In this paper we report results from an ongoing study about the diagnostic benefits of three-dimensional (3D) visualization and quantification of stenosed coronary artery segments. Biplane angiographic images do not provide enough information for the exact reconstruction of the coronary arteries. Therefore, a priori information about the 3D shape to be reconstructed must be incorporated into the reconstruction algorithm. One approach is to assume a circular cross-section of the coronary artery. Hence, the diameter is estimated from the contours of the vessel in both projections. Another approach is based on densitometry and searches for a solution of the reconstruction problem close to the previously reconstructed adjacent slice. In this paper we apply contour information as well as the densitometrical profiles of the two orthogonal vessel projections. We present a new probabilistic densitometric reconstruction algorithm, which extends the correct handling of the stochastic properties of the density profiles into the network flow based reconstruction algorithm. The reconstructed coronary segment is visualized in three dimensions. In order to assess the accuracy of the reconstruction, the method is applied to tubes with artificial obstruction of known geometry, modeling coronary artery stenoses. These catheter tubes are filled with normal iodine contrast material. The results of the reconstruction and visualization are discussed with respect to clinical usefulness.