AI- is: Part 1

• Research type

  Research Study

• Full title

  Artificial Intelligence - Informed Screening: Part 1

• IRAS ID

  288182

• Contact name

  Sue Astley

• Contact email

  sue.astley@manchester.ac.uk

• Sponsor organisation

  University of Manchester

• Clinicaltrials.gov Identifier

  N/A, N/A

• Duration of Study in the UK

  0 years, 6 months, 17 days

• Research summary

  This research aims to develop a method for identifying screening mammograms in which early signs of cancer are difficult to detect because of the pattern or quantity of radio-dense breast tissue. There is evidence that a proportion of breast cancers are missed by radiologists interpreting screening mammograms; some of these cancers are detected between screens but can be seen retrospectively or identified using other screening modalities. There is a need for an accurate method of identifying women who would benefit from additional imaging because there is a risk of cancer being missed. A method will be developed by training Artificial Intelligence algorithms to recognise difficult-to-interpret mammograms and evaluating performance with an without an additional physics-based software that quantifies and locates regions in which cancers might be obscured. In this initial stage image data will be retrospectively collected from that held at the Greater Manchester Breast Screening Centre from women who participated in a previous research study PROCAS(Predicting Risk of Cancer at Screening). These digital mammograms were previously pseudonymised at the screening centre. Cancers will be annotated for this research by an expert breast radiologist at the centre. The pseudonmymised images and annotations will then be transferred to the University of Manchester for software development and testing. The physics software will be developed by a company (Volpara) using images they already hold, and run on computing facilities at the the University. The AI algorithm will be trained on mammograms of women who developed a cancer before or at their next screen, mammograms in which cancer was detected, and mammograms in which no cancer was present at a subsequent screen. For cancer cases we will focus on the area of the image where the cancer developed. We anticipate that the software will take 12 months to develop and evaluate.

• REC name

  South East Scotland REC 01

• REC reference

  20/SS/0121

• Date of REC Opinion

  4 Nov 2020

• REC opinion

  Favourable Opinion