Testing breast cancer mammography screening artificial intelligence algorithms using the Cambridge Cohort database; study design, methods and statistical analysis
Two-million women aged between 50-70 are screened for breast cancer every year in the UK and each mammogram is read by two expert readers. Screening is therefore a labour-intensive repetitive task which could be improved through the use of Artificial Intelligence (AI), to automate screen reading or through priority triage of cases that could be cancer. With the aim of improving patient outcomes and screening efficiency.
There are now over five FDA approved algorithms as well as numerous academic algorithms that have been developed for either computer aided detection and diagnosis (CADe and x) or computer aided triage (CADt) approaches. However, all the current literature is from retrospective studies using cancer enriched cohorts, with limited research investigating the use of AI in the UK screening programme. We have created a mammographic imaging database which will be used to independently and systematically test mammography AI algorithms from institutions world-wide to evaluate performance as well as provide data for ongoing development and prospective testing.
This seminar will cover the study design, methods and statistical analysis planned for future testing using this database as well as provide a forum to discuss key areas of analysis that should be addressed as part of this testing.
Meeting ID: 925 7540 3744