World Health Day: an introduction to machine learning in healthcare workshop


Earlier this year, I attended the well anticipated An Introduction to Machine Learning in Healthcare Workshop by RE-WORK. Having attended a number of their conferences in the past I was ready to be informed and inspired.

The workshop was set over an afternoon and lead by Nophar Geifman from the Division of Informatics, Imaging and Data Sciences at the University of Manchester. She kicked off with an introduction to machine learning: supervised; unsupervised; reinforced learning. She introduced us to types of neural networks; types of clustering; and when to use each. She also gave us a few examples of how AI is being used in the healthcare space. It was a great way to quickly get everyone in the room up to speed on the basics of the topic and whet our appetite for the rest of the afternoon.

After a cup of tea and some snacks to let our brains unwind we were introduced to Andreas Theodorou from the University of Bath. On top of his PhD, Andreas advises the government on AI to help develop legislation. The legislation is pretty much non-existent at the moment, allowing great creativity. Andreas spoke of the three qualities he believes are necessary for great AI legislation; accountability, responsibility and transparency. He believes that developers should be held accountable for their ML algorithms. This should prevents groups from blaming the ‘black-box’ for responsibility. In addition, ML algorithms should be transparent. By this, he didn’t just mean sharing the source code. He means exposing the machines goals, it’s processes, it’s sensory inputs, how reliable it is in the context, and producing error messages comprehensible for laymen. I couldn’t help but think how significant the overhead of transparency would be in product development, it would inevitably slow development. In addition, protecting IP would become a little more difficult. It’ll be interesting to see what the House of Lords will report on 31st March 2018, and whether Andreas’ ideas shape policy.

Our next talk was given by David Clifton who runs the Computational Health Informatics Lab at Oxford University. It was clear from his talk that with the amount of available related healthcare data; vital signs, lab results, genomics, prescriptions, diagnosis, wearables, and the significant increase in processing power, make it a ripe time for deep learning.

He gave an example of some research from the University of Oxford on TB’s drug resistance. By taking a sample of TB, using a DNA sequencer, they could map the TB’s genetic similarity to other TB strains they knew of. They could then predict to a high degree of accuracy which medication the TB was already resistant to. This cut down treatment time from 2 weeks, to 24 hours. It was a great example.

Lastly, Maria Chatzou stepped up to talk to us about genomics. Maria is the Founder and CEO of Lifebit who analyse genomics and bio data quickly and cost effectively. The cost per genome has plummeted massively. It is expected the by 2020, 500 million genomes will have been sequenced. Alongside all the data David was speaking of, we have a wealth of knowledge in the data to be untapped. What’s more, this leads to the possibility of widespread adoption of personalised medicines; moving from symptom based medicine to algorithm based medicine.

The floor then took questions. One which inevitably came up was whether AI would take over doctors. There was an overwhelming sense that it would not, at least in our lifetime. At the moment it’s taking away the grunt work that doctors don’t want to do. It’s improving the analysis of patients so that we can get doctors to the bedsides of patients before they’re needed, with the information they need. It can enable more informed, preventative care, it has the capacity to reduce the cost by helping in the diagnosis, understanding which treatment will be most effective, and cutting down time to review tests. A poignant comment from a doctor in the room was that, as a doctor, part of his role was persuading a patient to take a certain drug, or adopt a change in lifestyle. Those soft skills are extremely complexed to build in a machine, and so the human element is likely to remain a human skill for a while.

In all, the event was a blast. I left feeling inspired and keen to get involved. Thank you to Nikita, and the fantastic team at RE•WORK for the workshop.

In the spirit of World Health Day, RE•WORK are currently offering complimentary access to video presentation PLUSS Pass upgrade from 2 of the following events, when you purchase a full price pass to one: Deep Learning in Healthcare Summit Boston 24 - 25 May, AI in Healthcare Summit Hong Kong 06 - 07 June, Deep Learning in Healthcare Summit London, 20 - 21 September. Offer available when you register before Friday 13 April.


Healthcare Deep Learning in Healthcare Summit World Health Day


    As Featured In


    Partners & Attendees

    Ibm watson health 3.001
    Rbc research.001
    Kd nuggets.001
    Mit tech review.001
    Maluuba 2017.001