Abstract 102: Coregistration of Serial Angiograms Using Point Cloud Matching

Abstract

Introduction: Digital subtraction angiography (DSA) is the gold standard to assess reperfusion during endovascular procedures. Visualization and quantification of changes between successive runs is challenged by patient motion and variations in acquisition parameters such as zoom and pose of the x-ray receptor. Automated coregistration of successive DSA sequences would allow for the visualization of serial changes in perfusion during the procedure. The objective of this study is to develop a fully automated framework for the coregistration of patient-specific DSA acquired at different time points during an endovascular reperfusion procedure. Methods: The dataset was established retrospectively from patients admitted at a stroke care center and diagnosed with acute ischemic stroke. Included patients underwent a clot retrieval procedure. Biplane DSA was performed before and after endovascular reperfusion intervention. A neurologist manually coregistered the anterior-posterior (AP) view of successive DSA sequences from each patient using anatomical reference landmarks. The developed computer vision framework processed each DSA to extract the vasculature using a vessel detector, followed by resampling of the responses. The resulting set of points in the pre- and post-intervention DSA were then matched using a point cloud matching algorithm. In this study, we provide an experimental analysis with the conventional RANSAC algorithm. Evaluation was performed by measuring the error between the estimated affine transform, that relates the pre- to the post-intervention DSA, and the groundtruth established manually. Results: A total of 20 patients were included in the analysis. Mean age was 66.8 (range 34-91). Distribution of the TICI scores was as follows: TICI 0(3), TICI 1(0), TICI 2a(5), TICI 2b(10), TICI 3(2). Overall coregistration error was as follows: angle (8.6+- 3.1 degrees), shift (24.8 +- 19 mm), respectively. Conclusions: RANSAC point cloud matching algorithm can be used to accurately coregister serial angiograms during endovascular procedures. This could lead to near real-time visualization and quantification of revascularization.