Segmentation of Retinal Vasculature using Active Contour Models (Snakes)

Characteristic of retinal vasculature has been an important indicator for many diseases such
as hypertension and diabetes. A digital image analysis system can assist medical experts to
make accurate diagnosis in an efficient manner. This project presents the computer based
approach to the automated segmentation of blood vessels in retinal images. The detection
of the retinal vessel is achieved by performing image enhancement using CLAHE followed
by Bottom-hat morphological transformation. Active contour model (snake) that based on
level sets, techniques of curve evolution, and Mumford-Shah functional for segmentation
is then used to segment out the detected retinal vessel and produce a complete retinal
vasculature. A Graphic User Interface (GUI) has also been created to ease the user for the
segmentation of the retinal vasculature. The algorithm is then tested with 20 test images
from the DRIVE database. The results shows that the algorithm outperforms many other
published methods and achieved an accuracy (ability to detect both vessel and non-vessel
pixels) range of 0.92-0.94, a sensitivity (ability to detect vessel pixels) range of 0.91-0.95
and a specificity (ability to detect non-vessel pixels) range of0.78-0.85.
IV