Focus Areas
Three threads run through my clinical and research work right now. Each is about what happens when a neurosurgeon takes technology seriously instead of treating it as an accessory.
Invasive Brain-Computer Interfaces
Functional neurosurgery has always been an interface discipline. We put electrodes into the brain routinely, we think in terms of circuits and target structures, and we run long-term clinical programs to manage implanted hardware for years. That is exactly what invasive BCI needs. It's why I think functional neurosurgeons are better positioned for this work than most of the engineering groups currently driving the headlines.
The field is early. Invasive BCI is only now moving from research demonstrations toward its first real clinical indications. At Maastricht UMC+ we are planning conservatively: a maximum of around five implants a year to start, growing as evidence, hardware, and reimbursement catch up. As far as I know, only UMC Utrecht is working on invasive BCI alongside us in the Netherlands.
Dr. Christian Herff, whom I helped attract to MUMC+, is internationally recognized in invasive BCI, particularly speech decoding. We co-supervise the BCI research pipeline together: his side is neural engineering, mine is the surgical and clinical work. In December 2023 we obtained an NWO Take Off phase 1 grant to build a proper business plan, which forces hard thinking about indication, workflow, and commercial viability that a purely academic line can avoid indefinitely.
I try to push back on BCI hype rather than add to it. That's the spirit behind my 2024 paper in JMIR Neurotechnology, "Invasive Brain-Computer Interfaces: A Critical Assessment of Current Developments and Future Prospects". The field is promising, real, and still early, with unresolved questions around long-term signal stability, regulatory pathways, and what "first-in-man" means once you move past a handful of selected patients. I'd rather say that plainly than sell a timeline I don't believe.
Deep Brain Stimulation, as context
Deep Brain Stimulation (DBS) is the established technique BCI builds on. It's well proven, with roughly 200,000 people implanted worldwide, and at MUMC+ we do 40 to 60 procedures a year. A few years ago I led a remote-programming and remote-monitoring pilot for Parkinson's patients, letting us adjust stimulation settings and follow up without an in-person visit for every change. Patient satisfaction went up, tertiary referrals increased, and the pilot was featured in a national RTL4 prime-time news segment. It's a smaller, proven version of the same idea behind BCI: the neurosurgeon's job doesn't end at the operating table when a device stays implanted for years.
Agentic AI in Medicine
An agent is different from a chatbot or a decision-support tool. It holds a goal, plans the steps to reach it, uses tools and data on its own initiative, and adapts as the situation changes, with a human setting the goal and the guardrails. That distinction matters clinically. The risk profile of an AI that answers a question when asked is nothing like that of an AI autonomously pursuing a multi-step task inside a hospital workflow.
Agentic AI interests me partly because of where it meets BCI. An invasive brain-computer interface is a high-bandwidth but noisy channel between intention and action, and an agent sitting between the decoded signal and the real world can fill gaps, disambiguate intent, and execute multi-step tasks from an underspecified signal rather than making the patient drive every low-level action. The same architecture shows up in ordinary clinical workflows. Triage, documentation, and follow-up scheduling have the shape of a BCI control loop, just with a keyboard instead of an electrode.
What I'm working on now is the boundary: how much autonomy an agent should get inside a clinical workflow before oversight becomes theoretical, and why that line sits differently for a documentation assistant than for something touching a BCI decoding pipeline. No named project yet, just the open question.
NeuroMind Academy
NeuroMind Academy is EACCME-accredited continuing medical education built around AI-supported, personalized learning. To my knowledge it's the world's first AI-supported accredited CME of its kind. Current courses cover Digital Neurosurgery, Digital Neurology, and Digital Psychiatry, walking clinicians through what AI, machine learning, and large language models actually mean for their own specialty.
It grew out of NeuroMind, a clinical decision-support app I began building in 2010, long before "digital health" was a category. The venture runs as NeuroMind Academy BV, where I'm co-founder, shareholder, and Chief Medical Officer, a commercial interest I disclose plainly on the About page.
For the course catalog and details, see neuromind.academy.