Purchase this article with an account.
Tristan Swedish, Karin Roesch, Devesh Jain, Ramesh Raskar; eyeSelfie 2.0: Display and Algorithm for Self Directed Bi-ocular Alignment for Fundus Photography. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5968.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Fundus photography typically requires a highly trained operator because eye alignment to patients is challenging. Self directed eye alignment makes possible streamlined screening methods and home monitoring of eye disease.Previous work has has included light field patterns for monocular eye alignment with a known pupil size. We have extended this work with a method to allow self alignment using a combination of a light field display and eye tracking algorithms for an unknown size pupil.
We analyze the optical path of light-field based eye alignment patterns and developed framework to define the class of patterns that allow alignment regardless of pupil size. We built a indirect ophthalmoscope that uses near infrared (NIR 850nm) illumination for active alignment and a visible flash to capture images.Pupil size independent display patterns are less accurate for axial alignment (distance to the imaging lens), so we developed an active imaging tracking algorithm to estimate the axial distance from the imaging lens. This tracking algorithm analyzes frames captured by the device using NIR illumination to find feature points such as blood vessels and the optic disc.
We captured images of the display using a 2.0 f-number camera to visualize the display pattern from the user’s perspective. We observe that for small displacements of the camera, the alignment pattern makes a visible change. We also find that alignment estimation from the active tracking algorithm performs well while imaging a model eye.
Pupil size independent eye alignment patterns can be constructed that enable self alignment of the fundus. The loss of accuracy in axial alignment can be overcome using active alignment algorithms that estimate eye position from IR fundus images. We demonstrate this combination with a prototype indirect ophthalmoscope that incorporates a light field display in the imaged eye and and LCD for the fellow eye.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
We observe large changes in the perceived light field with subtle misalignments. This can be explained by the pupil occlusion of the 4D light field. We can visualize this occlusion (top row) using a 2D slice of the light field.
Hardware prototype of the bi-ocular system. One eye (left in image) is shown the light field display while simultaneously being imaged. The fellow eye (right in image) is shown a pattern on an LCD to help stabilize the view.
This PDF is available to Subscribers Only