Rapid Access Macular Screening and Evaluation (RAMSEs): a datasheet for a dataset to diagnose and triage macular diseases

AbstractPurpose Macular disease can cause significant visual morbidity. Timely and accurate diagnosis and management is paramount. However, there is a lack of high-quality multimodal datasets reflective of real-world clinical practice to support research in this area. We present a large longitudinal real-world dataset that aims to address this gap. Design Datasheet describing a multimodal dataset of routinely collected ophthalmic data focusing on possible macular disease. Participants Adult patients (aged ≥18 years) attending the retina service at Moorfields Eye Hospital NHS Foundation Trust for the first time from 1 February 2017 to 31 July 2025 with macula-centered optical coherence tomography (OCT) scans (Topcon or Heidelberg). Methods Data was curated from the INSIGHT Health Data Research Hub, one of the world's largest ophthalmic imaging bioresources, which aims to provide researchers with controlled access to anonymised routinely collected data. Clinical and imaging data derived from routine clinical care were exported, processed, and de-identified for secondary research use. Main Outcome Measures This datasheet describes the demographic, clinical, and imaging metadata of the dataset, including a transparent overview of its strengths and weaknesses. Results RAMSEs (Rapid Access Macular Screening and Evaluation) is a large and diverse dataset specifically designed to facilitate the diagnosis and triage of possible macular disease by providing access to multimodal imaging and clinical metadata. The current version of this dataset (time-locked as of July 2025) consists of retinal images and linked demographic and clinical metadata from 85,444 patients with a median age of 63 (IQR 50-75) and fairly even gender distribution (52.2% female). It comprises over 5.1 million multimodal ophthalmic images (e.g. colour fundus photographs, ultra-widefield imaging, autofluorescence), including over 1.3 million macula-centered OCT scans, together with linked clinical and sociodemographic metadata. Conclusion We have developed a large multimodal real-world dataset, which was designed to address existing gaps in dataset size, disease distribution, and key clinical and sociodemographic metadata for suspected macular disease. This valuable longitudinal resource can serve multiple purposes, including the development and/or robust clinical validation of AI models, facilitating insights into real world patient pathways and outcomes, or enabling research relating to epidemiology or big data analytics. This dataset may be made available through INSIGHT via a structured application process.