AI bioresource Building datasets to develop AI for healthcare

The challenges for transforming diverse databases from multiple imaging modalities and electronic patient records to a curated, computationally tractable form with robust metadata are significant. The reading centre has developed unique expertise in data curation of large datasets and the on-going work involves:

  • A wide-ranging dataset of structural and functional ophthalmic diagnostic tests, that are converted from proprietary to open source file formats, linked with metadata obtained automatically from the Moorfields electronic health record with quality assurance process from Moorfields reading centre graders.
  • An associated framework so that all data subsequently obtained from patients with pre-specified ophthalmic diseases are automatically ingested in the bioresource.
  • A graphical user interface, plus robust methods for anonymisation and bulk export of datasets – with ethical and information governance oversight – to facilitate collaborations.