Praneeta Konduri, Postdoc at Amsterdam UMC, specialises in image processing in patients who have had an acute ischaemic stroke

Computer models for fast and effective interventions in the event of acute cerebral infarction

Computer models that predict the most promising intervention in the event of a cerebral infarction based on data from synthetic patients: it sounds futuristic, but the work of Praneeta Konduri and fellow researchers at Amsterdam UMC is bringing this animal-free solution closer. At TPI's request, she explains exactly what the innovation entails and how it can benefit both patients and laboratory animals. ‘Using laboratory animals for clinical research will unfortunately never entirely be a thing of the past, but this technology will enable us to use animal testing in a much more targeted and effective way.’

It is with good reason that medical researchers have been focusing on acute cerebral infarction for some time. This life-threatening situation blocks blood flow to part of the brain, depriving it of oxygen and nutrition. The blockage is usually caused by a blood clot in an artery in the brain from the heart or one of the carotid arteries. When this happens, immediate intervention is crucial for the patient’s survival and to limit permanent disability afterwards. The longer the vessel is occluded the more likely the patient is to suffer problems such as permanent paralysis.

Rapid surgical intervention

Praneeta explained that until around ten years ago, cerebral infarctions were only treatable with drugs. ‘This frequently proved ineffective. Fortunately, Dutch researchers were the first in the world to demonstrate the additional benefits of using an intervention to enable the mechanical removal of the blood clot. As effective as this treatment is, there’s still progress to be made, particularly in the rapid identification of a patient’s best treatment option. This is especially important considering that up to 50 percent of stroke patients remain functionally dependent after surgery.’

Every blockage caused by a blood clot is different

According to Praneeta, the difficulty is that each blockage looks different. ‘Sometimes the interventional radiologist needs to ‘suck out’ the blood clot, sometimes mechanical removal using a stent works best and sometimes a combination of the two is required. There are various methods that could be considered for a specific type of blockage. Taking the time to consider each method or trying out several methods is obviously not an option in an urgent, life-threatening situation. To ensure rapid and effective intervention, medics actually want to use a method that enables them to determine in advance and at high speed the most promising treatment for that particular bloodstream blockage.’

Customised solutions for each patient when every second counts

An additional complicating factor in acute cerebral infarctions is that a range of patient characteristics determine the suitability and effect of a certain type of intervention. Praneeta: ‘Basically, neurologists and interventional radiologists prefer to know exactly what type of intervention is likely to work out best for each blockage and for each patient, so they can take rapid and effective action. Because, to reiterate, in acute cerebral infarction every second really counts.’

Computer calculations versus real tests

Amsterdam UMC established a large-scale INSIST consortium five years ago to enable faster and more effective interventions in the event of an acute stroke. The aim was to develop a test platform on which computer models (aka ‘in silico’) predict and visualise a blood clot’s movement through a blood vessel and clarify when which type of intervention would be most effective. Praneeta: ‘This kind of computer modelling can enable significant reductions in the number of resource-intensive and time-consuming in vitro and clinical trials.’

AI database for synthetic patients and treatments

According to Praneeta, finding sufficient data as input for this kind of computational or in silico modelling is a challenge in itself. ‘The models require huge amounts of data, the collection of which is extremely time-consuming and labour-intensive. Moreover, for patient privacy reasons, you often simply can’t access these data. This led to the idea of using artificial intelligence (AI) to create synthetic patient and treatment data. We lacked the expertise for this in our team. Fortunately, Dr Erik Bekkers, a machine learning and AI specialist was prepared to take on this challenge. As a result of his work, we can now take the first steps towards generating the realistic virtual infarct data we need for this kind of computer modelling.’

First practical applications

Praneeta: ‘Although our research had already demonstrated the value of synthetic data and computer modelling, we’re not yet at the point of being able to use these in clinical decision-making. Our aim, of course, is for our models to be available for quick and accurate use in clinical practice. We’re currently working on this with another large European consortium: GEMINI.’

Fewer, more targeted and effective pre-clinical and clinical trials

With respect to the ethical benefits of the innovation, Praneeta referred first to the privacy issues that become a thing of the past as a result of the database of synthetic treatment data. ‘But I’m particularly pleased that our computer models will enable us to do much more digitally, which will probably lead to less, and more targeted and effective animal testing. Realism is important here, though. I don’t think we’ll be able to manage entirely without animal testing, as much as I’d like this out of respect for animals and life in general. Because no matter how precisely you develop a digital solution, you eventually need to test it against practice guidelines, especially with a medical innovation that concerns life and death decisions.’

Gaining the confidence of doctors and surgeons

According to Praneeta, the biggest remaining hurdle to the actual use of computer models in acute cerebral infarction is not so much in the innovation itself. ‘What’s much more important is that we gain the confidence of all involved parties, varying from the device developers to legislative parties and, of course, the doctors who will be working with this. We can only gain that confidence by providing even more evidence that it actually delivers a sufficient and accurate simulation of reality. I may sound impassioned now, but at the end of the day, we definitely don’t want to impose our solution. The technology can only be implemented in practice once we, as researchers, have proven its value.’

Glimpse of the future

When asked what she wants to have achieved by the end of her career, Praneeta had to think for a moment. ‘That's a big question as I’m still young. More generally, I can say that I find it immensely satisfying to develop these kinds of innovations for healthcare together with my team. The collaboration with all kinds of disciplines within consortia and in research teams is also a fantastic aspect of my work, as this enables me to develop across a range of areas. As well as lots of technical knowledge, you also gain insight of other perspectives on innovations. Societal acceptance of new technologies such as AI and machine learning for medical solutions is, for example, a focus point in innovations like ours.’

More information

Interview: Bard Borger
Photo: WUR