Meeting Banner
Abstract #2562

Performance of an Automated Segmentation Algorithm for MR Renography

Artem Mikheev1, Jeff L. Zhang2, Tariq Gill1, Marta Heilbrun2, Stella Kang1, Hersh Chandarana1, Henry Rusinek1, Vivian S. Lee2

1Radiology, NYU School of Medicine, New York, NY, United States; 2Radiology, University of Utah School of Medicine, Salt Lake City, UT, United States


A key prerequisite for analysis of MR renography (MRR) data is the ability to segment MRI images. We have developed and validated a new semi-automated renal segmentation technique based on edge-constrained region growing. The segmentation error is 7.6 6.5 cm3 and the interobserver disparity 5.4 4.5 cm3, a significant improvement over graph-cut method. The new algorithm achieves a ten-fold improvement in user processing time, from >20 min to 2.1 0.7 min per kidney. With expedited image processing, MRR has the potential to expand our knowledge of renal function and to help diagnose different types of renal insufficiency.