Abstract
Purpose: :
Light coming from structures in fundus images is differentially polarised, leading to clinical applications, such as assessment of glaucoma and foveal fixation. Using polarisation, we and others have shown objective image improvement in visualisation of the retinal structures. The purpose of this study was to explore the subjective quality of fundus images obtained through polarimetry using various image metrics.
Methods: :
A high resolution CSLO was modified to include a polarisation generator (linear polariser + quarter wave plate) in the input path. Four video segments of the ONH (10°/15° fields) were recorded for differing generator polarisation states in one eye of each of two normal, young adult subjects. At each polarisation state, eight frames were registered, averaged and used to calculate the top row of the Mueller matrix for each pixel. By incrementing the incident Stokes vector in degree steps on the Poincaré sphere, images were calculated that maximised and minimised several image metrics, including signal to noise ratio (SNR), entropy and acutance. Image series were repeated on subsequent days (both subjects) and for one subject at ±0.50D of defocus. A clinician, masked to the process of image generation, subjectively evaluated the entire set of images, including the first element of the Mueller matrix (corresponding to an image in unpolarised light). Clinically relevant features of the ONH (cup shape and size, lamina cribrosa visibility & visibility of the neuroretinal rim), the retinal nerve fibre layer (RNFL) and the ONH vasculature were rated (1: Not visible – 4: Good Visibility) and then ranked.
Results: :
The image for the minimum acutance metric was consistently ranked highest for ONH features across repeated image acquisitions. Images taken with circularly and elliptically polarised light were ranked similarly to those with minimum acutance. Images calculated with maximum acutance and minimum entropy were ranked highest for visualisation of the RNFL. Images with maximum SNR and unpolarised images were ranked best, and equally, for blood vessel clarity.
Conclusions: :
In this initial study, subjective quality was improved with ONH images generated by using complex polarisation information. Overall, these images were preferred to those using linear or unpolarised light. Acutance gave the highest rankings across the ocular structures evaluated.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retina