Abstract
Purpose :
cpRNFLT measured by optical coherence tomography (OCT) aids clinicians to diagnose optic neuropathies. To date, OCT device norms are solely based on age and neglect individual eye anatomy. Here, we derive personalized normative distribution models based on age, sex, and an estimate of the true scanning radius from participants of the population-based LIFE Adult study.
Methods :
Among the 10,000 age and sex stratified study participants, reliable (≥50 B-scan repetitions, quality ≥20 dB, ≤5% missing A-scans) Spectralis SD-OCT cpRNFLT scans (768 A-scans, diameter: 12°) of one randomly selected eye without clinically significant findings on fundus or OCT images were included. The true scanning radius on the retina, which depends on eye anatomy, was estimated based on scanning focus. At each location, an optimal model of a Gaussian RNFLT distribution was determined by Akaike's information criterion among the 64 possible models composed of all combinations of parameters age, sex, and radius for mean and variance, respectively.
Results :
5,646 eyes from 5,646 subjects (ages: 20-79 years, 54.8% females) were selected. The color bars in Fig. 1 illustrate the best models for each of the 768 locations, placed below the cpRNFLT mean curve as a spatial reference. Age, sex and radius have significant impacts on cpRNFLT mean and variance on most retinal locations. Apart from nasal areas, increasing age is associated with thinner RNFL. Larger scanning radius is related to thicker temporal RNFL and thinner RNFL on all other areas. Males have thicker superior RNFL, females thicker RNFL at most other locations. Additionally, we implemented these new cpRNFLT norms as open source software that mimics the Spectralis printout but provides a personalized normative color plot specific to age, sex, and radius. Fig. 2 A and B show cpRNFLT of two healthy eyes. While the conventional norms (top) result in seeming abnormalities (black line crossing yellow/red areas) on numerous locations, our personalized norms (bottom) only indicate few abnormalities.
Conclusions :
In addition to age, sex and scanning radius have significant impacts on normative cpRNFLT distributions. Based on over 5,600 subjects, we provide new, personalized cpRNFLT norms. To facilitate clinical applications, we implemented these norms in an open source software that superimposes cpRNFLT on a color plot closely resembling the current device printout.
This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.