Respiratory motion estimation in x-ray angiography for improved guidance during coronary interventions. (10th April 2015)

• Record Type: Journal Article
• Authors: Baka, N; Lelieveldt, B P F; Schultz, C; Niessen, W; van Walsum, T
• Is Part Of: Physics in medicine & biology. Volume 60:Number 9(2015:May)
• Page Start: 3617
• Page End: 3637
• Publication Date: 2015-04-10
• Subjects: 2D/3D registration -- PCI -- XA -- CTA -- breathing motion; Biophysics -- Periodicals; Medical physics -- Periodicals; 610.153
• DOI: 10.1088/0031-9155/60/9/3617
• Languages: English
• ISSNs: 0031-9155

Abstract

Abstract: During percutaneous coronary interventions (PCI) catheters and arteries are visualized by x-ray angiography (XA) sequences, using brief contrast injections to show the coronary arteries. If we could continue visualizing the coronary arteries after the contrast agent passed (thus in non-contrast XA frames), we could potentially lower contrast use, which is advantageous due to the toxicity of the contrast agent. This paper explores the possibility of such visualization in mono-plane XA acquisitions with a special focus on respiratory based coronary artery motion estimation. We use the patient specific coronary artery centerlines from pre-interventional 3D CTA images to project on the XA sequence for artery visualization. To achieve this, a framework for registering the 3D centerlines with the mono-plane 2D + time XA sequences is presented. During the registration the patient specific cardiac and respiratory motion is learned. We investigate several respiratory motion estimation strategies with respect to accuracy, plausibility and ease of use for motion prediction in XA frames with and without contrast. The investigated strategies include diaphragm motion based prediction, and respiratory motion extraction from the guiding catheter tip motion. We furthermore compare translational and rigid respiratory based heart motion. We validated the accuracy of the 2D/3D registration and the respiratory and cardiac motion estimations on XA sequences of 12 interventions. The diaphragm based motion model and the catheter tip derived motion achieved 1.58 mm and 1.83 mm median 2D accuracy, respectively. On a subset of four interventions we evaluated the artery visualization accuracy for non-contrast cases. Both diaphragm, and catheter tip based prediction performed similarly, with about half of the cases providing satisfactory accuracy (median error < 2 mm).