Fully Automated Fibroglandular Tissue Segmentation and Bias Correction in Breast MR Images Using Level Set Method

  • Mehri Owjimehr University of Calgary
  • Elise Fear University of Calgary


In this paper an image processing method is proposed for bias correction and fibroglandular tissue segmentation from Magnetic Resonance Images (MRI) of human breast. The proposed method is based on level sets and includes three steps. In the pre-processing step, a chest wall line detection method is applied to separate the chest wall from the breast region in the MR images of breast. In the next step, a new level set algorithm is employed to estimate the bias field. The bias field estimation is used for intensity inhomogeneity correction, which leads to an efficient segmentation of the fibroglandular tissue. Finally, in the post-processing step, the skin layer is detected using morphological operations, and the fibroglandular tissue is extracted after skin layer subtraction. The proposed method has been validated on 2D images of an MR scan of the human breast. The implementation results show efficient performance of this method in tissue segmentation of MR images with the presence of intensity inhomogeneity.