Independent Component Analysis for Two-Spectral-Band Functional Imaging Fundus Camera

Abstract

Purpose: : Intrinsic optical imaging on the retina has been intensively studied. We have established a method using a fundus camera to measure retinal reflectance changes caused by optical or electrical stimuli (Okawa, IOVS 2007, Mihashi, ARVO 2008). To extend the ability of the functional imaging, we built a functional imaging fundus camera with two-spectral-band observation (730-780 nm, 820-880 nm) (FIF2S). In this report, a method to analyze images obtained by the FIF2S was investigated.

Methods: : We obtained images from a cat retina using the newly built FIF2S. The two-spectral-band near-infrared imaging was performed with two high quality CCD cameras. One sequence of measured images consisted of 1000 images in one spectral range for 26-second imaging duration and 40 frames per second. The eye of the cat was studied under general anesthesia. We used two different kinds of stimuli; a visible light stimulus and transcorneal electrical stimulus (TES). After obtaining two-spectral-band images, we performed independent component analysis (ICA). The difference of the analysis from the ICA we used in our previous study for a single-spectral-band fundus camera was that we treated a couple of two spectral band images obtained simultaneously as a unit for the new ICA. We obtained ten independent components (ICs) for one analysis. Each IC consisted of an image for 730-780-nm band, an image for 820-880-nm band, and a common time course for both spectral bands.

Results: : We found richer information from the ICA applied to the images obtained by the FIF2S than the results from the single-spectral-band fundus camera. With the previous camera, we usually obtained darkening response of the artery to the stimulation. ICs to the light stimulation from the FIF2S were categorized into three kinds. Firstly, we obtained the same component (the artery response) as we found in the previous study. Secondly, we found a slow response in the vein. Thirdly, we also found fast responses in the choroid. The time courses of the ICs with the corneal electrical stimulation were more varied than those with the light stimulation. We found two kinds of responses in the artery and in the vein.

Conclusions: : The advantage of the new FIF2S with the ICA over the previous method was that the response in the vein and the choroid can be detected. This suggests that the method can obtain information concerning retinal and choroidal metabolism.

Keywords: imaging/image analysis: non-clinical • metabolism • neuro-ophthalmology: diagnosis