Optic cup and optic disc analysis for glaucoma screening using pulse-coupled neural networks and line profile analysis

This paper proposes an image processing algorithm that segments and measures the optic cup and the optic disc, by using a pulsed coupled artificial neural network and a line profile technique respectively. Our approach extracts two key glaucoma prediction features-the vertical cup-to-disc ratio (vCDR) and the Inferior Superior Nasal Temporal (ISNT) rule. A total of 126 fundus images have been used to evaluate the proposed algorithm. The vCDR and ISNT rule evaluation was manually determined by experienced eye specialists. The proposed algorithm was then used to automatically estimate the same parameters on the same images. The algorithm achieved a RMSE of 0.11. Furthermore, we conducted a similarity test between the values for the parameters extracted using our algorithm and that of the manual estimation, using a student's T-test. The probability of difference in datasets was 2.08∗10−13%. This could be a key step in providing good features for subsequent autonomous screening of glaucoma.