much oncologists optimize optimized oral parameterization percentage population populations post presentation previous previously principal probabilistic processed qualitative quantitative radiologist radiologists radiology rads randomly rather readers registration related reporting respectively rigid risk scientists screening segment segmentation segmented selected selecting series slices software specifically standardize studies subjective subtraction suggest suppressed system target third threshold tissue translated unaffected underway useful validated validation varying view volume wall whole wise women year