Reconstructing the beating heart in 3D: AI for more precise coronary interventions
Source: https://www.tue.nl/en/news-and-events/news-overview/18-03-2026-reconstructing-the-beating-heart-in-3d-ai-for-more-precise-coronary-interventions Parent: https://www.tue.nl/en/research
Share
Using AI and X-ray images to build 3D models of the heart for better treatment decisions
Reconstructing the beating heart in 3D: AI for more precise coronary interventions
March 18, 2026
PhD researcher Kirsten Maas developed AI methods to reconstruct coronary arteries in 3D using just a small number of X-ray images.
image: iStockphoto.com
Coronary artery disease is the most common form of heart disease worldwide. In this condition, the coronary arteries, which are the blood vessels that supply the heart with oxygen-rich blood, become narrower. The standard treatment is a percutaneous coronary intervention (PCI) in which a small metal tube called a stent is placed inside the artery to restore blood flow.
To guide this procedure, cardiologists use X-ray coronary angiography, short X-ray videos in which a contrast fluid is used to makes the arteries identifiable. However, these images are two-dimensional projections of a complex three-dimensional structure, which makes it difficult to accurately assess the severity of a narrowing or determine the best location for a stent in the artery.\ \ PhD researcher Kirsten Maas set out to address these limitations by developing AI-based methods to reconstruct a moving three-dimensional model of the coronary arteries using a small set of X-ray recordings. She evaluated whether advanced neural models can reliably reconstruct coronary anatomy under realistic imaging conditions, therefore improving reconstruction methods to account for heart and breathing motion, and increasing the reliability and transparency of AI-based reconstructions. She defended her PhD thesis at the Department of Mathematics and Computer Science on Tuesday, March 17.
From 2D projections to a moving 3D model
Although coronary angiograms are essential in clinical practice, they have a major limitation – images are two-dimensional projections of a complex three-dimensional structure. As a result, depth information is missing, making it difficult to accurately assess the severity of a narrowing in an artery or determine the best location for a stent.
While images are captured from multiple angles, the number of alternative views is limited in practice to reduce radiation exposure for the patient. As a result, cardiologists must interpret a moving 3D anatomy using a small number of 2D recordings.
Maas first examined whether neural radiance field models, originally developed for 3D reconstruction in computer graphics, can reliably reconstruct coronary arteries from a small number of X-ray images. Using synthetic data, she showed that these models provide a more accurate three-dimensional understanding of coronary arteries, even when imaging data is limited.
Accounting for motion and clinical constraints
Building on these findings, Maas developed new reconstruction methods that explicitly accounted for cardiac motion while using fewer X-ray images. This facilitated dynamic 4D reconstruction possible under realistic imaging conditions and significantly reduced reconstruction time. Further testing with real patient data and improving processing speed are the next steps toward ultimately bringing the method into clinical practice.
Given that real-world procedures also involve breathing motion on the part of the patient, Maas introduced a method to estimate respiratory motion directly from X-ray sequences. This was an important step toward applying AI-based 4D reconstruction to clinical data.
Beyond reconstruction accuracy, Maas also addressed reliability and transparency. As AI models are commonly viewed as so-called ‘black boxes’, she developed a visualization method that highlights uncertainty in reconstructed images, supporting safer and more informed use in clinical settings.\
Toward safer heart interventions
Maas’ research was carried out within the Eindhoven MedTech Innovation Center (e/MTIC), a collaboration between TU/e, clinical partners such as Catharina Hospital, and industry partners including Philips. By combining mathematical modelling, artificial intelligence, and close clinical collaboration, the project connected technological innovation directly to clinical needs.
While further research is needed before clinical implementation, these findings highlight the strong potential of the method to improve future diagnosis.
\ PhD researcher Kirsten Maas. Photo: Angeline Swinkels
- [### Promovendus
Kirsten Maas, department of Mathematics and Computer Science
Read more](https://www.linkedin.com/in/kirsten-maas/?originalSubdomain=nl) - [### Thesis title
Implicit Neural Representations for 3D Reconstruction of X-ray Coronary Angiography
Read more](https://pure.tue.nl/ws/files/379801857/20260317_Maas_hf.pdf) - ### Supervisors
Anna Vilanova Bartroli, Danny Ruijters, Nicola Pezzotti
Written by
Bouri, Danai
(Communications Advisor M&CS)
More on Health
[March 13, 2026
Making homes more sustainable leads to better health for children
Research by TU/e and VU shows that better insulation and ventilation lead to important health benefits for children.
Read more](https://www.tue.nl/en/news-and-events/news-overview/13-03-2026-making-homes-more-sustainable-leads-to-better-health-for-children) [March 4, 2026
Giving Healthcare Professionals Creative Control in Digital Health
AI supported tools help healthcare professionals create personalized, gamified content that supports patient engagement and improves digital health experiences.
Read more](https://www.tue.nl/en/news-and-events/news-overview/04-03-2026-giving-healthcare-professionals-creative-control-in-digital-health) [February 27, 2026
When Online Experiences Shape Care
PhD researcher Jhon Cerón Guzmán examined how patients can better find relevant experiential knowledge in online health communities, supporting users and...
Read more](https://www.tue.nl/en/news-and-events/news-overview/27-02-2026-when-online-experiences-shape-care)
Latest news
[March 18, 2026
TU/e awards an honorary doctorate to Christian Borgs
Borgs is a world-leading researcher at the interface of mathematics, statistical physics, theoretical computer science, and AI for science.
Read more](https://www.tue.nl/en/news-and-events/news-overview/18-03-2026-tue-awards-an-honorary-doctorate-to-christian-borgs) [March 13, 2026
Municipality of Roermond and TU/e investigate collaboration
First potential project: housing for students and starters next to Roermond station
Read more](https://www.tue.nl/en/news-and-events/news-overview/13-03-2026-municipality-of-roermond-and-tue-investigate-collaboration) [March 13, 2026
Making homes more sustainable leads to better health for children
Research by TU/e and VU shows that better insulation and ventilation lead to important health benefits for children.
Read more](https://www.tue.nl/en/news-and-events/news-overview/13-03-2026-making-homes-more-sustainable-leads-to-better-health-for-children)
keep following us
[Socials
Experience life at our university on Instagram. With every week a different person taking over.](https://www.instagram.com/tue_campuslife/)[Video
YouTube
On our YouTube channel you find the latest videos and animations about research, education and working at TU/e.](https://www.youtube.com/user/tueindhoven)
- 1
- 2
- 3
Previous
Next
Previous
Next