Using deep learning for the automated identification of cone and rod photoreceptors from adaptive optics imaging of the human retina

Mengxi Zhou; Nathan Doble; Stacey S Choi; Tianyu Jin; Chenwei Xu; Srinivasan Parthasarathy; Rajiv Ramnath

doi:10.1364/BOE.470071

Using deep learning for the automated identification of cone and rod photoreceptors from adaptive optics imaging of the human retina

Biomed Opt Express. 2022 Sep 2;13(10):5082-5097. doi: 10.1364/BOE.470071. eCollection 2022 Oct 1.

Authors

Mengxi Zhou¹, Nathan Doble^{2

3}, Stacey S Choi^{2

3}, Tianyu Jin¹, Chenwei Xu⁴, Srinivasan Parthasarathy¹, Rajiv Ramnath¹

Affiliations

¹ The Ohio State University, Department of Computer Science and Engineering, 2015 Neil Ave., Columbus, OH 43210, USA.
² The Ohio State University, College of Optometry, 338 W 10th Ave., Columbus, OH 43210, USA.
³ The Ohio State University, Department of Ophthalmology and Visual Science, Havener Eye Institute, 915 Olentangy River Road, Columbus, OH 43212, USA.
⁴ The Ohio State University, Department of Statistics, 127 Pomerene Hall, 1760 Neil Ave, Columbus, OH 43212, USA.

Abstract

Adaptive optics imaging has enabled the enhanced in vivo retinal visualization of individual cone and rod photoreceptors. Effective analysis of such high-resolution, feature rich images requires automated, robust algorithms. This paper describes RC-UPerNet, a novel deep learning algorithm, for identifying both types of photoreceptors, and was evaluated on images from central and peripheral retina extending out to 30° from the fovea in the nasal and temporal directions. Precision, recall and Dice scores were 0.928, 0.917 and 0.922 respectively for cones, and 0.876, 0.867 and 0.870 for rods. Scores agree well with human graders and are better than previously reported AI-based approaches.