DeepMind Health and research collaborations
Each scan and test result contains crucial information about whether a patient is at risk of a serious condition, and what needs to be done. Being able to interpret that information quickly and accurately is essential for hospitals to be able to save lives.
AI systems could be hugely beneficial in helping with this process. Rather than programming systems by hand to recognise the potential signs of illness, which is often impossible given the number of different factors at play, AI systems can be trained to learn how to interpret test results for themselves. In time, they should also be able to learn which types of treatments are most effective for individual patients.
This could enable a series of benefits across healthcare systems, including:
- Improved equality of access to care. Demands on these healthcare systems are felt more acutely in certain areas of the world, and even within certain departments in hospitals, than others. If we can train and use AI systems to provide world-class diagnostic support, it should help provide more consistently excellent care.
- Increased speed of care. We hope that AI technologies will provide quick initial assessments of a patient to help clinicians prioritise better, meaning patients go from test to treatment faster.
- Potential for new methods of diagnosis. AI has the potential to find new ways to diagnose conditions, by uncovering and interpreting subtle relationships between different symptoms and test results. In theory, this could lead to even earlier diagnosis of complex conditions.
- Continual learning and improvement. Because AI tools get better over time, they will also help hospitals to continually learn about the approaches that help patients most.
We're still in the early stages of AI research in health, and our work is based on deep collaboration with our clinical partners.
We currently work with several hospitals to combine their clinical expertise with our machine learning technology, and together we ensure that the research we undertake will have practical benefits and save lives. The results of this work will be subject to rigorous clinical scrutiny, and will be published in peer-reviewed academic journals. The first of these, which related to our collaboration with Moorfields, was published in August 2018 in Nature Medicine.
Today our research is centred on a technique called machine learning. Our algorithms interpret visual information in the form of de-personalised images (head and neck scans at University College London Hospitals NHS Foundation Trust, eye scans at Moorfields Eye Hospital NHS Foundation Trust, mammograms with the Cancer Research UK Imperial Centre) and de-personalised medical records with the Department of Veterans Affairs in the US. The algorithms learn how to identify potential issues within these images and results, and how to recommend the right course of action to a clinician.
As the algorithm processes more images, it refines its understanding and interpretation of the information. It then provides increasingly useful feedback, and segmentation, of the data for the clinicians to use for better diagnoses and treatment.
All research projects go through rigorous regulatory and Trust approvals and are conducted only on non-identifiable patient data. You can read more about the healthcare research permissions process on our data and security page.
For more details about our research collaborations with Moorfields, UCLH, the Cancer Research UK Imperial Centre and the Department of Veterans Affairs, please see the links below.