Detecting statistical differences between average glaucoma indices of the eye recovered from virtual 3D reconstructions of the eye’s optic nerve head using 2D stereo images

Abstract

In this article, we revisit a Shape-based Glaucoma Index previously examined in Derado et al. (2004) using Scanning Confocal Laser (Heidelberg Retina) Tomographic (HRT) images. Here, we examine the use of 2D stereographic images for detecting glaucomatous shape change. Similar to Derado et al. (2004), we are interested in detecting such change from the same Rhesus monkeys experimental data. In contrast to Derado et al. (2004), in this study we use 2D stereographic images that were taken on the same sample of Rhesus monkeys before and after receiving an eye injection, inducing glaucoma-like Intra-Ocular Pressure (IOP) in one of their eyes, while keeping the other eye as control. Once we retrieve the 3D Kendall Shape reconstruction for the stereographic images of the Rhesus monkeys eye using Agisoft PhotoScan, we summarize the shape by selecting three matched corresponding biological landmarks on the neural-rim and the fourth matched landmark as the point of maximum depth. Clinicians believes that the ratio depth to diameter of the eye cup provides useful measure of shape change. We calculate the glaucoma index, (G), a normalized depth of the eye cup of Rhesus monkeys before and after receiving an injection. Parametric and nonparametric analyses were performed. Both procedures detected a positive mean glaucoma index increase in the medically induced glaucomatous eye. Our analyses supports a similar conclusion obtained in Derado et al. (2004) using HRT images.

Keywords:

• Glaucoma index
• Medical imaging
• Anatomical landmarks
• Nonparametric bootstrap
• 3D Kendall Shape Reconstruction

Acknowledgments

Vic Patrangenaru thanks the National Science Foundation for partial support from DMS-1106935. The authors are very grateful to Hilary Thompson for providing them with the medical image files from the LEGS library.

Additional information

Funding

Vic Patrangenaru thanks the National Science Foundation for partial support from DMS-1106935.