Meeting Banner
Abstract #1551

The Influence of Image Registration and Segmentation Error on Functional MR Renography

Christopher C. Conlin1, 2, Jeff L. Zhang, 23, Marta E. Heilbrun, 23, Henry Rusinek4, Artem V. Mikheev4, Vivian S. Lee, 23

1Department of Bioengineering, University of Utah, Salt Lake City, UT, United States; 2Utah Center for Advanced Imaging Research, Salt Lake City, UT, United States; 3Department of Radiology, University of Utah, Salt Lake City, UT, United States; 4Department of Radiology, New York University, New York, United States

This investigation employed Monte Carlo simulations and a patient study to examine the effects of image registration and segmentation error on the accuracy of renal functional parameter estimates as measured with dynamic Gd-MRI. Image registration and segmentation is performed as the first step in quantitative analysis of the dynamic images. Imperfect registration and segmentation may lead to errors in the estimated single kidney glomerular filtration rate (skGFR). It was found that GFR is a robust functional renal parameter that is resistant to errors introduced during registration and segmentation of Gd-MRI and that clinically acceptable GFR estimates can be obtained quickly.

Abstract PDF

Keywords

accuracy added adding affects aortic arterial assessment audience averaging bilateral bioengineering build city clinically clinicians compartment compartments concentration consisted constrained constraints convolving cortex cortical counterparts curve curves decreased deviation deviations diseased displaying distributed distributions dynamic dysfunction dysfunctional effectively enhanced environment equal error errors established even examine example extent extreme female filtration fitted fitting fractions full function functional functions gamma generated generating greater healthier healthy henry ideal identical imperfect impulse included incremental indicated influence injection intravenous introduced kidney kidneys kinetic lake lead male manual measured measuring medicine medulla mild mildly minutes model modeled models monitored mutual noise noiseless normalization normally nuclear original painting patient peaked poor post processing protocol quantitative radiology radionuclide rather reflect registration renal representative represents required residuals resistant respiration restricting retention salt school segmentation severe severely simulated simulation software step studies table target technician tissue tracer trial trials true type unconstrained unregistered variate varied varying volume whole years zero