Coronary artery curvature extraction using physic-based model in intravascular ultrasound image sequence

Three-dimension reconstruction of intravascular ultrasound (IVUS) images allows to study atherosclerotic treatment in longitudinal views and to study dynamical artery features. Vessel curvature extraction is the main step to reconstruct volumetric coronary artery. Unfortunately, IVUS pullback image sequence has the lack of three-dimensional information due to vessel motion around the IVUS catheter and exterior factors such as heart pumping and breathing. It typically needs an additional external approach, such as angiography, to overcome the problem. In this paper, a physic-based model is proposed to extract vessel curvature without any external information. The vessel tissue is determined by an iterative method for each frame. Considering the movement of the tissue mass center, the displacements of IVUS frames are calculated. Finally, external terms are removed by applying the appropriate filters in the frequency domain. The extracted curvature is validated with an IVUS-Angiography data. The obtained vessel curve has mean difference of 0.033 mm^-1 parametric curvature in comparison with the curve achieved by fused IVUS-Angiography data.