Computerized Interpretation of Dynamic Breast MRI

Magnetic resonance images are often corrupted byintensity inhomogeneity, which manifests itself as slow intensityvariations of the same tissue over the image domain. Suchshading artifacts must be corrected before doing computerizedanalysis such as intensity-based segmentation and quantitativeanalysis. In this paper, we present a fuzzy c-means (FCM)based algorithm that simultaneously estimates the shading effectwhile segmenting the image. A multiplier field term that modelsthe intensity variation is incorporated into the FCM objectivefunction which is minimized iteratively. In each iteration, thebias field is estimated based on the current tissue class centroidsand the membership values of the voxels and then smoothedby an iterative low-pass filter. The efficacy of the algorithm isdemonstrated on clinical breast MR images.