Abstract
Purpose :
Recently, the brand new scientific theme was found by artificial intelligence. One of them was gender determination by ocular fundus photography. To our knowledge, there has been few manuscript to predict gender by fundus photography alone. However, it’s not clear which parameter of the fundus photographs affect the gender judgment. Therefore, the purpose of this study was to investigate gender determination by fundus photograph parameters using binominal logistic regression in young healthy eyes.
Methods :
A prospective observational cross-sectional study was performed in 112 right eyes of young healthy volunteers. Using the color fundus photograph, Tessellation fundus index (TFI, Yoshihara N, et al. Plos ONE, 2015) was calculated by TFI=R/(R+G+B) using the mean value of red-green-blue intensity in the eight locations around the optic disc. Optic disc ovality ratio, papillo-macular angle, retinal artery trajectory, retinal vessel angles were quantified by our previous reports (Yamashita T, et al. Invest Ophthalmol Vis Sci. 2013 and 2014). The binomial logistic regression was used to predict the gender by these fundus parameters.
Results :
Mean age was 25.8 years, 76 males and 36 females. In the forward selection method, five parameters were selected, and the predicted correct answer rate was 85.7%. In the backward selection method, 22 parameters were selected, and the predicted correct answer rate was 95.5%.
Conclusions :
it was possible to estimate men and women by color fundus parameters, such as TFI, the shape and position of the optic disc, and retinal vessel trajectory.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.