In this paper, we present an automatic method for extracting center axis representations (centerlines) of coronary arteries in contrast enhanced (CE)-CT angiography scans. The algorithm first detects the aorta which is used as an initial mask for ostia detection~cite{grady2006:fast}. Second, the ostia locations are detected via a vessel centerline extraction method~cite{Gulsun:Tek:008} which tracks the center axis of the coronaries starting from the aorta surface. The full centerline tree of the coronary arteries are computed via the multi-scale medialness-based vessel tree extraction algorithm which starts a tracking process from the ostia locations until all the braches are reached. The centerline extraction algorithm is a graph-based optimization algorithm using multi-scale medialness filters.