amdoct: Real life OCT segmentation data of patients with AMD

In eyedata: Open Source Ophthalmic Data Sets Curated for R

Description

This CSV dataset (AMD_baseline_MEH_v1.csv) is associated with the paper Moraes et al Quantitative analysis of optical coherence tomography for neovascular age-related macular degeneration using deep learning. Ophthalmology. (2020). The dataset comprises anonymised metadata and OCT segmentation data of patients undergoing treatment for wet AMD at Moorfields Eye Hospital, London, United Kingdom. The dataset includes 2966 rows (2966 individual patients)

Usage

data("amdoct")

Format

A data frame (tibble) with 2966 rows and 24 variables:

patID

Anonymized patient identifier

sex

Sex of patient (m = male, f = female)

ageStrat

Stratified age (years) as factor. One missing value

ethnicity

Ethnicity of patient following the categories from the UK ETHNIC CATEGORY CODE 2001. (asian = Asian or Asian British, white = White, black = Black or Black British, other_unknown = Other ethnic group or unknown)

eye

Left or right eye of patient (r = right, l = left)

first_eye

Is the eye the first injected eye

va

Visual acuity in Early Treatment Diabetic Retinopathy Study (ETDRS) letters and categories hand motion (hm) or counting fingers (cf) for values <1. One value = 0 was not specified in the original data set.

inj

Injection given (TRUE) or not

time

Time in days from baseline, i.e. injection number 1

Neurosensory_volume_voxels

Number of voxels segmented as the feature neurosensory retina (NSR)

RPE_volume_voxels

Number of voxels segmented as the feature retinal pigment epithelium (RPE)

SRF_volume_voxels

Number of voxels segmented as the feature subretinal fluid (SRF)

IRF_volume_voxels

Number of voxels segmented as the feature intraretinal fluid (IRF)

SHRM_volume_voxels

Number of voxels segmented as the feature subretinal hyperreflective material (SHRM)

Drusen_volume_voxels

Number of voxels segmented as the feature drusen

sPED_volume_voxels

Number of voxels segmented as the feature serous pigment epithelium detachment (sPED)

fvPED_volume_voxels

Number of voxels segmented as the feature fibrovascular pigment epithelium detachment (fvPED)

HRF_volume_voxels

Number of voxels segmented as the feature hyperreflective foci (HRF)

Neurosensory_thickness_um

Thickness (measured in microns) of the segmented feature neurosensory retina (NSR)

IRF_thickness_um

Thickness (measured in microns) of the segmented feature intraretinal fluid (IRF)

SRF_thickness_um

Thickness (measured in microns) of the segmented feature subretinal fluid (SRF)

SHRM_thickness_um

Thickness (measured in microns) of the segmented feature subretinal hyperreflective material (SHRM)

HRF_thickness_um

Thickness (measured in microns) of the segmented feature hyperreflective foci (HRF)

CST_um

Central subfield retinal thickness measured as the sum of NSR, IRF, SRF, SHRM, HRF thickness (measured in microns)

Missing values

There are 233 missing visual acuity entries in this data set. They result from data entry errors (ETDRS values above 100) and missing data in the original medical health records.

Study setting and design

Study setting and design: This study is a retrospective cohort study of patients that commenced anti-VEGF therapy for neovascular AMD between June 2012 and June 2017 at Moorfields Eye Hospital NHS Foundation Trust, London, UK. Approval was granted by the Institutional Review Board of the hospital (ROAD17/031). The study complied with the Declaration of Helsinki

Data source

All clinical information at Moorfields Eye Hospital is recorded within an electronic medical record (EMR) application (OpenEyes Foundation, London, UK). A SQL database (SQL Server Reporting Service, Microsoft Corporation, Richmond, USA) containing all the information from the EMR is in place and regular updates are performed overnight to keep the data warehouse up-to-date. VA is reported in ETDRS letter score (and categories for ETDRS < 1). The highest value (independent of measurement method) available at each visit was chosen.

OCT

Each voxel equates to 2.60 x 11.72 x 47.24 microns in the A-scan, B-scan, and C-scan directions, respectively. All OCT data is captured using 3DOCT-2000 devices (Topcon Corp., Tokyo, Japan). All images comprise 512 x 885 x 128 voxels covering a volume of 6x6x2.3mm. In this dataset, only one image is available per eye. The image with the lowest segmented artifacts was chosen if multiple scans were available at the same visit.

Time variable

For first-treated eyes, time will generally equal 0. Where a scan at day 0 was not available, a scan up to 14 days prior to the first injection (i.e. up to -14 days), is used for analysis.

Segmentation

Segmentation data was output using a deep learning segmentation model described further in De Fauw et al. and Yim and Chopra et al.

Source

https://doi.org/10.5061/dryad.2rbnzs7m4

See Also

Scientific article to which this data set was supplement: https://doi.org/10.1016/j.ophtha.2020.09.025

eyedata documentation built on Dec. 9, 2020, 9:07 a.m.