Junction Resolution Projects
These project aim to form a retinal vascular graph from segmented retinal vessels,
by determining the branching pattern of retinal arteries and veins. These projects
are the first extensive study of an algorithm to resolve the geometry of retinal
junctions, and to build a retinal graph, from vessel segments.
We have conducted our analysis using digital fundal images from diabetic screening
populations. However, the algorithms exploit geometrical features of vessels that
are not highly dependent on the imaging modality, although excessive sensor noise
might affect performance. We expect that the techniques will be suitable for fluorescent
angiography and other image capture techniques, although verifying this supposition
remains for future work.
The public-domain
DRIVE database
was used to evaluate the algorithms. The blood vessels of the DRIVE test set were
segmented using the
ESP algorithm.
Click here to download these segments.
Each segment is represented by a set of profiles. Each profile consists of two edge
points. The distance between these points represents the width of the profile while
the perpendicular vector to the line segment joining them represents the vessel
profile direction.
The resulting system is able to determine the connectivity of most junctions in
the retinal graph. Overlapping segments cause specific problems, and are identified
by their unusual configuration and processed separately.
Currently we are developing automated methods to extract measurements from the identified
junctions, and evaluate the predictive and diagnostic potential of these for a variety
of retinal diseases.
Also, a bayesian framework for the local configuration of retinal junctions is under development and as a part of this project a new dataset fro bifurcations based on the DRIVE dataset will be extracted.
The Drive Segment-Junction Set is under construction.
The output of this project were published in a paper entitled ”Automated analysis
of retinal vascular network connectivity.” The abstract is given below:
This paper describes an algorithm that forms a retinal vessel graph by analysing
the potential connectivity of segmented retinal vessels. Self organizing feature
maps (SOFMs) are used to model implicit cost functions for the junction geometry.
The algorithm uses these cost functions to resolve the configuration of local sets
of segment ends, thus determining the network connectivity. The system includes
specialized algorithms to handle overlapping vessels. The algorithm is tested on
junctions drawn from the public-domain DRIVE database.