Purpose:
Intelligent computing techniques are used to automaticallyregistering retinal images that are separated in time. Any differencebetween the registered images may show changes of disease progressionthat have occurred.
Methods:
Image processing techniques are used to locate thebrightest region in each retinal image, that of the optic–head.The centroid of this region is then derived, and is used toalign the optic–head of the two images, and act as centreof rotation to register the vasculature of the two images.Theoutput from the applied fuzzy classifier to each image is agrey–scale image where each pixel has a value between0 (likely background) and 255(likely vessel). These two outputimages are then aligned, rotated and subtracted to find therotation angle that gives the minimum value, and the optimumregistration. The aligned output images are then subtractedon a pixel by pixel basis to produce the difference image.
Results:
The procedure was applied to time–separated theabnormal images, and typically results are shown below for twotime–separated but registered images [Fig.3], and subtraction[Fig. 4], where changes significantly changes in the imagesmay be observed.
Conclusions:
The general approach appears to provide an automatedapproach to the registration of time–separated retinalimagers, and the highlighting of changes. This procedure willform the basis of further clinical studies.
Keywords: image processing • imaging/image analysis: clinical • age-related macular degeneration