To treat a brain tumor, obtaining a piece of tissue through brain surgery is required for analysis. Prof. dr. Marion Smits is working with advanced MRI and image analysis methods to find ways to eventually make that invasive surgery unnecessary. "At the moment, our MRI analysis using algorithms is already more accurate than human estimation."
Marion Smits is professor of neuroradiology at Erasmus MC and one of the Scientific Leaders of the Medical Delta program Cancer Diagnostics 3.0. This was followed in November by an appointment as Medical Delta Professor with a dual appointment at Delft university of Technology. "This professorship is a tremendous honor and I see it as a crowning achievement," Smits says. "The appointment at TU Delft feels like a way to be able to take the next step. As a clinical person in a very fundamental environment, I notice that our worlds are quite far apart. If you then have a formal appointment, the door opens more easily. I now plan to be in Delft more, join more consultations and do more together. For example, I can then be much more directly involved with PhD candidates and students. That way I am really part of the organization and therefore collaboration is easier."
"I work half clinically and half scientifically. As a radiologist, I assess scans of brain and the head/neck area. In doing so, I focus on neuro-oncology, or brain tumors. This is also the focus of my scientific research on physiological imaging of the brain. I am researching whether new techniques, especially MRI techniques, will allow me to image the processes in the brain much better. I do that at Erasmus MC and within the Medical Delta program.
In the Medical Delta program, our dot on the horizon is to extract all the diagnostic information you need for treatment of a brain tumor from an MRI scan, so that no tissue biopsy is needed anymore. We call it tissue-level diagnosis via MRI. Think of it as a virtual biopsy or MRI biopsy. For this we use new MRI techniques combined with AI techniques and lots of data. For this we work together with MRI physicists in Leiden, image analysis scientists in Delft and Rotterdam, and clinicians at Erasmus MC. That's a very strong combination."
"People come to a doctor with complaints. Imaging is then done and then the assessment follows as to what kind of tumor it is and what needs to be done. Those are human judgments. We have studied how correct these estimations are. A radiologist with a lot of expertise scores about 60-70% on accuracy based on an MRI. Our MRI analysis using algorithms is currently 80% accurate, so that's already more accurate than human estimation.
It will be some time before we can replace brain tissue sampling with MRI alone in all patients, but in certain groups of patients it may be applicable in the near future. For example, if someone is too sick for brain surgery. For these people who would otherwise not get a diagnosis at all, 80% accuracy is already a good advance.
One obstacle we still have to overcome is that people have to get used to it.We have retrospectively tested our technique in large data sets, and we are now testing it in practice, without it affecting diagnosis at this point. We should not immediately make everything depend on an algorithm, but we can get a better estimate of what type of tumor it is with this. There is still much to be gained there and this is also a way to get all hospitals to the same level of expertise.
One obstacle we still have to overcome is that people have to get used to it. It's exciting for doctors to rely on something that's happening outside their field of vision. It's something new that we have to learn to work with."
"In the radiology department, the demand and therefore the workload is increasing rapidly, about ten percent per year. This is due in part to new therapies that require imaging. There is a lot of scanning and, rightly so, we rely on it more and more. Scans are also becoming more extensive and take up more work. If you can automate some of this, it saves time and you can probably get more info out of a scan. A simple example is measuring a tumor. This is still done manually now, but automated it is much faster."
"What I find very important is that scientific research and innovations also land in practice. That developments concerning Artificial Intelligence and algorithms do not remain stuck in the ivory tower of our academic hospital. Colleges are closer to practice and it is therefore very refreshing to work with people who are involved in this. With oncological surgeon Joost van der Sijp, for example. He is affiliated with The Hague University of Applied Sciences and together we have just applied for a grant for research into how we can ensure that new technologies also reach clinical practice.
For me, this runs like a thread through my work. I became a doctor for a reason: I want to mean something to someone and I want the best for my patients. I think medicine is important and I think new technology is important and interesting. Those two worlds are pretty far apart and I can be the translator or the bridge."
"Not every encounter with another discipline leads to a collaboration, but almost always to inspiration. Leading up to the start of our Medical Delta program a few years ago, I met with several people from Delft. Some of the things they are working on are far away from my experience, but that is also because I had never thought about them at all. I am very pragmatic: this is the MRI device, this is what we have to do with it. Dreaming about new things is very inspiring. Can we develop new probes with which you can make certain things visible that you cannot yet see? Conversations with others then help to widen your own mind a little. To realize that there is more than just the hospital."
"Sincere interest in each other's work. Not even trying to understand everything. Not immediately looking for where the collaboration is, but asking out what someone does and drives it. Then you automatically find out whether it generates something in yourself and whether collaboration is likely to succeed."
"I work within our Medical Delta program very nicely with the other Scientific Leaders. Thijs van Osch from LUMC has really become a buddy. He works from the MRI physics side and I work from the clinical side. He has a lot of affinity with the clinic and therefore we complement each other well. Sjoerd Stallinga and now Jeroen Kalkman of TU Delft are more concerned with the tissue side. I had never given much thought to the fact that all kinds of advanced imaging techniques could be applied there as well. That in turn has given me lots of new ideas."
This article is part of a series in which we highlight the eight new Medical Delta professors. Click here for the other portraits published so far.