Simulated ultrasound images help develop better cardiovascular diagnostics

Which of these two ultrasound images is real? Daniek van Aarle regularly posed that question to audiences of ultrasound researchers at scientific conferences. On Sept. 29, she defended her Ph.D. research on computer-generated, or virtual, ultrasound images that are almost indistinguishable from real ones. One image came from a patient, while the other was generated entirely by a computer.

“About half the audience got it wrong. It was almost like flipping a coin,” she says. That is exactly what Van Aarle has spent the past four years working on: creating virtual ultrasound images that look as realistic as possible.

At first glance, this may seem like an unusual goal. Why recreate an ultrasound image when real patient scans are available? The answer lies in the development of new technologies. Software designed to automatically identify the wall of a blood vessel, for example, requires large numbers of ultrasound images for training and testing.

Built digitally from scratch

With patient images, experts often need to manually outline the vessel wall before the data can be used. This is time-consuming, and different experts may draw the boundary slightly differently. “If you ask three physicians to outline the same vessel wall, you’ll get three slightly different drawings,” says Van Aarle.

With a virtual ultrasound, things are different. Researchers build the digital blood vessel themselves and therefore know exactly where the vessel wall is located, how large the vessel is and how much it moves. Software can then be tested using images for which the correct answer is already known. “We’re essentially trying to fool the computer a little,” Van Aarle explains. “The simulated ultrasound should resemble a real ultrasound as closely as possible.”

For patients, this research will not immediately change how examinations or treatments are performed. Van Aarle is clear about that. “As a patient, you don’t directly encounter my research. We’re still a few steps earlier in the process.” The goal is to provide researchers with better tools for developing new medical devices, software and artificial intelligence applications.

Tiny ultrasound probe

A major part of the research focused on patients with an abdominal aortic aneurysm, a widening of the largest blood vessel in the abdomen. During nearly 30 procedures at Catharina Hospital, ultrasound images were captured from inside the blood vessel itself. Physicians inserted a miniature ultrasound probe attached to a thin catheter into the artery. Van Aarle then compared these clinical images with virtual ultrasound images she had generated from medical scans.

In this way, her work connects to earlier research by Floor Fasen, who studied how physicians can extract more information from the vessel wall in patients with the same condition. Characteristics such as vessel wall thickness, calcifications and blood clots may help improve future assessments of aneurysm risk. For that research, simulated ultrasound images complemented data obtained from both laboratory studies and patients.

Joining forces

Van Aarle conducted her research within e/MTIC (Eindhoven MedTech Innovation Center), a collaboration between Catharina Hospital, TU/e, Philips, Máxima Medical Center and Kempenhaeghe. Within e/MTIC, these partners combine their expertise to accelerate the translation of medical innovations from the laboratory to clinical practice. This helps ensure that technological developments better address the needs of physicians and patients.

For Van Aarle, this collaboration was particularly valuable. Although her work usually takes place behind a computer screen, this project frequently brought her into the operating room. “I recreated the equipment and even the catheter digitally as accurately as possible. It was fascinating to see in the hospital how the equipment is actually used in practice.”

Not perfect yet

The virtual ultrasound is not perfect yet. Van Aarle therefore worked to incorporate subtle properties of real biological tissue into her models. At the same time, highly detailed simulations require substantial computing power. Further research will determine how much detail is truly necessary.

She sees her work primarily as a foundation on which others can continue to build, something that is already happening. The simulation toolbox developed during her Ph.D. is now being used in a variety of follow-up projects within the research group, with more projects on the way. “We can create datasets that would not otherwise exist. That allows us to develop new ways of generating images and extracting information from ultrasound scans.”

And if even experienced ultrasound researchers can no longer tell which image is real, the virtual images are already coming remarkably close to the real thing.

Who’s behind this story?

Gaby Clark

MA in English, copy editor since 2021 with experience in higher education and health content. Dedicated to trustworthy science news.

Full profile →

Andrew Zinin

Master’s in physics with research experience. Long-time science news enthusiast. Plays key role in Science X’s editorial success.

Full profile →