SCALE-TORT

Introduction

Tortuosity could be defined as something winding or twisted, therefore a retinal vessel is tortuous if it has abnormal twists and turns along its length. Changes in the tortuosity of retinal blood vessels have been reported as a characteristic of great importance in some diseases, such as retinopathy of prematurity (ROP). Scale variation is an important property in retinal imaging, since during longitudinal epidemiological studies, fundus images can be taken with different equipment, at different times and with different parameter settings such as resolution and magnification (field of view -- FOV). Thus, a measure that is scale invariant and that finely detects changes in turns and twists, regardless of the equipment used to take it, is a challenge.

SCALE-TORT database was built with the objective of comparing different tortuosity metrics computed on the same retinal blood vessel segment extracted from two very different imaging modalities. The current database was built with the objective of comparing the same metric of tortuosity on the same segment of blood vessel extracted from two very different imaging modalities.

Data

The images used in SCALE-TORT database were obtained as part of the population-based LIFE Adult Study conducted by the Leipzig Research Centre for Civilization Diseases at Leipzig University [1]. The LIFE Adult Study includes 10,000 randomly selected participants from the population registry of Leipzig, Germany. Among many other physical and medical examinations taken during this study, information from the eye of the participants via ophthalmological imaging was registered.

The study employed high quality non-mydriatic (i.e. pupil dilation is not required) fundus photography (FP) (Oculus Nidek AFC-230, 45 FOV) and a infra-red reflectance (IR) en face image obtained by scanning laser ophthalmoscope (SLO) (Heidelberg Spectralis, 30 FOV) to produce representative images of the retina and optic nerve. In order to prove the scale invariance property of tortuosity metrics, ophthalmological photographs from 18 subjects, both eyes, were randomly extracted from the LIFE-Adult-Study, 36 FP images and 36 IR (fovea centered) images, a total of 72 images of the ocular fundus.

The original image sizes were of 3744×3744 and 768×768 pixels, for FP and IR respectively.

[1] M. Loeffler, C. Engel, P. Ahnert, et. al. The LIFE-Adult-Study: Objectives and design of a population-based cohort study with 10,000 deeply phenotyped adults in Germany, BMC Public Health 15 (2015) 691. doi:10.1186/s12889-015-1983-z.