A novel method for automatic extraction of apparent diffusion coefficients in breast MRI
Paper in proceeding, 2011

Diffusion weighted (DW) MRI—and in particular the apparent diffusion coefficient (ADC)—shows potential for improving the characterization and classification of enhancing breast lesions identified using dynamic contrast-enhanced (DCE) MRI. Nevertheless, to date there does not exist a well defined and objective method for computing a representative ADC value for such lesions. Typically an average ADC is computed for a manually selected region of interest (ROI) [1]. This is problematic for two reasons. Firstly the choice of ROI is subjective. Differences in ROI selection between individuals, as well as the reproducibility of selection for a given individual, can lead to variation in the mean ADC. In addition ROIs are often defined to be circular or elliptical which imposes an arbitrary geometry on the ROI [2]. Secondly, given the heterogeneity in breast lesions, an ensemble average of ADC may not provide a truly representative value. It is assumed that a representative ADC will be present in the area of neovascularisation, as indicated by rapid contrast enhancement. In order to improve the objectivity, reproducibility and efficiency of representative ADC computation, we propose an automated method based on the selection of hypo-intense areas on the ADC map corresponding to regions of greatest initial contrast enhancement identified in the DCE-MRI data. We also present an evaluation of the method using routine clinical data.

Author

Darryl McClymont

Andrew Mehnert

Chalmers, Signals and Systems

Adnan Trakic

Dominic Kennedy

Stuart Crozier

Proc. Intl. Soc. Mag. Reson. Med. 19 (2011)

2014-

Areas of Advance

Life Science Engineering (2010-2018)

Subject Categories

Computer Vision and Robotics (Autonomous Systems)

Medical Image Processing

More information

Created

10/7/2017