We have studied retinal image registration in the context of the National Institutes of Health (NIH), Early Treatment Diabetic Retinopathy Study (ETDRS) standard protocol. The ETDRS protocol defines seven 30-degree fields of each retina with specific field coverage. A robust ETDRS image registration algorithm is required to (1) assess image quality in terms of ETDRS field coverage, and to (2) support ETDRS-based disease staging. Three major challenges are present. First, small overlaps between adjacent fields lead to inadequate landmark points (crossovers and bifurcations) for feature-based methods. Second, the contrast and intensity distributions within an image are not spatially uniform or consistent. This can deteriorate the performance of area-based techniques. Third, high-resolution ETDRS images contain large homogeneous nonvascular/textureless regions which result in difficulties for both feature-based and area-based techniques. In this work, we propose an ETDRS image registration algorithm which effectively combines both area-based and feature-based methods

Retinal image registration example.

Related Publications

• T. Chanwimaluang, G. Fan, and S. Fransen, “Hybrid Retinal Image Registration”, IEEE Trans. on Information Technology in Biomedicine, Vol. 10, No. 1, , pp129-142, Jan. 2006.

• T. Chanwimaluang and G. Fan, “Retinal Image Registration for NIH’s ETDRS", in Proc. International Symposium on Visual Computing, Lake Tahoe, Nevada, Dec. 2005, also in LNCS, Vol. 3804, Editors: G. Bebis, Springer, 2005.