New AI Tool Promises to Slash Radiation and Speed Up 3D Medical Imaging

Researchers in Hong Kong have created an artificial intelligence algorithm that could dramatically reduce radiation exposure for patients—by as much as 99%—while streamlining the process of producing 3D anatomical images for surgical use.

The technology, developed at the Hong Kong University of Science and Technology (HKUST), reconstructs 3D images of bones and organs using just 2 to 4 standard X-rays. This is a significant shift from traditional computed tomography (CT) scans, which require hundreds of images and expose patients to much higher radiation levels. The AI approach uses less than 1% of the imaging data needed for a typical CT scan, reducing radiation exposure by 95%–99%.

According to the research team, the imaging process takes under a minute, offering not only health benefits but operational efficiencies. Reduced imaging time can decrease patient wait times, boost clinical throughput, and cut imaging costs, which are substantially higher for CT scans compared to conventional X-rays.

“There is tremendous demand for 3D printing solutions in orthopedics, and rapid, high-definition imaging technology could significantly optimize surgical workflows,” said lead researcher LI Xiaomeng, PhD, assistant professor in HKUST’s Department of Electronic and Computer Engineering. “Our technology is already assisting surgeons in preoperative planning, and we aim to expand into creating implantable bone structures while advancing imaging precision of 3D anatomical structures beyond our current 97% accuracy benchmark compared to conventional CT scans.”

To bring the technology into real-world clinical settings, HKUST has partnered with Koln 3D, Hong Kong’s pioneering company in orthopedic and metal component 3D printing. The two teams are developing a roadmap for clinical integration of the AI algorithm.

Koln 3D’s founder and CEO, Edmond Yau, praised the potential impact of the collaboration. “HKUST’s AI technology solves our biggest challenge: time. Previously, designing a single implant took days of CT scans and manual modeling, not to mention the wait time for CT scan services at public hospitals,” Yau said. “Now, we can generate 3D bone models using just 2 X-ray inputs in under 30 seconds. This is transformative for patients needing urgent interventions. This collaboration enables us to deliver personalized solutions at scale, cutting costs by 50%–70% while improving outcomes."

The next phase of the project includes clinical testing and validation at a public hospital, set to begin within the coming months. If the algorithm proves effective and safe, full implementation into clinical workflows could begin as early as next year.