Surface reconstruction is rarely been investigated to reconstruct 3D surfaces from data obtained from the internal human body organs. The 3D cloud of points can view objects in R 3 space, however, it will not clearly demonstrate the curvatures of the outside of objects. Our 3D clouds of points are obtained from coronary artery trees. Each single-view angiogram can produce a 3D coronary artery tree. An approach to reconstruct 3D-oriented surfaces of 3D coronary artery trees is proposed. The approach leverages the Poisson problem for 3D surfaces from oriented data. The approach does not require the establishment of topological relations between adjacent points and involves no implicit parameter fitting. The approach consists of three stages: first, calculation of Euclidean distances between the clouds of points. Second, 3D-oriented data structuring. Finally, Poisson surface reconstruction of the 3D-oriented data (oriented cloud of points). An additional stage is added to measure the curvatures inside the 3D surfaces. Experimental evaluation has been done to raw of clinical data sets and results revealed that the proposed approach is efficient to reconstruct 3D surfaces of coronary artery trees. Results show that our proposed approach has high robustness for a variety of 3D cloud of points. The output surface can clearly display all the details and curvatures of the cloud of points. Our proposed algorithm of surface reconstruction plus the curvatures estimation is able to automatically indicate the locations of arteries and warn specialists of any abnormal medical cases of artery’s penetration inside the heart.
