Diaphragm tracking in cardiac C-arm projection data

Abstract

Long acquisition times of several seconds lead to image artifacts in cardiac C-arm CT. These artifacts are mostly caused by respiratory motion. In order to improve image quality, it is important to accurately estimate the breathing motion that occurred during image acquisition. It has been shown that diaphragm motion is correlated to the respiration-induced motion of the heart. We describe the development of a method that is able to accurately track the contour of the diaphragm in projection space using a 2D quadratic curve model of the diaphragm to simplify the process. In order to provide robust and stable tracking, additional constraints based on prior knowledge of the projection geometry and human anatomy are introduced. Results show that the tracking is very accurate. A mean model error per pixel of 0.93 ± 0.44 pixels for the left and 0.79 ± 0.19 pixels for the right hemidiaphragm was observed. The diaphragm top is tracked with an even lower error of only 0.75 ± 0.84 pixels for the left and 0.45 ± 056 pixels for the right hemidiaphragm respectively. © Springer-Verlag Berlin Heidelberg 2012.