Real-time digital holography of the retina by principal component analysis

TL;DR

This paper demonstrates real-time digital holography of the human retina using principal component analysis to process high-speed interferograms, enabling enhanced contrast discrimination related to blood flow and absorption.

Contribution

It introduces a novel real-time processing method combining digital holography with principal component analysis for retinal imaging.

Findings

Achieved real-time hologram processing at 500 fps.

Successfully discriminated blood flow signals from noise.

Implemented on standard computer hardware.

Abstract

We demonstrate the feasibility of high-quality digital holography of the human retina in real-time with a fast camera and commodity computer hardware. High throughput rendering of digital Fresnel holograms from optically-acquired inline interferograms is performed in conjunction with temporal demodulation by projection of hologram sequences onto a data-derived basis in order to discriminate local narrowband coherent detection contrasts, mostly due to blood flow and optical absorption, from spurious interferometric contributions. Digital holograms are calculated from a sustained input stream of 16-bit, 1024-by-1024-pixel interferograms recorded at up to 500 frames per second, processed by principal component analysis. This temporal signal demodulation scheme consists in the projection of stacks of 32 consecutive holograms onto a basis calculated by eigendecomposition of the matrix of their time-lagged covariance; it is performed up to 20 times per second with commodity computer hardware.