'Data and context are key for robust AI in healthcare'

Giving evidence on Artificial Intelligence to a House of Lords Select Committee, PHG Foundation's Dr Sobia Raza highlighted the need for an NHS-wide strategy on using health data for algorithm development to realise the benefits of AI for patients.

Following written evidence from PHG Foundation submitted to a House of Lords inquiry on Artificial Intelligence, PHG Foundation was invited to give oral evidence on the implications of AI for healthcare to the Select Committee on 21 November.

In evidence to the House of Lords Select Committee on Artificial Intelligence (AI), PHG Foundation’s Head of Science, Dr Sobia Raza highlighted the potential value of an NHS wide strategy on using health data for algorithm development to realise the potential of AI for patients and to ensure data sets are sufficiently representative of the UK population. The committee is considering the economic, ethical and social implications of advances in artificial intelligence.

Responding to a question on how to ensure the robust evaluation of machine learning tools before they are used on patients, Sobia used the analogy of driverless cars, which may have been trained and then successfully tested on broad open highways in California but are faced with a very different set of decisions on narrow British country lanes.

She said ‘It’s the same in healthcare – the AI algorithms have to be tested under the conditions and the population in which they will be used.’

Discussing the development of policy around AI, Sobia emphasised the need for appropriate digital infrastructure in relation to storing, collecting, processing and transferring data - and the importance of retaining a focus on building patient and public trust for sharing health data. In order to gain access to large, representative data sets for AI development, it is key that there is greater engagement on the potential benefits and more widely on why data is vital to improving health, care and services.

Watch the evidence session here



Read more

Looking for something specific?