Edge intensity normalization as a bias field correction during balloon snake segmentation of breast MRI

Paper in proceeding, 2008

Segmentation of fat suppressed dynamic contrast enhanced MRI (DCE-MRI) image data can pose significant problems because of the inherently poor signal-to-noise ratio (SNR) and intensity variations due to the bias field. Segmentation methods such as balloon snakes, while able to operate in a poor SNR environment, are sensitive to variations in edge intensity, which are regularly encountered within DCE-MRI due to the bias field. In order to overcome the effects of the bias field, an intensity normalization based on the strength of the strongest edge, i.e. the skin-air-boundary, is proposed and evaluated. This normalization allows balloon segmentations to be run three times faster while maintaining, or even improving accuracy.