Extracting vessels from the fundus image of an eye

[maverick34]

We are trying to extract the blood vessels from a set of fundus images. The magnetic lasso tool and the quick selection tool don't give very good precision and take very long. We tried separating the green channel (for maximum contrast) and then selecting the vessels by colour range, but that doesn't work either. We want results like the one in the image. Is there a way to do it with relatively good precision that doesn't involve tracing along the vessels?

 

[Tom Mann]

Extracting the shape of a vascular network is one of the most common tasks in medical image processing, and there is considerable literature on the subject. Photoshop isn't a good choice for scientific or medical image processing. PS is more for graphic designers, photographers, artists, etc., not scientists and physicians. You would be much better served with Image J from NIH. The image processing toolkits for Matlab and Mathematica would be distant second choices, but still far ahead of PS because PS simply does not have the many of the most common primitives needed for such processing, e.g., morphological operators, image regularization, a good interface to mathematical operations, etc., whereas these are exactly the areas in which ImageJ, Matlab and Mathematica shine.

For example, many ImageJ plugins / add-ons have been written for angiogenesis and similar analyses:

Vascular Network Toolkit for ImageJ:
http://sourceforge.net/projects/ntwrkanlystlkit/
http://ntwrkanlystlkit.sourceforge.net/overview.php

Angiogenesis Analyzer for ImageJ:
http://image.bio.methods.free.fr/ImageJ/?Angiogenesis-Analyzer-for-ImageJ

A Computational Tool for Quantitative Analysis of Vascular Networks:
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0027385

HTH,

Tom M

PS - I presume the images you are concerned with are all taken without fluorescing or other chemical contrast agents, and that for your application, these can't be used, right?