Rapid virtual FFR assessment using deep learning

  • Research type

    Research Study

  • Full title

    Rapid virtual fractional flow reserve assessment using deep learning

  • IRAS ID

    238438

  • Contact name

    Jack Lee

  • Contact email

    jack.lee@kcl.ac.uk

  • Sponsor organisation

    King's College London

  • Clinicaltrials.gov Identifier

    n/a, n/a

  • Duration of Study in the UK

    2 years, 11 months, 31 days

  • Research summary

    Research Summary
    Coronary heart disease is the leading cause of death and the most common form of cardiovascular disease in the UK. The clinical decision of whether to re-open a diseased coronary vessel rests centrally on the flow-limiting severity of the lesion. To this end, research over the past decades have shown interventions guided by pressure-derived Fractional Flow Reserve (FFR) to significantly reduce major cardiac events, compared to using the angiographic appearance alone. Nevertheless, clinical uptake of FFR has remained low due to increased procedure time and the risk of complications. This project will develop and clinically evaluate a real-time virtual FFR (vFFR) assessment strategy to directly address these shortcomings in a less invasive manner. By integrating computational fluid dynamics (CFD) with deep learning, a powerful AI technique, the time-consuming and expertise-intensive parts of the vFFR work flow will be automated, while the diagnostic accuracy will be maintained by training the scheme to high fidelity CFD simulations. The pilot trial of the new method will demonstrate its potential impact to reduce patient risk and procedure time, with implications for improved CHD management in the clinics via vFFR.

    Summary of Results
    This study set out to develop a fully automated computational method to process x-ray angiography images of coronary vessels, and then to estimate a clinical index (Fractional Flow Reserve) which is used to determine whether a patient should receive revascularisation. The patient data collected were intended to be used in a final test for the algorithm. Due to the pandemic, the patient recruitment experienced severe difficulties and the plan was modified to use previously collected data (n=100). Currently, patient recruitment has ended, but the computational part of the project is ongoing. We have developed an accurate AI-based method that can calculate the clinical index given a coronary geometry, and several approaches by which to reconstruct the coronary anatomy are being evaluated.
    Has the registry been updated to include summary results?: No
    If yes - please enter the URL to summary results:
    If no – why not?: As outlined in the original project plan, the principal focus of this study was to develop and evaluate the computational techniques for cathlab diagnosis, which will be used in a follow-up clinical trial. The patient data collected in this observational study was strictly meant for technique validation and did not intend to answer a specific clinical question. This was made clear to the patients at recruitment time. Several publications on the computational technique development have been published (that does not use patient data) and the final summary publication involving the patient data is currently under preparation.

  • REC name

    North West - Preston Research Ethics Committee

  • REC reference

    19/NW/0332

  • Date of REC Opinion

    5 Jul 2019

  • REC opinion

    Further Information Favourable Opinion