Thursday, July 11, 2024
A 'glioma' is a tumour in the glial cells. These are the supporting cells of the brain. Every year, about a thousand people in the Netherlands are diagnosed with this disease. They cannot be cured. The right treatments can prolong life and reduce complaints. A prerequisite for this is that radiologists closely monitor tumour development.
Medical Delta PhD student Dr. Karin van Garderen (Erasmus MC) researched methods to help radiologists make this assessment. She conducted this research within the scientific program Medical Delta Cancer Diagnostics 3.0: Big Data Science of in & ex vivo Imaging and recently successfully defended her dissertation. Her research contributed to a new methodology that is now being used at Erasmus MC.
This interview is the fourteenth in a series of interviews with PhD candidates and postdoc researchers funded by Medical Delta. Karin's research is funded by the Medical Delta Cancer Diagnostics 3.0 program.
What is your area of research, and how did you come to this topic?
“I studied physics and then computer science at TU Delft, with a specialization in machine learning and artificial intelligence. When I finished my studies, I didn't really know what I wanted. Machine learning and artificial intelligence are of course 'hot topics', so you can work anywhere - from large online retailers to industry.
My partner works at Erasmus MC on the development of pacemaker technology and came home with the coolest stories about what they did and the impressive technology they use for it. Then the penny dropped for me and I knew what I wanted to do: using technology to make a difference for people. I came across the vacancy for a PhD researcher and got to work.
Then the penny dropped for me and I knew what I wanted to do: using technology to make a difference for people.A characteristic of glioma is that the tumour can remain stable for some time, but at a certain point there is progression of the tumour. That may be the time to start a new treatment or perform an operation, in consultation with the patient. The period up to this decision is also called 'watchful waiting'.
People undergoing treatment for glioma are closely monitored. Patients get regular MRI scans during that period; the image of the tumour on the MRI often guides decision-making on treatment. The radiologist uses scans to estimate whether there is tumour growth. This involves comparing multiple scans over time. The dimension 'time' is what makes this research so interesting for me as a data scientist.”
What do you hope to achieve with your research?
“I developed methods to help radiologists, amongst other things, by using artificial intelligence to automatically outline the tumour and quantitatively measure tumour growth. The radiologist's analysis remains the gold standard, with the technology supporting the radiologist by, for example, making an outline in a 3D scan much faster - something that takes a radiologist a lot of time.
The technology of this automatic 'volume measurement' is not new, but thanks to our research it has now been so advanced and so well evaluated that it is actually used in our clinic. A very nice result. The next step is that others outside Erasmus MC can also use it, but this requires further certification. Hopefully a product will be released in the future.
Gliomas can be divided into grades. Particularly with low-grade gliomas, you see that they initially grow relatively slow and respond well to therapy. But at some point some kind of evolution takes place that makes them more aggressive. It is very important to be able to see this on the MRI and preferably to predict it. To do this, you have to look for the moment before that growth takes place. This could be a next step in the research.”
Photo: Guido Benschop
Photo: Guido Benschop
What motivated you to do this research?
“That is twofold. I wanted to contribute something to society with my knowledge of machine learning. It is very exciting that we have now managed to get the technology to the point where a protocol has been drawn up that radiologists can actually work with. It really solves problems.
In addition, as a data scientist it is also the type of data you work with that is interesting. There is much to be gained from artificial intelligence for other types of medical data. Machine learning will not always be effective or useful. But I am convinced that there are still many possibilities for this specific field of work.”
What are your experiences with regard to collaboration within the Medical Delta program?
“Our research group includes quite a wide variety of researchers, including clinicians, the people who know a lot about the technology behind MRI scans and data scientists. For my research, I also collaborated with biologists who look at the molecular evolution of glioma and with neurologists. I had to have a very broad view, which was very interesting and from which I learned a lot.
With the help of people who actually treat patients, I was able to help find a solution for a specific problem.I found it very pleasant and useful to work with people who actually treat patients. From them I received the necessary input to help find a solution to a specific problem. I had an idea of the problem, but by talking to them and by experiencing for yourself how difficult it is to make volume measurements and delineate tumours in MRI scans, you gain new insights. You better understand the problem and are able to look for solutions in a more targeted manner.
I have also been to a few Medical Delta meetings and by doing that, you see how varied the applications actually are and how many different things are being done. That was cool to see. I had the feeling that we are all doing important work.”
One of the previous interviewees, Antonio López Marín, has a question for you: does your research take into account the developments of AI and how can this be further developed in the future?
“Part of my research was on AI and especially its application. Developments are moving very quickly, but what I have mainly seen is that it is very important to test AI in the situation in which you want to apply it, because you often experience that it does not fully work as it should.”
Karin's PhD thesis can be downloaded here.
