By Lesley Smart, Senior Analytics Consultant, GE Healthcare
Behind every successful Artificial Intelligence (AI) implementation lies a sea of rich and bountiful data. . It is the data that enables the floating of a deep learning idea and buoys the building of innovative man and machine intelligence to optimise clinical services, overcome staff shortages and enhance patient care.
Never before has the importance of data readiness been so key in digital health.
The importance of data readiness for AI success
Let’s look, by way of an example, at the real-world testing of a certified AI app in breast cancer screening being run at The East Midlands Radiology Consortium (EMRAD), a UK collaboration of seven NHS Trusts running 11 hospitals and supporting more than five million patients across the region.
Currently, mammograms are double read by radiologists and specialist reporting radiographers, but the growing shortage of these clinicians and the skills they possess is delaying patient results and putting pressure on operational teams.
A solution to automate elements of the breast screening workflow at EMRAD – via an intelligent software solution – is at testing stage and promises efficiencies to deliver faster patient diagnosis and avert a potential radiology workforce crisis.
The first phase testing uses historic, anonymised images (sourced from emrad’s immense data stores) to assess the accuracy of the AI app to flag potential cases of breast cancer. The results are then being tested for accuracy when compared against the human accuracy of the breast screening reading team. If deemed safe – it is performing better than humans to date – the next stage will be to clinically test the system within the workflow for first read scans before they are reviewed by the clinical experts.
This example highlights that to realise the opportunities of AI depends significantly on the availability and accessibility of data sources that AI tools can be trained and tested upon.
Float gracefully in the health data lake, rather than sink in a data swamp
The opportunities that AI promises in healthcare abound. But creating AI value for all requires scale and currently this is being hampered by the barrier of not having a defined and clear strategy to source the data that AI feeds upon.
Indeed, a recent survey1 of cross-industry organisations indicated that only 18% had a strategy to source data for AI and only 15% had the right technological infrastructure and architecture in place to support the development and implementation of AI systems.
So, to swim strongly towards AI success requires a flowing and healthy data lake governed by a data access strategy and sourced by a robust digital framework and scalable data repository. This approach unlocks the value and veracity of your data safely and securely so that you can maximise the innovation that AI will bring to your staff and patients. Without it, you’ll simply be stuck in a data swamp.
1.McKinsey, The Adoption of AI, Nov 2018 [https://www.mckinsey.com/featured-insights/artificial-intelligence/ai-adoption-advances-but-foundational-barriers-remain]