Compressed Sensing Dynamic Reconstruction in Rotational Angiography H´ el` ene Langet 1,2,3,⋆ , Cyril Riddell 1 , Yves Trousset 1 , Arthur Tenenhaus 2 , Elisabeth Lahalle 2 , Gilles Fleury 2 , and Nikos Paragios 3,4,5 1 GE Healthcare, Interventional Radiology, Buc, France 2 Sup´ elec, SSE department, Gif-sur-Yvette, France 3 ECP, Center for Visual Computing, Chˆ atenay-Malabry, France 4 ENPC, Center for Visual Computing, Champs-sur-Marne, France 5 INRIA Saclay, GALEN Team, Orsay, France Abstract. This work tackles three-dimensional reconstruction of tomo- graphic acquisitions in C-arm-based rotational angiography. The rela- tively slow rotation speed of C-arm systems involves motion artifacts that limit the use of three-dimensional imaging in interventional proce- dures. The main contribution of this paper is a reconstruction algorithm that deals with the temporal variations due to intra-arterial injections. Based on a compressed-sensing approach, we propose a multiple phase reconstruction with spatio-temporal constraints. The algorithm was eval- uated by qualitative and quantitative assessment of image quality on both numerical phantom experiments and clinical data from vascular C- arm systems. In this latter case, motion artifacts reduction was obtained in spite of the cone-beam geometry, the short-scan acquisition, and the truncated and subsampled data. 1 Introduction Rotational angiography provides three-dimensional (3D) qualitative and quanti- tative information of high clinical interest as the complexity of minimally invasive procedures increases. However, its spread in the clinical practice is limited by the nature of the measurements: 3D reconstruction of C-arm system data is chal- lenging due to the use of cone-beam (CB) geometry, short-scan orbits, as well as truncated and angularly subsampled data. Besides, physiological times (e.g. heart beat, breathing time) are small compared to the gantry rotation speed of C-arm systems. The resulting temporal variations within a scan are so challeng- ing that they are usually not corrected for in the clinical practice. Temporal variations are addressed through the decomposition of the dynamic (3D+t) data into phases where the object can be considered “static” so that 3D reconstruction is feasible. Each phase defines a subset of projections that must fully sample the volume at this particular time point [9]. To alleviate ⋆ Corresponding author: helene.langet@ge.com. This work was supported in part by ANRT grant number CIFRE-936/2009. N. Ayache et al. (Eds.): MICCAI 2012, Part I, LNCS 7510, pp. 223–230, 2012. c  Springer-Verlag Berlin Heidelberg 2012