Text Classification Problems in Healthcare

• Research type

  Research Study

• Full title

  Text Classification Problems in Healthcare

• IRAS ID

  246361

• Contact name

  Nir Oren

• Contact email

  n.oren@abdn.ac.uk

• Sponsor organisation

  University of Aberdeen

• Duration of Study in the UK

  0 years, 11 months, 30 days

• Research summary

  Summary of Research
  In health care data systems, free text is common: letters, x-ray reports etc. These electronic health records hold a wealth of important clinical data. These data are difficult to access for managing health care systems, planning care and for research.\n\nIt is possible to train a computer to read text, identify key information and classify this information in a more structured way that helps clinical care planning and enables information to be found more easily for managing health care systems and research. This process is called natural language processing (NLP).\n\nThis project aims to train a computer to identify key information from X-ray reports and clinical letters. It represents a vital first step in using these data to improve health and health care. For example, to create summary information about an episode of care that would help the hospital doctor efficiently report care to a patient’s GP; to summarise X-ray reports so that a doctor or research can efficiently find all patients with, for example, a hip fracture for a study or to audit and improve care.\n\nAll real patient data will have identifying information removed and will be handled in the accredited secure data safe haven. \n

  Summary of Results
  The project was delayed initially as we were unable to recruit a suitable student and then due to the COVID-19 pandemic. The project has since been superseded by a new data study.

• REC name

  North of Scotland Research Ethics Committee 1

• REC reference

  18/NS/0109

• Date of REC Opinion

  2 Oct 2018

• REC opinion

  Favourable Opinion