Nonparametric Shape Analysis Methods in Glaucoma Detection

A statistical method for glaucoma detection using tomographic images is discussed. It is known that the ONH (optic-nerve-head) area contains all the relevant information on glaucoma. Mean change of the angles of the tetrahedron determined by four control points, three on the neural rim and the other one corresponding to the maximum depth of the ONH is tested for significance. Apart from Hotelling’s T 2 -test, a nonparametric bootstrap and permutation method are used for statistical analysis, because the assumption of normality of the data set seems clearly violated. Moreover, a projection pursuit approach based on the sign test is applied as an alternative to these nonparametric procedures. The statistical analysis is done using data from Louisiana State University, Eye Center.