Algorithms that work – how higher math drives innovation in healthcare — CWI Amsterdam

Mathematical algorithmic modeling may seem like a rarefied academic activity, but CWI scientists who pursue research in the life sciences and health (LSH) are committed to designing models with direct applications to challenges in healthcare. From optimizing treatment plans for cancer patients to reducing wait times for ambulances to improving geriatric care and preventing suicide, their work, says lead researcher Peter Bosman, “is not window dressing”. Our algorithms are actually used. We go from foundation to application.

The Centrum Wiskunde & Informatica (CWI, the Dutch National Research Institute for Mathematics and Computer Science) wants the real value of its work to be better understood. That’s why, says CWI Professor Leen Stougie, the partnership with Smart Health Amsterdam is so important to the institute. “We do a lot of applied scientific research in the broad field of LSH, so it was only natural to support the Smart Health Amsterdam initiative. One of the things we are interested in is making new contacts for and with whom we can continue to develop and apply our methodologies. With CWI’s extensive expertise in mathematical algorithmic modeling etc., we can really enrich the LSH domain within Amsterdam Smart Health.

Leen Stougie

Researchers focus on methodologies, based on modeling and formulating problems in a quantitative way. Stougie explains that the developed algorithms have a wide range of applications in the LSH field, including medical imaging, genome research related to COVID-19, radiation therapy and geriatric care.

Practical applications

The CWI models are very advanced and, according to Stougie, those who work in the LSH industry are not always sure how they can be applied in practice. “There are people who know a little math and know what we’re offering, but usually people have no idea,” he says. “I was at one of the very first Smart Health Amsterdam meetings on a boat near the maritime museum and I talked to people there, and they thought our work was very interesting but they couldn’t nothing to do.” Practical, real-world examples, he says, are key to promoting a better understanding of how mathematical modeling can advance healthcare.

Advanced optimization

Peter Bosman leads a research group at the CWI which works, among other things, on medical informatics. A field that involves, he says, “using advanced forms of optimization and machine learning to help medical experts make better decisions.” Medical experts will always have the final say, but by ensuring there are better ways to process the data available, they can gain better insight before making a final decision. CWI’s work with medical imaging, for example, aims to help doctors analyze images more effectively. “How do you contour the prostate, for example,” says Bosman, “so you can actually see where it is and how big it is, and then plan treatment based on that.” His group is also working with UMC Amsterdam to help doctors create more effective plans for administering radiation therapy to treat cancer – another example of using optimization to support medical decisions.

Pierre Bosmann

When working in this field, Bosman says, it’s important that the data you’re working with represents “the best thing possible.” “If you don’t know what’s best yet, you need to optimize first, then create data, then learn from data to accelerate it. In the field of radiotherapy, many recent innovations still focus on optimization. You need to help doctors optimize planning to maximize the likelihood of tumor destruction and minimize the likelihood of damaging nearby organs. Doctors are already making plans with existing software, but we can make better plans with the optimization we have created. You have to optimize to know what is best, otherwise you will just be repeating what you have done in the past. »

Reduce waiting lists in health care

The Dolce Vita Project, a partnership between CWI, UMC Amsterdam, Vrije Universiteit (VU), Amsterdam Institute of Health and Technology (ahti) and SIGRA, a regional partnership of health organizations and of well-being. Dolce Vita is short for “Data-driven Optimization for a Vital Elderly Care System” and CWI’s Rob van der Mei is the project manager. The goal, he says, is to reduce wait times in the acute care system for the elderly. “That’s one of the huge problems,” he says. “It often happens that a patient enters the acute care system and has to be placed in a nursing home, but there is no space available. There are waiting lists in many nursing homes and many problems related to bottlenecks in the system. The idea of ​​the project is to identify bottlenecks and solve them, if possible. The acute demand for health care for the elderly will continue to grow, and that is why it is really urgent that we address this issue.

The project team developed an allocation algorithm that explicitly takes patient preferences into account when deciding how best to allocate care beds. “We have a list of patients who need to be placed and their individual preferences, we have a list of beds that have become available, and we assign them in a mathematically optimal way, often resulting in dramatic reductions in wait times.” Not only are shorter wait times better for patients, they also reduce the overall burden on the healthcare system. Launched in 2019, the project is already delivering promising results: “The algorithms we have developed have reduced waiting times by two to four in realistic use cases.

Create, refine, reuse

Van der Mei is now seeing interest in the model from other healthcare sectors that have similar bottlenecks and wait times issues. “The fun thing about the mathematical modeling we’re developing is that these models to some extent abstract from the specifics of elderly care, youth care or mental health care. So that means we can pretty much reuse patterns from one domain for another. This is the power of mathematics. He adds that he appreciates “that we can do things with our background in AI and math that are relevant to society.”

A cross-sectoral approach and collaboration between the different parties involved in the Dolce Vita are crucial for the success of the project, according to Van der Mei. “We bring the knowledge needed to work with data, modeling and optimization, and healthcare professionals, who work with patients, bring the practical knowledge. And in addition, SIGRA brings an extensive network of healthcare providers and insurance companies. This aspect is also very important as a link between us, academics, and the politicians and decision makers of the system. I am convinced that you need all three parties to get things done.

Rob van der Mei

The power of math

Stougie, Van der Mei and Bosman are all evangelists when it comes to spreading awareness of the power of math and computer science to create solutions to real-world problems. “We’re the ones who can help you do what generic software can’t,” says Bosman. “We can tell you how to use cutting-edge AI techniques, especially machine learning and optimization, and do something a little beyond what you can do today. If your problem is large and complex, which is where we can help. Van der Mei is also keen to highlight the potential of their work to have broad impact. “AI and mathematical modeling are generic and can be reused in a wide This is the power of the type of research we conduct. It can be used anywhere.

That versatility is something people don’t always understand, Stougie says, which can cause them to undervalue CWI’s work. “To clarify our work, we can describe a project like Dolce Vita, but the downside is that people may think, ‘Oh, that’s just one specific example.’ Van der Mei agrees. “Yes, in the case of Dolce Vita, we created a use case for Amsterdam, and some people then think that’s an Amsterdam-only model. When you tell them you can use it for Rotterdam too because it’s just a different set of parameters, then that’s a miracle, isn’t it? Whereas for us this is a completely trivial change to make to a model.

Complete solutions

The three researchers want to make it clear that CWI is not an ivory tower. “Yes, we do cutting-edge science,” Bosman says. “But we know how to put our science into practice. It’s not just window dressing. I often say “from foundation to application” – and by application I mean something that is used. For example, our radiotherapy algorithms are used at UMC Amsterdam to treat patients. We don’t just say “oh look, it works on paper”. No, we also build the software and create a full-fledged solution that works in practice.

CWI is one of The partners of Smart Health Amsterdam – To see the partner page for CWI. The center works with other partners of Smart Health Amsterdam: UMC Amsterdam, ahti and the Vrije University (VU). CWI research generated a number of high-tech spin-off companies working in fields such as data management, operational analysis, optical tracking systems, financial mathematics and biotechnology.

This article was published by the Smart Health Amsterdam platform (SHA). SHA is the network for data-driven and AI-driven innovation in Amsterdam’s life sciences and healthcare sector.

Sharon D. Cole