Project Summary

The background to the award was the trial of new Artificial Intelligence by staff at Royal Hallamshire Hospital and University of Sheffield, looking to revolutionise breast cancer diagnosis. The project used the UKCA-approved PANProfiler Breast (PPB) tool which had the potential to revolutionise breast cancer diagnosis in the region, improve efficiency/speed of diagnosis, reduce delays and costs in the region, and provide data for wider adaptation across the country.

Following a period of project planning; quality improvement activity, agreement of a research project and training - retrospective and proactive testing was completed. Significantly more training and optimisation was required than originally envisaged. The testing showed that in some categories of samples the accuracy rate was high, whereas other areas were not as accurate as initially expected. Furthermore, the system lacked visual identification of areas assessed and some of the important relevant features and markers assessed in histopathological practice.  After careful consideration, and completion of a cost analysis, it was decided that in its current state the system needs more work and should not be rolled out to the wider region. The project team found piloting this technology an informative process, and lessons learned are being utilised by both the organisation and Provider.

Q&A

More feedback from Mr Khurram, Project Lead can be found below

Could you describe your project, including your aims?

  • The project aimed to test an AI tool which has been licensed for use in breast cancer (PPB). PPB has been reported to predict biomarker status in breast cancer tissue images (without the need for doing the test). We aimed to test it on breast cancer biopsies from the region to determine its efficiency, accuracy and cost-effectiveness.

Did your project meet anticipated aims/objectives?

  • Yes. We did manage to undertake the training, retrospective and prospective analysis as intended.

What were your key lessons learned?

  • Clinical validating and testing of such tools is essential even if they have CE and UKCA approval. There was significantly more work/time commitment for this work from our end than originally thought. However, the industrial partner was very approachable and worked with us to overcome some of the issues and challenges.

Will the project be continued – are there any plans for sustainability?

  • We are still in discussions with the provider to contribute to further development and optimisation of the AI tool, to improve results, and make it more clinically usable.

For more information on this project, please contact Ali Khurram s.a.khurram@sheffield.ac.uk