New to Meducation?
Sign up
Already signed up? Log In

CPD Tracker



Machine Learning in Medicine.

Are we about to be taken over by machines?

Written by Dr Alastair Buick · Thursday 13th September 2018

The first key step in realising any new technology in healthcare is communicating its value to both patients and healthcare staff. This can present a challenge when the discussion is around Machine Learning (ML) and Artificial Intelligence (AI) due to the associations, and often, misunderstandings, many people carry surrounding these terminologies. Nevertheless, informed patient consent in this area is vital and only fully informed when a decision has been made with the weighing up of accurate information. This is especially important in the issue of data ownership, where there can be confusion about what the implications of consent to 'use of data' means in healthcare today.

But if we can get beyond this issue of understanding and demystifying concerns about ML and AI then the vast benefits for modern medicine and the NHS are apparent; improved efficiencies in practice and resource management, enhanced precision and personalised medicine, more accurate diagnoses, treatments and prognoses.

Of course there are some concerns too that if ML algorithms are recognising patterns in large data sets and then spotting these same trends in new data then the alogortihm is only as good as the data it recieves. Currently the NHS IT infrastructure is fragmented and multi-layered which may pose challenges for acquiring reliable usable data. There may also be problems with algorithms overfitting to patterns in data and application may not suit the different context of primary and secondary care. Algorithms may also make mistakes, and keeping track of the reasoning and logic behind their outputs may be difficult and may limit how much their credibility and acceptance is seen in practice.

This technology will undoubtedly disrupt ways of working in modern medicine - but as human oversight is vital it can only be successful when applied in collaboration and partnership with clinicians and healthcare staff. For example, the role of a radiologist will likely change to more of a supervisor overseeing algorithms reporting hundreds of scans and clarifying discrepancies, but then, new roles and responsibilities will emerge in these new ways of working...more interventional working, or perhaps even more patient-facing care?

Underpinning all of this is of course the issue of accountability and where liability sits with the use of such technology. Will the clinician be responsible? Or the manufacturer? Or the healthcare provider? Whatever the end-point it seems now is a crucial time to bring together clinicians, industry and policy makers to support the development of governance regulation and standards to ensure the potential of this technology can be realised, but also trusted by society.

Original post from Technical Health

Original post written by George Brighton