Meeting Banner
Abstract #1821

Automatic 3D Probe Localization and Iceball Segmentation for MRI-Guided Kidney Cryoablation

Xinyang Liu1, Kemal Tuncali1, William M. Wells, III1, Gary P. Zientara1

1Radiology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United States

MRI-guided cryoablation provides direct visualization of the target tumor and iceball. Visual assessment of therapy progress during the freezing cycle can be challenging, though. Nonetheless, real-time quantitative assessment of ablation is needed to ensure safety and effectiveness. We developed software that includes methods for: (1) automatic cryoprobe localization, able to detect multiple inserted probes (in 3D) from pre-freeze MRI; and, (2) automatic iceball segmentation, guided by the labeled probes, that can extract 3D iceball configuration from periodic MRI scanning during the freezing cycle. Both methods showed good accuracy when applied to five cases of 3T MRI-guided cryoablation of kidney tumors.

Keywords

abdominal ablated ablation according accuracy accurate anatomy applied applying approaches approximates artifact artifacts assess assessment automatic automatically axis black border breath bundle certain challenging combining comparing composed computational computer configuration consecutive containing controlling created curves cycle cycles cylindrical derived detect detects determined develop difficulty direct directly distances dropping energy euclidean every evolving experienced experimentally extracts feasibility feature flow freeze freezing fully good graph guided hart haste hold hough identifying incorporated initial initiate inserted intensity involving kidney labeled limited localization localized locates major manner manual manually margin measuring minimization minute minutes modeled modeling monitoring need novel overcome periodic periodically phantom pictures placement planned position prior priors probe probes problem progress prolate qualitative quantitative radiology rapid real represented respectively retrospectively scanned scanning school segmentation segmented segmenting selected sequentially shape similarity slices software space supported surrounding system taken target therapy tips transform tumor tumors ultimate validate validation vision visual void volume volumes women yielding