U.S. flag An official website of the United States government

U.S. Food and Drug Administration

AO-OCT Dataset on Human Photoreceptor-Retinal Pigment Epithelium Macular Topography

Catalog of Regulatory Science Tools to Help Assess New Medical Devices 

This regulatory science tool (RST) is an open-access dataset containing adaptive optics – optical coherence tomography (AO-OCT) images of the outer retina, along with the cell markings and topographical characterizations of the retinal pigment epithelium (RPE) cells and photoreceptor (PR) mosaics across the temporal macula from healthy individuals. This RST can serve as a normative baseline for future studies of retinal pathology and as labelled ground truth data for the assessment of novel algorithms for automated PR/RPE cell segmentation.

Technical Description

This open-access dataset contains AO-OCT images of the outer retina from healthy individuals, along with the cell markings and characterized topography of the PR and RPE cell mosaics. Eleven healthy volunteers were imaged with a custom, high-speed Fourier domain mode-locked laser (FDML)-based AO imaging platform [1]. Only data from the AO-OCT channel were analyzed, which were collected at 13.0 Hz using the 3.4 MHz FDML swept source laser (λc = 1060 nm, Δλ = 76 nm), with estimated lateral and axial resolution of 2.9 μm and 8.4 μm, respectively. One eye of each participant was imaged, and that eye was dilated and cycloplegia induced with Tropicamide 1%. After alignment in the system, the fixation target was set to direct the gaze of the participant to nine regions of interest (ROIs) from the fovea to 12° temporal (12T), where each imaging location was separated by 1.5°. At each location, multiple AO-OCT videos with field-of-view (FOV) of 2° × 2° were recorded. Each video had 10 AO-OCT volumes separated by 6 s to allow for optimal contrast enhancement from RPE organelle speckle decorrelation [2]. With the aid of a deep learning algorithm, key metrics were extracted from the PR-RPE complex of averaged AO-OCT volumes including PR and RPE cell density, PR outer segment length (OSL), and PR/RPE ratio. Outer retinal morphology was quantified at 13 selected ROIs from the fovea to 12T, with ~1° separation between adjacent selected regions. The values for the measured RPE metrics are the mean values from two expert graders, while the PR measures were obtained by manual correction of results obtained with a deep learning algorithm [3]. Both RPE and PR density values were scaled for each participant to correct for magnification differences due to the axial length.

Intended Purpose 

This dataset RST can serve as a normative baseline of PR-RPE morphology in healthy individuals for future studies of retinal pathology focused on cellular-level changes to the retina, and for assessing the capability of future high-resolution retinal imaging technologies in resolving the PR-RPE topography. It can also serve as a tool to assess novel algorithms for automated segmentation, detection, and/or classification of the outer retinal structures.

Testing

The measurement results were compared with other published human in vivo and ex vivo reports, details of which are provided in [4]. In general, our data matched previous AO-OCT and clinical OCT values and compared favorably with results from histology and other high-resolution imaging systems. The reproducibility of these AO-OCT measures was also assessed in a subset of the cohort. Seven participants were re-imaged on two additional days at two select locations (fovea and 7.5T) for this purpose. The selected regions were graded independently. Due to subject fixational imprecision, no attempt was made to track or map individual PRs or RPE cells over the three AO-OCT imaging sessions. Intraclass correlation coefficients (ICCs) were used to characterize the precision of measured inter-session parameters. All measures showed low variance in repeat measurements and excellent to moderate reproducibility was obtained.

Limitations

The limitations of this dataset include:

  • The speckle noise and longer imaging wavelength used with the AO-OCT system make the resolution of PRs in the foveola (central 0.25°) extremely difficult, and this is a task better suited for AO-SLO. While our device includes AO-OCT and AO-SLO channels, we did not collect data from the foveola with AO-SLO to extract PR density values.
  • The imaged cohort is relatively young (35.5 ± 4.4 years).
  • Our definition of the cone outer segment boundaries may not be consistent with some previous published studies. However, controversy over the origin of the OCT signals in the outer retina remains.

Supporting Documentation

Additional data description and detalis on data access can be found here:  https://www.synapse.org/Synapse:syn63546986/wiki/629721 

Peer-reviewed journal article describing the dataset [4] and the imaging system [1, 4]:

  1. Z. Liu, F. Zhang, K. Zucca, A. Agrawal and D. X. Hammer, Ultrahigh-speed multimodal adaptive optics system for microscopic structural and functional imaging of the human retina. Biomed Opt Express, 2022. 13(11): p. 5860-5878. DOI: 10.1364/BOE.462594
  2. Z. Liu, K. Kurokawa, D. X. Hammer and D. T. Miller, In vivo measurement of organelle motility in human retinal pigment epithelial cells. Biomed Opt Express, 2019. 10(8): p. 4142-4158. DOI: 10.1364/BOE.10.004142
  3. S. Soltanian-Zadeh, Z. Liu, Y. Liu, A. Lassoued, C. A. Cukras, D. T. Miller, et al., Deep learning-enabled volumetric cone photoreceptor segmentation in adaptive optics optical coherence tomography images of normal and diseased eyes. Biomed Opt Express, 2023. 14(2): p. 815-833. DOI: 10.1364/BOE.478693
  4. Z. Liu, S. Aghayee, S. Soltanian-Zadeh, K. Kovalick, A. Agrawal, O. Saeedi, et al., Quantification of Human Photoreceptor-Retinal Pigment Epithelium Macular Topography with Adaptive Optics-Optical Coherence Tomography. Diagnostics (Basel), 2024. 14(14). DOI: 10.3390/diagnostics14141518

Contact

Tool Reference