Automated identification of metallic implants on x-rays

  • Research type

    Research Study

  • Full title

    Automated identification of metallic implants on x-rays with neural networks

  • IRAS ID

    277725

  • Contact name

    James Howard

  • Contact email

    jphoward@doctors.org.uk

  • Sponsor organisation

    Imperial College London

  • Duration of Study in the UK

    2 years, 0 months, 1 days

  • Research summary

    Approximately 200,000 patients per year in the UK have either metallic hip or metallic knee implants inserted, for a variety of clinical reasons. Patients can develop issues associated with their metallic implant and require further surgery to rectify this. Revision surgery requires different specialist equipment for different makes and models of metallic implant.

    Patients often do not have the information with them to tell physicians what make and model of implant device they have. If they had the implant at a different hospital or too far in the past, it can be difficult to identify. Difficulties identifying implants preoperatively can cause delays to treatment, consume additional clinician time, and can complicate surgery, all of which increase costs. However, it is possible to work it out from a simple hip or knee x-ray which can be performed in any emergency department in the UK. The steps for identification from an x-ray can use a flowchart which can be difficult and time-consuming, but we believe this could be done more quickly and accurately by a computer.

    Recently, people have utilised a form of computer science called 'machine learning' to try and help in analysing medical images. It works on the idea that a computer trains itself to recognise certain images by seeing enough examples. Our team have experience in using machine learning previously, and we are confident it could provide us with an easy and fast system for identifying metallic hip and knee implant makes and models.

    Imperial College NHS Trust implants several hundred metallic hip and knee implants per year, and the make and model is recorded as part of routine clinical practice. Furthermore, all patients post-implantation of a metallic implant have an x-ray for routine clinical reasons. I plan on examining the appearances of these metallic hip and knee implants on x-rays to see if a computer can learn to distinguish them.

  • REC name

    N/A

  • REC reference

    N/A