Retinal vasculature segmentation using principal spanning forests

We propose an automated morphology reconstruction method for curvilinear network analysis. The proposed approach first projects samples to the ridge of the intensity image of the curvilinear system. Then, a manifold deviation measure is utilized to approximate the ridge with piecewise linear segments between the projected samples. A nonparametric system workflow based on the kernel interpolation and density estimation is provided for the derivations without any user defined meta-parameter, i.e. hard threshold for segmentation. Lastly, a rigorous sampling strategy using the manifold deviation measure that can be used for robust sparse tree reconstruction is provided. The proposed approaches have been tested on a small set of representative retinal scans. Preliminary qualitative results indicate the effectiveness of the method.