The Mathematician Advancing the Future of Stroke Care
Professor Ozan Öktem develops advanced mathematical and AI‑based methods that can make medical image diagnostics both faster and more reliable. With a career spanning academia, industry, and several leading international universities, he has built an extensive and multifaceted track record in mathematical imaging and AI‑driven medical diagnostics. “I have taken risks and not always chosen the safe path, but it was necessary for me to follow my curiosity,” Öktem says. That drive has led to a new research grant from the Knut and Alice Wallenberg Foundation’s prestigious Proof of Concept program, providing an opportunity to transform his results into technologies that can make a real difference in healthcare.
The research is grounded in the need for more reliable and efficient methods to interpret complex medical imaging data, particularly in situations where time and precision are critical. By combining theoretical models with clinical insight, Öktem and his research group develop methods tailored to real healthcare environments rather than idealized laboratory conditions. The work is built on close collaboration between mathematicians, engineers, and physicians, creating a structured foundation for developing solutions that can be applied in practice. It is within this broad interdisciplinary setting that the current project has grown and gained the conditions needed to take its next steps.
Background
Öktem was four years old when he arrived in Sweden from Turkey. The family first settled in Rinkeby, later moving to Lappkärrsberget, which enabled him to attend Engelbrektsskolan and later pursue the natural science program in upper secondary school. His interest in chemistry and physics was strong throughout his school years, and his initial plan was to study biochemistry.
“My family has a long history of taking risks for something they believe in. My parents came to Sweden as political refugees, and both my grandfather and great‑grandfather were involved in nation‑building during a turbulent period in Turkey. This shaped my view of what matters in life, to stand for something and follow what feels meaningful,” Öktem explains.
He grew up in an environment where education and science were highly valued. His mother has a background in statistics and chemical engineering and later worked as a teacher in Sweden, while his father is a philosopher specializing in philosophy of science and related questions. The combination of scientific and humanistic perspectives at home created an early and broad academic awareness.
Career Path
The journey to a professorship in numerical analysis at KTH’s Department of Mathematics has been far from traditional. Although science was strong in his family and he initially intended to become a biochemist, his trajectory changed when his father suggested he “take one year of mathematics, it’s always useful.” What began as a temporary detour became a path that shaped both his career and the research fields he would later contribute to.
During his studies, he discovered the potential of mathematics and how applied mathematics and mathematical analysis could be used to understand and model complex natural systems. His interest grew quickly, and he continued to study, deepen his knowledge, and ultimately complete a PhD in pure mathematics at Stockholm University in 1999, despite never planning to stay in the field for long.
After defending his thesis, Öktem chose a path few mathematicians take: he moved into industry. First to the financial sector, where he worked with advanced risk modeling, and later to a research‑intensive startup at Karolinska Institutet developing algorithms for visualizing proteins using 3D electron microscopy.
These years in industry broadened and deepened his expertise. They forced him to think in new ways, away from formal theory and toward rapid problem‑solving, concrete results, and constant adaptation. There he learned how mathematics actually functions in complex practical environments, how models must be adjusted, simplified, and optimized to be useful. At the same time, his interest in academic research gradually returned as he worked in a setting where the technology was closely connected to scientific questions.
“I think it was the freedom to decide for myself and explore different areas that ultimately led me back to research,” he adds.
In 2009, he returned to academia and joined KTH’s Center for Applied and Industrial Mathematics. This became the beginning of a period that shaped both his research and his role in KTH’s growing engagement in biomedical research. Although Öktem started at KTH in an administrative position, his responsibilities expanded quickly, and he took on roles such as director of KTH’s strategic research platform in Life Science Technologies, deputy director of the Stockholm Mathematics Centre, and KTH representative in the establishment of EIT Health. These years placed him at the intersection of mathematics, technology, and clinical research — a place where his unusual background made a significant difference. He built networks, created structures, and opened doors between disciplines that had rarely worked closely together.
Between 2018 and 2021, Öktem spent time in research environments in Göttingen, at the University of Cambridge, the Alan Turing Institute, and the University of Edinburgh. There, he encountered the forefront of mathematics and AI research. International collaborations and new perspectives broadened his outlook and provided tools to further develop his work.
This international experience strengthened his academic profile and confirmed his position as a researcher at the very forefront of his field. After a period at Uppsala University, he returned to KTH in 2022 and was promoted to professor of numerical analysis the following year.
Research Area
His research combines mathematics, statistics, and machine learning to interpret and improve medical images. His work in inverse problems and mathematical imaging has made him an internationally recognized figure in the field. The development of the software library ODL* (Operator Discretization Library) has enabled researchers worldwide to work with advanced reconstruction algorithms developed in his group.
This research is especially relevant in biomedical applications such as low‑dose CT reconstruction* and 3D cryo‑electron microscopy*, where the need for clear and reliable images is essential.
A Fika That Changed the Direction of Stroke Care
During a coffee break with neurosurgeon Marcus Ohlsson at New Karolinska Hospital, Ohlsson described a practical challenge in stroke care: perfusion imaging, a diagnostic procedure in which the brain is repeatedly imaged with CT over a short time interval.
“This is a time‑consuming procedure that cannot always be performed for all patients,” Öktem explains.
To better understand the clinical need, Öktem asked Ohlsson to describe his dream scenario, how diagnostics would ideally work if workflow constraints were removed. The answer was clear: the most valuable improvement would be to eliminate the need for perfusion imaging entirely and instead obtain equivalent diagnostic information directly from the first non‑contrast CT scan taken when the patient arrives with stroke symptoms.
To test the concept, he assigned it to thesis student Joel Wållberg, supervised by doctoral student Jevgenija Rudzusika. The idea was to use AI to read a single non‑contrast CT image and predict the diagnostic information usually derived from perfusion imaging. The results were unexpectedly strong, and the work grew into the startup Aipek, founded by Öktem together with Wållberg, Rudzusika, and neurosurgeon Ohlsson. Today, the group collaborates with hospitals in Sweden, the UK, and South Korea to validate the technique in real clinical environments. The goal is to drastically shorten decision time in stroke care and ensure that more patients receive the right treatment in time.
In December 2025, the project AI‑guided Stroke Identification (AI Stroke ID) was awarded a grant through the Knut and Alice Wallenberg Foundation’s Proof of Concept program, led by Öktem and his team. The grant provides both funding and support via the Wallenberg Launch Pad, making it possible to move from promising research results toward actual clinical implementation.
Research Description: AI Stroke ID
The purpose of AI Stroke ID, developed through the Aipek platform is to make stroke diagnostics faster and easier. The project aims to replace the lengthy blood‑flow assessment (perfusion imaging) with an AI‑based analysis of the first non‑contrast CT scan taken when a patient seeks care for stroke symptoms.
“At present, clinicians often require a full image sequence to determine which areas of the brain are irreversibly damaged and which can still be saved,” Öktem continues.
Because the procedure cannot always be performed immediately, some patients must return for additional examinations, which in practice can lead to multiple visits before a complete diagnostic picture is available. This delays clinical decisions and affects the ability to provide timely treatment.
AI Stroke ID instead uses mathematical methods and machine learning to reconstruct the equivalent diagnostic information directly from a single standard *CT image. The model is trained on extensive clinical data to identify patterns that correspond to blood flow and tissue damage , information traditionally visible only in a perfusion series.
The goal is to give stroke specialists rapid and reliable decision support, regardless of patient conditions and without the need for repeated examinations.
"With this method, diagnostic time can be reduced from 20–30 minutes, or several visits, to just a few seconds. In stroke care, where every minute affects the amount of salvageable brain tissue, this improvement can be crucial for patient outcomes, Öktem points out.
The research is conducted in close collaboration with clinicians to ensure that the technology works both in theory and in real healthcare settings.
AI Is Changing Mathematics — and the Future of Engineering
Rapid developments in artificial intelligence are transforming both research practices and the roles of engineers. Techniques that once required extensive computation, specialized algorithms, or expert knowledge can now be partially automated using large language models and advanced machine‑learning systems. As a result, both researchers and students must adapt to a new reality in which AI is not an add‑on but an integrated part of mathematical work.
“Mathematics and AI are coming together in a way we have never seen before. AI isn’t just a new tool that complements mathematics, it fundamentally changes how mathematical problem‑solving is done, ” Öktem explains.
He emphasizes that this shift places significant demands on engineering education, both at KTH and internationally: “Future engineers will need to understand how AI tools function, when they can be trusted, and how they should be critically evaluated. Traditional methods will remain important, but they must be complemented with knowledge of how AI models are trained, what their limitations are, and how to independently assess their reliability in technical and medical applications.”
Looking ahead, he believes the challenge is to combine classical technical skills with the ability to critically examine and use AI tools, forming a new kind of engineering role.