Template definition and neural networks for pattern recognition?

Hello, I am working on my thesis with dermoscopic images. My purpose is to extract and identify the vascular patterns in skin lesions. The first part is quite done, that is to say, I have extracted the veins thanks to thresholding and clustering. Now my image consists of black background and veins. However, I am not sure at all of how to go on. The shapes can be any, from lagoons to hairpins, dots, strikes, etc. So should I first do some machine learning for the template definition? Or how could I do it? And once I have defined the templates, what would be the best for identifying the patterns? The matlab pattern recognition toolbox? Or neural networks???? Any help would be more than welcome... Best, Nerea