With the opening of the VU Campus Centre for AI & Health, another piece of Amsterdam’s wildly ambitious ‘AI Technology for People’ initiative falls into place. The online launch event highlighted how a diversity of specialists is required to bring a people-centric artificial intelligence to healthcare.
The long road to the greater good
On Friday, 23 October 2020, a Zoom-based launch event celebrated the opening of the VU Campus Centre for AI & Health.
As host and one of the initiators, Mark Hoogendoorn (VU Computer Science) explained that the centre is very much a network organisation out to fuel collaboration for improving healthcare by developing, implementing and evaluating AI technologies.
With partners including UvA, CWI, Smart Health Amsterdam and many others, the centre features 60 staff members from six faculties and hundreds of researchers. A soon-to-appear website will track all participants and their interactions and collaborations.
Collaboration, team spirit, impact
Mirjam van Praag, president of VU’s executive board, explained how the centre is an integral part of ‘AI Technology for the People’, which involves a collective investment of one billion euros over the next decade. “There are huge opportunities in terms of e-mental health, imaging, decision support, personalised treatment and prevention. We can have a huge impact on the society we serve,” says Van Praag.
Yet issues remain. “Privacy, confidentiality and informed consent all need to be addressed,” says Chris Polman, president of Amsterdam UMC’s executive board. “We also need to ensure that the data ownership stays with the citizens. We have to deal with inequalities – for example, to make sure that white male data is not applied to black females.”
“But the opening of this centre is a huge step forward,” says Polman. “It’s just a shame we have to wait until we can raise our glasses of champagne in person.”
Red dot alert
It was now time for the research presentations. Among other titles, Mihaela van der Schaar is a professor of machine learning, artificial intelligence and medicine at the University of Cambridge and a fellow at The Alan Turing Institute in London.
She opened her keynote ‘AutoML and interpretability: powering the machine learning revolution in healthcare’ with a warning: “Some of my slides are really only for the geeks – I’ve marked them with a red dot. I did try to use as few of these slides as possible,” says Van der Schaar before diving deep into a blur of red dots.
Turning black boxes into white boxes
The work of Van der Schaar’s lab seeks to lower the threshold in the successful application of AI by non-experts, for example by automating the use of machine learning (AutoML). Meanwhile, interpretability seeks to deal with ML’s often effective but unexplainable ‘black box’ outputs (which, unfortunately, only get darker when automated).
The goal is a ‘white box’ where the outputs are transparent, trustworthy and meaningful – and thereby making it acceptable to patients, clinicians and regulatory bodies. Happily, Van der Schaar and her team are developing various clinically-proven tools that integrate interpretability into AutoML frameworks.
The power of cross-discipline fertilisation
The presentations by junior researchers showed the importance of cross-fertilisation – may it be between disciplines or across the academic-business world divide.
‘Improvement and evaluation of AI techniques for personalised mental health interventions’ brought together clinical psychologist Marketa Ciharova with computer scientist Ali el Hassouni. They use reinforced learning to personalise internet-based interventions – to minimise the change of users disengaging and maximise the chance of those seeking help to follow quality advice (e.g. “do something fun every day.”).
Managing better healthcare through AI
With ‘The intersection of AI and knowledge work: a case study in radiology’, VU researcher Bomi Kim took a more anthropological approach to uncover best practices when applying AI in an area where the changes will be immense.
“We need a middle management to get their hands dirty and follow the whole complex process – keeping those involved in the loop without overselling the tech or underselling the necessary prep work and costs,” says Bomi.
Outcomes for cancer patients
With ‘Deep learning for tumour response evaluation’, Nina Wesdorp (Amsterdam UMC – Cancer Center Amsterdam) brought us back to basics: that AI can really have a direct impact on patient care and cure. Her collaborative project with analytics software leader SAS is improving outcomes for patients with colorectal cancer – the third most common cancer worldwide – by better identifying those 20% of patients who respond well to chemotherapy and thereby more likely candidates for life-saving surgery.
And now with the opening of the VU Campus Centre for AI & Health, we can welcome the acceleration of more such collaborations.
AI technology for people
AI technology for people is a collaborative partnership between the Amsterdam Economic Board, Amsterdam UMC, Antoni van Leeuwenhoek/Netherlands Cancer Institute, Centrum Wiskunde & Informatica, the City of Amsterdam, Amsterdam University of Applied Sciences, Sanquin, University of Amsterdam and Vrije Universiteit Amsterdam.