advertisement
Editors Selection IGR 15-4
Basic Research: Fundus photography of localized and early defects
Comment by Tae-Woo Kim on:
26813 Detection of retinal nerve fiber layer defects on retinal fundus images for early diagnosis of glaucoma, Muramatsu C; Hayashi Y; Sawada A et al., Journal of biomedical Optics, 2010; 15: 016021
Evaluation of the retinal nerve fiber layer is an important tool to diagnose glaucoma. It is particularly helpful to detect glaucoma in the early stage of disease because structural damage often precedes the detectable functional change in glaucoma. Moreover, it is considered that RNFL examination will provide more information about minor loss of axons than evaluation of the optic nerve head where axons are densely packed.
Muramatsu et al. (1172) describe a new computerized method to detect localized, early nerve fiber layer defect (NFLD) in the color fundus photographs. After removing the major blood vessels, they transformed the images so that the curved paths of retinal nerve become approximately straight on the basis of ellipses. Then, Gabor filters were applied for enhancement of NFLDs. Bandlike regions darker than the surrounding pixels were detected as NFLDs. In the comparison with the gold standard which were defined as NFLDs identified by two ophthalmologists, the computerized system initially detected 11.8 defects per case. To reduce the false positive detection, six image features including the area, vertical length, the average pixel values in the original and filtered images, and the contrasts in the original and filtered images were determined and the likelihood scores of the defects being true NFLDs were determined by using two classifiers: a linear discriminant analysis and an artificial neural network (ANN). Finally, they observed the sensitivity of 91% to detect NFLDs in at one false positive defect per image by ANN in 81 eyes. With this threshold, 52% of the cases without NFLD could be correctly dismissed as containing no suspicious region. This publication contributes to the potential use of automated systems to diagnose NFLDs in fundus photographs. Although, such a system may not be used as a confirmatory examination to define NFLDs, it may be useful for mass screening. However, it seems that several limitations remain to be solved. This system needs manual identification of the optic disk. Although, it is a simple task, a fully automated system would be desired. More importantly, the system was designed to detect localized, early defects. Thus, as the authors suggest, development of different methods or a new detection method that may detect relatively wide defects is needed. Then, the performance of the computerized method should be re-evaluated.
Comments
The comment section on the IGR website is restricted to WGA#One members only. Please log-in through your WGA#One account to continue.Log-in through WGA#One
