Artificial Intelligence (New website coming soon)
30/05/2019 / Artificial Intelligence
Software to help mammography services run more efficiently and to check mammograms are being developed by companies working with NHS partners.
Faculty AI are focusing on the operational aspects of breast screening services, including: a breast screening round-length and optimisation tool; an intelligent clinic and staff scheduling tool; and a smarter clinic booking tool.
Kheiron Medical have developed a CE marked Mammography Intelligent Assessment tool (Mia). If successful, it will act as the second reader in a breast screening workflow, and EMRAD hope to move to a prospective model running alongside the human-led process.
Simon Harris, EMRAD project manager, said, “Ultimately, if we can deploy these two pieces of AI and scale up within EMRAD and hopefully the wider NHS, we will make a positive difference to how breast screening services are run and support the severe workforce challenges that the NHS is currently facing.”
This article was first published on the Society of Radiographers website, for more information please go to https://www.sor.org/news/new-ai-tools-breast-screening-services.
In 2018, EMRAD formed a partnership with two UK-based Artificial Intelligence (AI) companies, Faculty and Kheiron Medical, to help develop, test and - ultimately - deploy AI tools in the breast cancer screening programme in the East Midlands. The project is one of seven ‘wave two’ NHS Test Beds, and is administered by NHS England and the Office for Life Sciences. The Test Bed project is focused on Capacity, Care, and Confidence: it aims to improve and optimise clinical service capacity, to enhance patient care at significant scale and to increase NHS confidence in the utilisation of innovative machine learning tools.
The project aims to develop and test both non-clinical (operational) and clinical AI tools. Faculty’s ‘Platform’ software has the potential to help optimise operational processes such as clinic scheduling and staff resourcing, while Kheiron’s ‘Mia’ tool has the potential to support the clinical workforce issues in the service by acting as the second reader in the dual-read mammography workflow.
Non-clinical (operational) applications of AI – Faculty
The Test Bed project is seeking to apply AI tools and techniques to the operational and administrative aspects of the breast screening programme, considering how AI can help to run the service in the most efficient and effective way possible.
Faculty’s role in the Test Bed is to test the potential application in the breast screening programme of process optimisation tools and techniques developed on their ‘Platform’ software and aims to deliver exactly the same breast screening programme in a more efficient way (see Fig. 1). For example, we will be training and testing an AI-powered intelligent capacity and demand planner. This should help screening round managers to more easily and accurately identify likely pinch-points where demand exceeds the service’s capacity, to identity ways to mitigate these pressures, and to simulate and estimate the likely knock-on impact of – for example – unplanned machinery or site down-time, or workforce changes, on the maintenance of the round length. We will also be testing a dashboard which aims to help programme managers identify their most and least efficiently used clinics, helping to best target constrained staff resources.
Fig. 1 - An AI-powered planning and scheduling tool
At the June EMRAD AI Project Board, the Lincolnshire Breast Screening Programme lead described the pressures as “tremendous”, with “no end in sight”. Deep learning is increasingly being proposed as a solution to the breast cancer screening workforce crisis. As part of their role in the Test Bed, Kheiron are conducting a large-scale retrospective study on mammograms from two of the NHS sites within the EMRAD Consortium. The aim is to test the generalisability of Mia™, their novel deep learning mammography software. They hope to conclude that their generalisable model is suitable for consideration as an independent reader in double-read screening programmes (see Fig. 2). This would have significant implications for the future of the breast screening workforce throughout the UK.
Fig. 2 - How Mia™ fits into the double-reading breast screening workflow
Realising all these opportunities will depend significantly on the availability and accessibility of the data which is required to train AI tools. Finding ways to securely and safely share appropriately de-identified client-level data for the purposes of service improvement will be vital. Building functional, modern connectivity into the core breast screening programme software will also be essential: programme managers already have to contend with a variety of IT systems, so core service delivery systems will need to be able to connect seamlessly with new AI tools if those tools are to be deployed effectively in the real world.