Calling all data scientists

The Medical Data plus Pizza Meeting is hosted by the Right Data, Right Now Consortium, which unites the hospitals UMC and OLVG with VU University, local MedTech company Pacmed and Amsterdam Economic Board. The monthly event is a type of speed-data-dating, which hooks up the medical community with data researchers, data managers, policymakers, philosophers and legal experts.

As is traditional, the 13th meeting on 15 October opened with a rough headcount. The split between doctors and data scientists is usually quite even. But tonight, the number of attending number crunchers vastly outnumbered the medical professionals.

The state of AI

Luca Roggeveen is a PhD student at Amsterdam UMC’s intensive care unit. He was not only the meeting’s second presenter but also the host. In his hosting duties, he gave a rundown of recent medical data news. As preparation, he googled “AI, where are we now?” and saw that the question was already being asked in the 1990s.

“We haven’t actually come that far. But in the medical domain, we’re doing quite well in applying AI,” says Luca. “And the Dutch are doing quite well. Only the Nordic countries are doing better than us. Which, to be honest, is quite typical.” The audience chuckles.

“And there’s more good news. While I think Boris Johnson won’t be PM for long, he’s done one good thing in his short time. He pledged 250 million pounds to AI for the National Health Services,” says Luca. When it comes to number crunching and pizza munching, a sense of humour seems a prerequisite.

“And that’s the direction we’re going,” says Luca. “As a result, we’ll be seeing more of this – a lot more investments in the healthcare market. Indeed, some say by 2025 we’ll have ten times the investments we have now.”

So, what is AI exactly?

Meanwhile, the term “AI” has become largely meaningless for the first presenter Zoltán Szlávik, even though he’s AI Research Director at myTomorrows. “I’ve been called many things during my career,” says Szlávik as an opening to his witty deep dive into ‘Learning to Understand Patient Language’. “I began studying computer science, and then became an information retrieval specialist, a data miner, a data scientist and then an AI researcher – even though my job has not really changed that much.”

myTomorrows is a pharma-meets-tech company out to confront unmet needs by connecting incurable patients with treatments in development across the world. Currently, there are 10,000 diseases with only 500 treatments – however, countless treatments are in development. And it’s potentially a win-win situation with the patient’s data contributing to the inevitably long development and licencing process. Consequently, one of the main challenges is how to make fast decisions on whether early access is possible for a patient. Time, after all, is of the essence.

One project Szlávik and his team are working on is how to ascertain the source of the information from vast amounts of unstructured data. To determine if it came from doctors, administrators or patients. This differentiation could be very important for when doctors make decisions.

To automate this process, Szlávik’s team is not only using deep learning-based natural language processing (NLP). They also use old school linguistic-based embedding (knowing a word by the company it keeps – sun, moon, stars) or the next-level NLP of Google’s new-yet-already-ubiquitous BERT (Bidirectional Encoder Representations from Transformers).

“Combine! If something works better, go for it!”

ICU data and the battle ICU readmission

In his presentation ‘Machine Learning at the Intensive Care’, Roggeveen explored the present and future of developing predictive models that can make patient monitoring automatic. These models aim to minimise a patient’s chances of readmission to ICU – a point when survival chances plummet.

For example, cameras are now being used to see if a patient is getting out of bed. “But what happens if we quantify a patient’s actions to monitor if they are actually getting better,” wonders Luca. “We can also use reinforcement, supervised or unsupervised learning to look at the sequence of medical decisions and their outcomes?”

Luca can imagine a future of fully automated and personalised monitoring – one that dispenses with cumbersome bleeping machines. “But how far do you want to go? Administering food and insulin will be automated within five or 10 years. Who knows how far we’ll have come in that time?” According to Luca, to move forward: “We don’t really need more data. We just need more data scientists.” Immediately the crowd smile approvingly before separating into smaller groups to talk shop over pizza.

Please join our Amsterdam Medical Data Science group if you want to be on board!

About Amsterdam Medical Data Science

The Medical Data plus Pizza meeting aims to bridge the gap between health professionals and data scientists by bringing both together in an informal setting for presentations and pizza. At Amsterdam UMC data scientists, medical professionals and researchers discuss how AI (Artificial Intelligence) and medical data can be used to benefit patients and improve medical practices. Since it’s launch in August 2018 it has steadily grown in popularity with more and more people attending to hear the latest news and innovations in medical data and spark new collaborations.

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.

Copy: Steve Korver, EdenFrost