Vascular cross-section is an important basis for computer-assisted diagnosis of vascular diseases. In this study, a new algorithm of calculating cross-sectional area of blood vessels was proposed. The algorithm first extracts centerline of the blood vessel from the image through thinning algorithm, and obtains neighborhood coordinates of each point on the centerline. The principal component analysis is employed on and the blood vessel cross-section is achieved via coordinating transformation with the tangent vector. Experiments were performed with the real data and simulated data, and the results demonstrate that the proposed method can efficiently detect stenosis of blood vessels with high accuracy, and achieve 2.06% deviation from the simulated ring model.