Possible Automated Screening of Age-Related Macular Degeneration

Abstract

Purpose: : To derive an objective and quantitative parameter of hyperspectral data differentiating age-related macular degeneration (AMD) from normal macula.

Methods: : This work was approved by the Ethics Committee of Tokyo Medical University and performed according to the Helsinki declaration. Thirty three eyes of twenty five patients with AMD were includes in this study. Color fundus photograph, indocyaninegreen angiography, and the following near-infrared hyperspectral images were performed. Sixteen eyes of eleven healthy volunteers were also hired as control.To develop a near-infrared hyperspectral imager for ocular fundus, a commercialized conventional fundus camera was customized. It could acquire a set of hyperspectral data within five seconds. The data consisted of spatial information of 480 x 321 pixels covering a 50-degree field at the ocular fundus and a reflectance spectrum in the effective wavelength region from 720 to 920 nm for per each pixel.A spectrum averaged over each macular was regarded as a macular vector. The macular vector averaged over all healthy volunteers was defined as a reference vector. Variation in the relative length and angle of each vector on the basis of the reference vector as a plot in the two-dimensional plane was calculated. A parameter which minimized the overlap between plots of patients with AMD and those of control was defined as an AMD differentiation index.

Results: : Difference in the proposed index between the two groups was statistically significant. On resubstituion, receiver operating characteristics analysis showed a sensitivity of 91 % and a specificity of 85 %.

Conclusions: : We proposed a simplified diagnostic index based on near-infrared spectra. Even though the diagnostic accuracy of AMD is satisfactory as a screening examination, additional prospective study would be required to make sure the utility of this method. The intensity of near-infrared light reflected from macula could be an important factor in diagnosing AMD.

Keywords: age-related macular degeneration • clinical research methodology