Medtech breakthroughs with Amsterdam startup Pacmed
From apps that encourage healthier behaviour to robots that carry out surgery, artificial intelligence and machine learning is becoming a bigger part of healthcare as applications and robots become more sophisticated at treating patients. In fact, some can even do it more efficiently, quickly and at lower costs than regular medical professionals. Of course, the thought of a world in which people who need treatment only see screens and machinery when being diagnosed and treated doesn’t seem immediately appealing. However, Amsterdam startup Pacmed is exploring exactly how it can use AI and machine learning to help medical professionals do their job as efficiently as possible, by using applications to guide doctors to provide the best treatment possible for patients.
Developing new AI and machine learning solutions in Amsterdam
It all started in 2014, when Willem Herter, Wouter Kroese and Hidde Hovenkamp met while participating in the Dutch National ThinkTank, which that year focused on how data can be used to make the Netherlands healthier. One of the problems the trio identified is that medical knowledge was not always based on a group of patients that a doctor might see in a hospital or medical practice, but usually done on relatively healthy patients. “It’s different how a patient reacts to treatment in an isolated setting than in practice,” Herter explains, “there is an enormous amount of data that’s been generated between doctors and patients these days, so we thought ‘why not learn from this observational data?’”
Photo: Willem Herter
And so Pacmed was born. Soon after, Herter, Kroese and Hovenkamp – all now directors of the firm - started to develop machine learning and AI models using this data, which includes everything from a patient’s medical history to the blood pressure that’s constantly monitored in an intensive care unit (ICU).
New software solutions for real-life healthcare problems
Wanting to show healthcare professionals exactly what the added value could be, Pacmed started building software to support GPs in choosing the best treatment for patients with urinary tract infections. “A lot of people were talking about using AI and machine learning in healthcare, but we didn’t see many examples of it in practice, or where it already added value to the patient,” explains Herter. After more than 100 GPs used Pacmed’s system during an implementation study, the startup is now developing algorithms and software to help treat diabetes, hypertension and kidney failure. It’s also using its research to make its systems fit as efficiently as possible into both a doctor’s and patient’s treatment process, keeping ethics in mind throughout all its decisions.
Easing the burden on human practitioners
As well as systems that can help recommend treatment for patients, Pacmed is also working on algorithms to support doctors and patients throughout hospital care. One model will help the Emergency General Practitioner's Center (huisartsenpost in Dutch) predict the actual urgency of a patient’s condition – which could help vastly reduce the number of people being sent for emergency care incorrectly, and in turn save costs and time for medical practitioners.
“A problem that the emergency care units have is that they have seen a big increase in the number of patients that are classified as urgent, but a decrease in the number of patients actually in need of urgent care,” Herter says, “so that creates a lot of pressure on emergency care doctors, and it impacts the quality of care.”
Another machine learning model is being developed to help assist doctors to make a discharge decision at an intensive care unit (ICU), reducing the chances of a patient being readmitted or developing more complications further into their recovery. Pacmed is now working hard to get these models into the implementation stage so they can see the results of their systems in practice.
"Working in AI in Amsterdam is very exciting,” Herter explains, “you have consortiums and research groups – like the Big Data Alliance – and there are great professors of AI here, like Mark Hoogendoorn and Max Welling, who started Scyfer. The other founder of Scyfer, Jörgen Sandig, is working with us. In Amsterdam there’s the opportunity to learn from academics, and there’s also a lot of great talent coming from the universities as well from abroad.”
Collaborating with medical centres in Amsterdam and beyond
As AI and machine learning systems become more advanced and effective, Herter sees them playing a much bigger role within healthcare. But a problem remains that the current applications of the technology do not answer the relevant questions doctors often have in practice. This is why the startup is keeping these professionals at the heart of what they do, working closely with hospitals, GPs and educational centres – including the University of Amsterdam’s Academic Medical Center, the Leiden Medical Center and the University Medical Center Utrecht – to develop their products. “We have had great support from the people at Amsterdam Economic Board, who really believe in our technology. They invited us to some inspirational events and believe that cooperation is key to what we do, which fits in exactly with our view.”
Part of Pacmed’s work also involves working with healthcare professionals on how best to record data so that it can be used more efficiently within AI and machine learning systems. “Doctors are becoming more open to the technology nowadays,” Herter explains. “It’s a new way of learning about the effects of treatments in practice, and it could facilitate a feedback loop between doctors and patients. If you do want to make personalised healthcare possible, then it’s important to have more digital information about the effects of treatments. But of course, it’s only possible to change that behaviour when you can show the added value of the data that’s produced right now.”
It’s this collaborative approach which Herter says is key to successfully launching and running any AI startup. “If you’re planning on starting a new AI or machine learning company, you have to be in touch with the field to gain expert knowledge about the questions you are trying to answer. In our case, we need to be in touch and work with doctors for our models to be successful. You need to keep that end user in mind and bridge the gap between your model and them.”
Building a business for the future
After starting with a modest investment from an innovation fund in 2015, Pacmed has been able to grow to its current size using its own revenue. Herter says it’s important for the firm to only work with parties with the same long-term view, so it can develop and implement technology in healthcare responsibly. “We’re in this for the long run, thinking about how we can gradually grow and develop technology to sustain us for the next 10 or 20 years,” Herter explains. “Not having investors that are solely commercially minded, allows us to have a more long-term vision.” Looking to the future, Herter says that the biggest development in AI in healthcare that we’re likely to see in the next few years is its use in interpreting images from diagnostic machines – such as identifying tumours from MRI scans.”
For Pacmed, the future is seeing their applications being used in hospitals and GP practices, and helping patients and doctors alike. “For us, we’re transitioning from a project company to a product company,” Herter adds. “Projects don’t help patients, but products do. I don’t think that there are many companies that have our experience in AI in healthcare, so we want to share that knowledge on a broad scale, through scientific papers, meetups and blogs. By doing that, we become an AI thought-leader and help hospitals in the Netherlands to make their centres AI driven, and so play a big part in how this technology influences healthcare in the future.”
See more testimonials from Amsterdam’s life sciences and health industry.