Abstract
Purpose::
Improved visualization of retinal features could aid in the detection, localization and tracking of ocular disease. The aim of this work is to enhance fundus imaging by using overall image quality metrics and a system which combines Mueller matrix polarimetry and a confocal scanning laser ophthalmoscope (CSLO).
Methods::
A generator, composed of a linear polarizer and a quarter wave plate was incorporated into the illumination path of a CSLO. The 4 young adult participants had normal vision and no pathologies. A series of four retinal images with independent incoming polarization states for each fundus region (optic nerve head and retinal blood vessels) were recorded and used to compute the pixel-by-pixel elements of the top row of the Mueller matrix. From these elements, images with the highest and lowest values of different overall image quality metrics (signal to noise ratio, entropy and acutance) were constructed by changing the Stokes vector values. The constructed images for non-polarized incident light and the average of 32-frames measured with linear polarized incoming light were also used for comparison.
Results::
The polarization state of light, which produced the recorded image with the best quality, depended on the participant and retinal area. For images containing blood vessels, the highest values for the three metrics for the constructed images were higher than all recorded images and higher than those of the non-polarized light images. Images associated with the maximum and the minimum values of the metrics showed different details of the optic nerve head; lamina pore structure was best for images of both minimum SNR and entropy, whereas maximizing SNR and entropy improved the visualization of the neural retinal rim.
Conclusions::
We have reported a simplified polarimetric technique to improve the quality of fundus images using three global image quality metrics. Although different metrics emphasize differing retinal structures, entropy was shown to be the most sensitive metric to differences in the image quality. This method provides a step forward in improving CSLO fundus imaging, since more structural details and small features, which were not discernible in the original images, can be observed.
Keywords: imaging/image analysis: non-clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • optical properties