Meeting Banner
Abstract #1845

Automated Selection of Hypointense Regions in Diffusion-Weighted Breast MRI

Darryl McClymont1, Andrew Mehnert2, Adnan Trakic1, Dominic Kennedy3, Stuart Crozier1

1University of Queensland, Brisbane, QLD, Australia; 2Chalmers University of Technology, Sweden; 3Queensland X-Ray, Australia

Recent research suggests that diffusion-weighted MRI can be used to improve the sensitivity and specificity of dynamic contrast-enhanced MRI for the detection of breast cancer. However, to date the methods proposed for determining a representative ADC value for a suspicious lesion are highly varied. We propose an automated method based on the converging squares algorithm, which is a noise-robust minimum finding technique. We present an evaluation of the method for computing a representative ADC. The method is also compared to ensemble averaging of ADC values over the entire lesion and the selection of the global minimum ADC value.

Keywords

administered anal another apparent appears applying approval assessed automated averaging avoid avoided benign best better bias bottom boundary breast cancer characteristic chosen clinical coefficient combination committee complicated compute computing confirmed constrained contrast converging cubic curve curves date defined defining degree delineation detection determining diffusion directional discriminating discrimination discriminatory dynamic earliest enhanced ensemble entire ethics evaluation exam examinations example examples exist experienced explanation farthest finding global greatly growing heterogeneous highest highly homepage human identify improve included influence inherent inherently interpolation involve last latter lesion lesions linear located malignant manually maps mask match materials medical middle necrosis nevertheless next noise objective offers operating originate overlaid particular pattern performance pixel placing power practice problem propose proposed radiographer receiver recent representative resolution robustness routine scanner seems segmentation segmented selection sensitivity slice smaller smallest spacing spatial specificity squares subjectivity suffers suggests suspicious systems taking technology terms tolerant tool trans typically varied women worst years yield