The glaucoma disease causes a successive dying of the retinal nerve fibers and thus is one major reason for blindness. The resulting morphological changes of the optic nerve head (ONH), called glaucomatous opticus atrophy, are one major indicator for the disease. Current state of the art methods describe the morphology of the ONH using geometric parameters such as normalized distances to gain well discriminating features.

In the current project period a novel Glaucoma Risk Index (GRI) was developed that is based on data driven approaches to gain relevant features well describing the appearance of th ONH. Thus, the approach no longer exclusively relies on exact measurements of the ONH and other prominent regions. Preprocessing eliminates non glaucomatous mutation such as variation ofthe vessel tree, ONH size and illumination in order to allow a pure image based classification approach. To capture different characteristics the ONH images are compressed by principal component analysis and interpreted as features in a 2-stage glaucoma classification. This monomodal setup on the color fundus images achieves an area under ROC of 0.85 that is competitive with established, but more expensive methods. Due to the utilization of the inexpensive fundus imaging the developed GRI provides first hints to early detect the glaucoma disease. A intuitive and location independent access to the calculation of the GRI is provided to the clinical personal by an web interface.

A further combination of the fundus image based GRI with further indices of the Heidelberg Retina Tomograph (HRT) and functional FDT led to futher increased AUC of 0.92.