Portrait and video Meike Vernooij: "The computer puts you on the right track. But the doctor remains at the helm"

One in five people develops dementia, making it one of the most common brain disorders in older age. Professor Dr. Meike Vernooij is investigating the origins of dementia and how we can better identify those at increased risk. "Because we don't stop at our boundaries, we harness much more and can do better for patients."

Meike Vernooij, a neuroradiologist and head-neck radiologist at Erasmus MC, employs a wide range of imaging techniques to establish the most accurate diagnosis possible. As a researcher, she works on expanding knowledge about the origins and prevention of neurological diseases such as stroke and dementia.

Vernooij is one of the Scientific Leaders in the Medical Delta Diagnostics 3.0: Dementia and Stroke scientific program and a consortium member of the METABODELTA program: Metabolomics for Clinical Advances in the Medical Delta. Recently, she was appointed as a Medical Delta Professor at TU Delft.

You are appointment as a Medical Delta Professor. What does that mean for you?

"It is, first and foremost, a great honor and recognition for the interdisciplinary and transdisciplinary work I have been doing for a long time. This appointment increases visibility and attention, making it easier for others to find you, including those who were not on your radar. I expect this will open new doors and strengthen the collaboration with Delft University of Technology. It pulls you more actively into the world of the other university."

Could you give an example?

"In anticipation of this appointment, I had a conversation with Paulien Herder, the dean of the Faculty of Applied Sciences at TU Delft. We discussed my role as the chair of the Talent & Innovation Council at Erasmus MC. In this role, we focus on how to best support researchers in their development. From Paulien, I learned that certain issues we are currently contemplating in Rotterdam are already resolved in Delft and could serve as examples — and vice versa. Through this conversation, I am now going to meet with those in Delft who are involved.

I look at my role as a professor more broadly than just in research. As you progress in your career, you become less hands-on and more of a wheelbarrow and mentor for others. I enjoy contributing to talent, and this professorship allows me to explore this within TU Delft."

Can you briefly explain your expertise?

"I am both a neuroradiologist and an epidemiologist. In my daily practice, I combine clinical work with research. My expertise lies in brain aging and neurodegenerative conditions such as dementia. In my research, I study MRI scans of thousands of middle-aged and older individuals, aiming to identify brain changes that could be predictors of the risk of dementia or its precursor.

We analyze the structure and function of the brain. Has the tissue changed? Is the blood flow different? Traditionally, we assessed these aspects by looking at scans. For example, we might observe that the hippocampus, a brain structure involved in Alzheimer's disease, has become smaller. This was a coarse evaluation.

AI helps discover markers that one might not have thought ofHowever, there is much more information in these images than what the naked eye can perceive. Over the last twenty years, I have been extracting more information from scans and quantifying assessments, collaborating closely with technologists from the Biomedical Imaging Group Rotterdam or the Imaging Physics department at TU Delft. With modern technology in image analysis, we can extract biomarkers or features from a scan that are not visible to the naked eye. This includes the strength of connections between brain cells or the shape of a structure rather than just its volume. Artificial Intelligence, such as deep learning, helps discover markers that one might not have thought of, allowing the data to reveal patterns.

The goal is to understand the onset of dementia, enable prediction and early diagnosis, and eventually translate that knowledge into clinical practice. When people come to a neurologist with memory problems, it needs to be determined if they have dementia, what type, and at what stage. To do this effectively, you must first understand what is normal. Then, by comparing an individual to this baseline, you can better assess how successfully or unsuccessfully someone is aging."

How far along are you in this research?

"Within ERGO (Erasmus Rotterdam Health Study, internationally known as 'the Rotterdam Study'), we have collected a wealth of information about brain aging over the past twenty years and extracted quantitative data from it. As a spin-off, a company emerged that processed this information into an algorithm that can be used in clinical practice. It is akin to the growth curves used at the pediatrician's office to determine if a baby or toddler is growing well compared to what is normal. Similarly, we can use this for the brain but in terms of shrinkage. Does someone deviate significantly from age and gender peers on the shrinkage curve? We now use this daily in decision-making. Is what we are seeing normal?

These types of curves represent a simple form of AI. The computer supports us in the diagnosis. But this is just the beginning. We are currently developing more machine learning and deep learning models to better combine image biomarkers, blood biomarkers, and neurological examinations for a more precise diagnosis. As humans, we are limited in this regard."

Do you still understand what happens with AI and machine learning?

"There is, of course, a limit to how one can master certain fields. It is better to seek out experts and ensure you can communicate with them. My strength lies in providing relevant clinical questions and conveying them to experts who can maximize the potential of technology. Integrating the clinical question with technological expertise is where I find my strength."

How much can you trust an algorithm in making a diagnosis?

"If you give a computer a supporting role, you need to know how dependable it is. What are its limits, and what are its weaknesses? You must trust that the result applies to the patient in front of you. If it is a black box where something comes out, and no one understands it, then it is not useful. Ultimately, a human makes the decision, not the computer.

I start with my visual assessment of a scan. Then, I incorporate quantitative information. If they match, it adds extra confidence and speeds up the process. If they do not match, there might be information in that discrepancy. Sometimes, there might be a technical error, or the computer might detect something not visible to the naked eye. Upon reviewing it again, you might have a better idea of what to look for, and you might see it. The computer sets you on the right track. But the doctor remains at the helm."

What role does transdisciplinary collaboration play in your research and work? How do you involve patients?

"In my clinical work, I don't see the patients directly, but as a radiologist, I see scans of their brains. I discussed the results with the neurologist. Therefore, I must thoroughly understand what the neurologist needs to discuss this with the patient.

Within the ERGO study, we also look at the end-users in practice. In conversations with general practitioners, for example, it became apparent that they are not entirely ready for predictive models for dementia. They might say, 'There's nothing you can do about dementia anyway,' as there are no medications available. Also, they believe that the patient cannot do anything with this information. They might tell a patient, 'You have a twenty percent chance of developing dementia in the next five years.' According to the general practitioners, the patient does not know what to do with that information.

You can think up things in your ivory science tower, but, ultimately, the field must be able and willing to do something with itThat is why we are now engaging in conversations with participants in the population study through focus groups. Suppose this is the outcome, can you do something with it? Does it influence your decision to adjust your lifestyle? Activities like more exercise, cognitive training, and reduced alcohol consumption help prevent dementia. Or would that knowledge make you very uncertain and create a lot of pressure? You can think up things in your ivory science tower, but, ultimately, the field must be able and willing to do something with it."

How is it to start collaborating with someone from a completely different discipline?

"I was always very accustomed to relying on my knowledge. The danger of that is that you stay within your comfort zone. Then, you automatically do not know what is happening outside. If I had done that, my research would have taken a different direction. We would have only looked at how we could better utilize existing resources. Now, we are discovering new imaging biomarkers. This provides a wealth of information hidden in those images. Because we do not stop at our boundaries, we harness much more and can do better for patients.

In these collaborations, it is essential to ensure that you speak each other's language. You must take each other along and be open to each other's backgrounds. Both of you need to understand enough to comprehend each other. From two separate paths, you then continue on one road."

What drives you to collaborate and conduct research?

"Curiosity. The desire to know for myself how things work. My field is exciting because the brain is the most complex organ in the body. When I did my doctoral research, we had a large quote on the door that explained how challenging it is: 'If the human brain would be so simple that we could understand it, we would be so simple that we couldn't.'

I find it very satisfying that, as a radiologist, like a detective, you can search the images for what is happening I find it very satisfying that, as a radiologist, like a detective, you can search the images for what is happening. Also, knowing that this is such a terrible condition affecting many people. Dementia is a major public health issue, and contributing to that is very meaningful. As a researcher, you work on a small puzzle piece, but if you do it as well as possible and apply it in practice, that piece becomes relevant."

What future collaboration do you envision? What call do you want to make?

"There is a lot of pressure to be very efficient and effective all the time. However, exploring potential collaborations is not necessarily efficient. I advocate for more space to engage in collaborations and conversations where you do not yet know exactly if and what it will yield. That is never wasted time. Sometimes, something needs a few years to simmer, and we should allow that to happen."

By whose work have you been genuinely surprised, and why?

"I cannot mention just one name without doing many others an injustice. I genuinely enjoy collaborating with people from a technological background. It is surprising to see where you can go by presenting a clinical question or a medical problem and then hearing how someone has been working on it from their perspective for a long time, without having the same clinical question. It is amusing how things brew in those silos. People from the technical side sometimes say, 'I want to develop this because I know it's possible.' That person is not primarily driven by what you can achieve with it.

Everyone has their motivation. When you bring these things together, the clinical question and technical drive, beautiful things emerge."