MoDRIA Data Codebook

Multimodal Database of Retinal Images in Africa (MoDRIA) platform with 38 codebook variables.

Multimodal Database of Retinal Images in Africa

(MoDRIA)

Dataset Overview

Dataset Description

The Multimodal Database of Retinal Images in Africa (MoDRIA) is a comprehensive dataset developed to drive advancements in machine learning and artificial intelligence for ophthalmological research. It addresses a critical gap in the availability of diverse and representative retinal imaging datasets from African populations, helping to reduce bias and improve the accuracy of AI-driven eye disease diagnosis. MoDRIA combines high-quality retinal images with detailed clinical data, reflecting real-world clinical environments across the African continent. This resource provides a robust foundation for training, validating, and benchmarking AI models, ultimately supporting the delivery of more equitable and effective ophthalmological care.

Study Population

The dataset comprises high-resolution retinal images collected from patients across multiple eye care centers in Africa, accompanied by comprehensive metadata that includes demographics, clinical history, and diagnostic details. Each image has been expertly labelled by certified ophthalmologists, ensuring both diagnostic accuracy and research reliability. By capturing data from diverse patient populations across different African regions, the dataset provides a representative foundation for developing inclusive and unbiased AI models in ophthalmology.

Data Components
  • Retinal Images: High-resolution fundus photography
  • Visual Acuity: Comprehensive vision assessment data
  • Clinical Assessments: Diabetic retinopathy and other eye condition evaluations
  • Clinical Metadata: Patient demographics and medical history
  • Laboratory Results: Blood glucose
  • Treatment Records: Medication history and therapeutic interventions
Quality Assurance

All images undergo rigorous quality control procedures including standardized imaging protocols, expert validation, and comprehensive metadata verification. The dataset adheres to international standards for medical imaging research and patient privacy protection. Quality metrics include image resolution standards, annotation consistency, and clinical data validation protocols.

Research Applications

This dataset supports the development of AI-driven diagnostic tools for diabetic retinopathy, glaucoma, age-related macular degeneration, and other eye diseases. The focus on African populations helps address healthcare disparities and supports the development of more inclusive AI models for global ophthalmological care, particularly in resource-limited settings where early detection can prevent blindness. MoDRIA enables multi-modal research by linking retinal imaging, clinical assessments, laboratory results, and treatment outcomes for comprehensive ophthalmological research.

Data Dictionary - All 38 Variables

A list of all variables from the MoDRIA dataset with detailed specifications for ophthalmological research

Variable Name Description Data Encoding/Codes
IMG_IDImage ID numberImage identifier
Pt_IDPatient ID numberPatient identifier
CameraCamera type
IMG_OD_or_OSImage of the Right or Left eyeOD= right eye, OS = left eye
IMG_Disc_or_Mac_CenterImage with disc or macula centeredD= disc center, M= macula center
AgeAge
GenderGender0= male, 1= female, nc = not collected
VAOD_NumeratorNumerator of visual acuity, right eye e.g., 6/18e.g., 6/12 or text: HM = hand motion PL = perception of light, NPL = no perception of light, nc = not collected
VAOD_DenominatorDenominator of visual acuity, right eye, e.g., 6/18e.g., 6/12 or text: HM = hand motion PL = perception of light, NPL = no perception of light, nc = not collected
VAOS_NumeratorNumerator of visual acuity, left eye e.g., 6/18e.g., 6/12 or text: HM = hand motion PL = perception of light, NPL = no perception of light, nc = not collected
VAOS_DenominatorDenominator of visual acuity, left eye, e.g., 6/18e.g., 6/12 or text: HM = hand motion PL = perception of light, NPL = no perception of light, nc = not collected
IOP_ODIntraocular pressure, right eyeICARE
IOP_OSIntraocular pressure, left eyeICARE
WeightWeight
HeightHeight
BMIBMI
SBPSystolic blood pressuretype of device
DBPDiastolic blood pressuretype of device
Random_BSRandom blood sugarfinger stick, device
HTN_MedicationDoes patient take medication for hypertension?0 = No, 1=Yes
DM_MedicationDoes patient take medication for diabetes mellitus?0 = No, 1=Yes
HIV_MedicationDoes patient take medication for HIV?0 = No, 1=Yes
DM_MedicationDoes patient take medication for diabetes mellitus?0 = No, 1=Yes
No_MedicationDoes patient take no medications?0 = No, 1=Yes
Current_SmokerIs patient a current cigarette smoker?0 = No, 1=Yes, nc = not collected
Current_AlcoholIs patient a current drinker of alcohol?0 = No, 1=Yes, nc = not collected
OccupationWhat is patient's current occupation?0-10, nc = not collected
Image_Include?Should image be included or not included based on quality assessment?0 = No, 1=Yes, nc = not collected
Img_Quality_Issues?Does image have quality issues?0 = No, 1=Yes
IMG_Poor_Focus?Does the image have poor focus?0 = No, 1=Yes, -99 = intentionally not collected
IMG_Too_Dark_or_Bright?Does the image have illumination defects?0 = No, 1=Yes, 99 = intentionally not collected
IMG_Artifacts?Does the image have artifacts?0 = No, 1=Yes, 99 = intentionally not collected
IMG_ICDR_ScoreInternational Classification of Diabetic Retinopathy Score0 = no DR, 1 = micros only, 2 = microaneurysm, cws, or retinal hemorrhages (<25/quadrant), 3 = >25 hemes per quadrant, 4 = proliferative diabetic retinopathy, 5 = unable to determine
IMG_Macular_EdemaPresence of Diabetic Macular Edema0 = no DME, 1 = DME, 2 = unable to determine, 99 = intentionally not collected
C_D_RatioCup-to-disc ratio1 = 0.0-0.6, 2 = 0.7-0.8, 3 = 0.9-1.0, -99 intentionally not collected
Optic_Nerve_Normal?Is optic nerve normal?0 = No, 1=Yes, -88 = not collected, -99 = intentionally not collected
Quick_Qual_MetricImage quality metric from Quick Qual, probability of the image being "bad"link to QQ, 99 = intentionally not collected (image defect)
Fractal_Dimension_MetricFractal dimension metric from DARTlink to DART, 100 = intentionally not collected (image defect)
Data Dictionary

This represents all 38 variables from the MoDRIA dataset. The complete dataset includes variables for retinal imaging, clinical assessments, laboratory results, diabetes management, and research protocols. All variables are designed to support comprehensive ophthalmological research and clinical care, with particular focus on diabetic retinopathy screening and management in African populations.