Medical professionals, data scientists, entrepreneurs, researchers and others came together on 19 September for the second meetup of the Amsterdam Medical Data Science Group to discuss how medical data can be used to benefit patients and improve medical practices.
Taking place at the Intensive Care Unit (ICU) at Amsterdam UMC, attendees took a deep-dive into AI and machine learning to explore how both could help shape the future of medical care, and to share projects, ideas, papers and the latest developments in the field.
Shaping the future of medical care
The monthly meetup, which aims to connect medical professionals with data scientists, features project leaders and presentations from experts in medical data science. Bringing these groups together in an informal setting is key to overcoming some of the main challenges involved in improving a patient’s treatment and medical care with data science.
‘Medical professionals and data scientists who are interested in projects around medical data don’t tend to know each other very well, and we want to solve that,’ explains Dr Paul Elbers, an intensivist at UMC and one of the event’s organisers. “Hosting the meeting in the hospital also makes it easier for doctors to attend.’
After an introduction from Dr Elbers in which he introduced the four Ds of data science: Data, Data scientists, Developers and Doctors, and a short talk by fellow organiser Dr Lucas Fleuren, the presentations began.
Collaborating to create new applications
Amsterdam startup Pacmed and the Amsterdam UMC are currently collaborating to investigate how machine learning can help medical professionals to do their job as efficiently as possible, by using applications as a guide to provide the best treatment possible for patients. Dr Patrick Thoral and Pacmed’s Hidde Hovenkamp introduced this collaboration. They discussed a machine learning model they are developing together to help assist doctors to make a discharge decision in the ICU. Using the enormous amount of data that’s generated during a patient’s care and past readmission, this model provides doctors with a readmission probability at the moment of (possible) discharge for each patient.
It’s hoped that this model can be used to reduce the chances of a patient being readmitted or developing health complications further into their recovery, in turn helping to reduce costs and staff workloads in hospitals. “It’s sometimes difficult to predict whether the patient is well enough to transfer to the general ward,” explained Dr Thoral. “We wanted to develop a model to predict the probability of readmission rate after discharge from the ICU, using data that’s routinely available so there’s no extra workload for nurses and physicians.” In the coming months, an implementation study using the model is due to start, and a paper will be published about the project.
A deep-dive into machine learning
Presented by VU University PhD student Amin Tabatabaei, the next part of the session focused on a particular type of machine learning: reinforcement learning. Tabatabaei explained how reinforcement’s key principles of taking actions in an environment to maximize a reward could be applied to a medical setting to work out the best way to treat
patients. He also offered explanations about how this process could be refined using different techniques to improve the efficiency of the way the models worked. Tabatabaei also introduced the two other types of machine learning (unsupervised and supervised), in order to explain the benefits and challenges associated with each. Tabatabaei used the example of a health application which uses a machine learning model to work out the best time to send messages to users to prompt them to exercise. In this example he showed how the data could be ‘pooled’ in order to get better results faster, or ‘personalised’, which would take longer to develop but would attain better results.
Doctors, data scientists and pizza
The event ended with the group sharing ideas, stories and connecting over pizzas and soft drinks which had been provided by the Amsterdam Economic Board. ‘It’s only the second meeting but the balance between medical professionals and data scientists has been good,’ Dr. Elbers said. “It’s brilliant that the data that we have is now being used, and something like the machine learning model Pacmed is developing shows that clearly. It’s very important to get things right, but what we can see is that people are getting more used to being supported by decision-support systems, and doctors are becoming much more open to using them in their duties.”
The Amsterdam Medical Data Science Group meetup takes place on the third Tuesday of every month in the Delta Room at the VU University Medical Center Amsterdam’s Intensive Care Unit. The next meeting will take place on 16 October.
To find out more visit: https://www.meetup.com/amsterdam-medical-data-science.
For more reports from previous Medical Data plus Pizza Meet-ups , click here.
The Amsterdam Medical Data Science Group meetings are supported by The Right Data Right Now consortium, which includes Amsterdam UMC, OLVG, Vrije Universiteit, Pacmed, and the Amsterdam Economic Board.