Purchase this article with an account.
D. W. Wong, J. Liu, N. Tan, Z. Zhang, F. Yin, H. Li, J. Lim, L. Tong, S. Saw, T. Wong; Automatic Detection of Peripapillary Atrophy in Digital Fundus Photographs. Invest. Ophthalmol. Vis. Sci. 2010;51(13):1798.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine the performance of an automatic system for the detection of peripapillary atrophy in digital fundus photographs
We tested the performance of a proposed approach for the detection of peripapillary atrophy (PPA) using a sample of digital fundus photographs from the SCORM study (Saw et al, IOVS 2006). The optic nerve head is first detected using an approach similar to our previously reported method (Wong et al, ARVO 2009). Subsequently, an entropy transform is applied on the peripapillary area, the output of which is divided into sections corresponding to the inferior, superior, nasal and temporal regions. Principle component analysis is used to determine the variables which best describe the data. These variables are then used as inputs in a naïve Bayesian classifier to determine the presence of PPA in the fundus photograph.
The sample consisted of 40 photographs with PPA and 40 without PPA, clinically assessed by an ophthalmologist. Using the described approach, 86.3% of the photographs were correctly classified (sensitivity: 0.83; specificity 0.90). Cross-validation was performed using a leave-one-out approach, attaining an average accuracy of 83.8%.
We have tested an approach proposed for the automatic detection of peripapillary atrophy. The test results are promising for the further development of this approach into an automated tool to aid in the detection of early glaucomatous damage in digital fundus images.
This PDF is available to Subscribers Only