Assessment of nasogastric tube placement using deep learning

  • Research type

    Research Study

  • Full title

    Application of deep learning to chest radiographs for the assessment of nasogastric tube placement

  • IRAS ID

    229082

  • Contact name

    Dermot Mallon

  • Contact email

    d.mallon@imperial.ac.uk

  • Sponsor organisation

    Imperial College NHS Trust

  • Clinicaltrials.gov Identifier

    17SM4053, Sponsor reference

  • Duration of Study in the UK

    1 years, 11 months, 31 days

  • Research summary

    The first aim of this study is to develop a deep learning algorithm to identify incorrectly positioned nasogastric (NG) tubes. NG tubes are commonly used to provide enteral intake to acutely unwell patients. Nasogastric tubes are generally inserted at the bedside without specialised guidance equipment. Feeding through an incorrectly placed NG tube is a so-called “never event“ due to the potential to cause significant morbidity and mortality. Therefore, numerous bedside and radiological tests are performed to ensure satisfactory positioning prior to use. Despite these precautions, there remain cases where the position of an NG tube is incorrectly assessed on the chest radiograph.\n\nDeep learning represents a potentially fast, cheap and reliable method of determining NG tube position. The deep learning algorithm will be trained on chest radiographs that have already been performed and assessed by a radiologist at our centre. The accuracy of the the deep learning algorithm will be compared with assessment by clinicians and radiologists.\n\nUsing the methods developed in the above study, the ability of the deep learning algorithm when applied to other radiological investigations to predict further clinical outcomes such as length of stay and mortality, which cannot be determined by the radiologist, will be assessed.

  • REC name

    South West - Frenchay Research Ethics Committee

  • REC reference

    17/SW/0210

  • Date of REC Opinion

    19 Sep 2017

  • REC opinion

    Further Information Favourable Opinion