Abstract
Purpose:
To test the use of variable scan sizes to generate variable field sizes giving low resolution images over a wider region of retina, which then provide references for image alignment even in the presence of modest sized saccades.
Methods:
Three normal subjects participated in this experiment. The Indiana AOSLO was used to generate both larger size (3.4 deg x3.3 deg) and standard size (typically either 1.9 deg x 1.8 deg or 1.3 deg x1.2 deg) fields. These smaller field sizes, which provide pixel sizes of approximately 1 micron per pixel or 0.66 microns per pixel, are desirable for capturing small features but in patients with poor fixation can be difficult to align and average due to eye movements. The steering capability of the AOSLO was used to make a rapid set of measurements on the eyes that covered 10x10 degrees. Subjects were also imaged using the normal fields sizes within the same session. In post-processing a montage of the large field size image was first generated and resized to the same number of pixels per degree as the small field images. Next each frame of the small field video sequences was cross-correlated with the large field montage and the location within the larger field identified. The location of the peak of the cross-correlation was used to determine the best template location within the larger field for each frame.
Results:
The larger field data acquisition required about 3 minutes additional time within a session. As expected the peak locations for the cross correlation were stable when fixation was stable. However, during eye movements between frames which cause problems for our standard alignment approach, the new approach was able to find a corresponding retinal location. Overall, the normalized cross-correlation was lower, presumable due to a drop in image quality for the large field template, due to both decreased AO control over the wider field size, but also due to under sampling of the images.
Conclusions:
The Multi-resolution approach holds promise for increasing the yield of usable frames in subject with relatively poor fixation. It adds only a slight amount to the total imaging time required per patient. The major limitations are that improvements may not extend to subjects with very bad fixation because in those cases acceptable frames of even the large field size may be difficult to obtain.
Keywords: 552 imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)