A new hybrid approach for automated leakage detection in fundus fluorescein angiography images

The existing methods for quantitative investigation of leakage in fluorescein angiography (FA) images involve manual delineation and lack annotated datasets and high accuracy. The proposed hybrid approach can automatically detect leakage in FA images with high accuracy. The hybrid approach combines the fuzzy c-means (FCM) with distance regularised level set evolution (DRLSE). Two expert radiologists test the accuracy of the proposed system. We have also tested inter-and intra-observer variability between the two experts. The study observed high accuracy of 0.98 ± 0.02 and 0.98 ± 0.03 with experts 1 and 2, respectively. Further, the accuracy of manual inter-observer (expert 1 and expert 2) and intra-observer (expert 1) were observed as 0.95 ± 0.05 and 0.98 ± 0.02, respectively. The proposed fully automated hybrid approach can accurately detect and quantify the leakage in FA images and thus has great potential value in diabetic retinopathy (DR).