Yin Huang1, Deborah Yagow2,
Nathan Artz1,
1Medical Physics, University of
Wisconsin-Madison, Madison, WI, United States; 2Biomedical
Engineering, University of Wisconsin-Madison, Madison, WI, United States; 3Radiology,
University of Wisconsin-Madison, Madison, WI, United States; 4Nephrology,
University of Wisconsin-Madison, Madison, WI, United States
The
K-means segmentation method was implemented to automatically segment kidney
cortex and medulla on MR images of 24 subjects based on two kidney feature
values -- T1 and perfusion weighted information. Manual segmentation results
on the same subjects were used as reference and three similarity measures
were calculated to evaluate the effectiveness of K-means segmentation. The
segmentation time was radically shortened by K-means compared with manual
operation. However, there are about 30% of all subjects that K-means
segmentation did not work well so that a semi-automated strategy can be
suggested to incorporate manual segmentation when necessary.
Keywords