A major site for metastases in cancer patients is the liver. Due to incoherent motion; diffusion-weighted images of liver are best obtained with echoplanar acquisitions. However, these data suffer from low SNR, fuzzy boundaries due to partial volume effects and anisotropic motion, and poor contrast. In addition, motion occurs between acquisitions at different b-values, precluding direct registration and pixel-by-pixel calculations of apparent diffusion coefficients. The focus of this work was to enable automated segmentation and registration of lesions from echo planar images of the liver.

We developed algorithms for automatic lesion segmentation, and tracking deformations and translations associated with these lesions.

Here is an example of a diffusion-weighted image of the liver, where the lesion is the bright region on the left side of the image:

This work was a collaborative effort with Prof. Robert J. Gillies (Dept. of Biochemistry and Arizona Cancer Center).