OLIVES: Ophthalmic Multimodal Dataset with Biomarkers and Time-Series Treatment

1268 fundus eye images, each paired with 49 OCT scans, 16 biomarkers, 3 clinical labels, and a DR/DME diagnosis. The dataset contains data from 96 eyes tracked over at least two years, with each eye receiving an average of 66 weeks of treatment and 7 injections. It was introduced by Mohit Prabhushankar of the Georgia Institute of Technology to research relationships between clinical labels, biomarkers, fundus images, and OCT scans.

Use Cases

• Multimodal disease prediction based on the combination of fundus images, OCT scans, and clinical labels.
• Biomarker analysis and interpretation based on the 16 vectorized biomarker attributes.
• Longitudinal treatment outcome modeling based on time-series patient data over an average of 66 weeks.
• Self-supervised learning for medical imaging based on described medically-grounded augmentation schemes.

Strengths

• Includes four aligned data modalities: fundus images, OCT scans, biomarkers, and clinical labels.
• Provides longitudinal data for 96 eyes tracked over a period of at least two years.
• Contains 1268 fundus images, each associated with 49 OCT scans, offering paired imaging data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count for the complete multimodal dataset is unspecified, which may limit suitability assessment.

Provenance

Source

Georgia Institute of Technology (author: Mohit Prabhushankar), sourced from paperswithcode.

Collection Method

Likely collected from associated clinical trials, as mentioned in the description.

Time Range

Treatment period averaged over at least two years per patient.