A research team led by Professor Kim Cheol-hong at Pohang University of Science and Technology POSTECH, together with a joint team led by Professor Park Deok-ho at S-POHANG Hospital, a regional collaboration hub hospital, developed an artificial intelligence (AI)-based technology to improve ultrasound image quality.
Ultrasound is safe and economical because there is no risk of radiation exposure and it can view inside the body in real time. However, there are strengths and weaknesses depending on the scanning method. Focused-beam ultrasound concentrates sound at a single point, producing sharp images but at a slow frame rate, while plane-wave ultrasound spreads sound widely for very fast imaging but has the limitation of reduced image quality.
To solve this, the researchers designed a two-stage AI technique. In the first stage, they used a model that learns the data processing pipeline to generate high-quality images, converting blurry plane-wave ultrasound images into images as clear as focused-beam ultrasound. It works on the same principle as a smartphone app that enhances a blurry photo to high resolution.
In the second stage, they developed a deep learning model that reconstructs high-quality images starting from the raw signal collected by the ultrasound machine. This model simultaneously considers diverse frequency information to clearly render even fine tissues and vascular structures.
Above all, this study is significant because the team validated it using real patient data at S-POHANG Hospital. Using clinical ultrasound equipment, they were able to reliably obtain high-resolution images in areas such as the carotid artery, thyroid, and musculoskeletal system (muscles of the arms and legs and around bones). This is not merely a result that ends in the lab, but an example showing that, based on organic teamwork between a regional hospital and a university, the technology can be applied directly at the patient's side in clinical settings.
Professor Kim Cheol-hong said, "This is a technology that can obtain high-definition images directly at the raw signal stage produced by the ultrasound machine," adding, "By improving the accuracy of ultrasound examinations, it will not only reduce the burden on patients but also serve as a catalyst to greatly expand the use of ultrasound imaging technology, from regional clinical settings to large hospitals going forward." Professor Park Deok-ho emphasized, "Because it was validated based on real patient data through collaboration between a regional hospital and a university, the significance of the research is even greater."
This study was selected as an early-acceptance paper, given only to the top 9%, at MICCAI 2025, the most prestigious conference in the field of medical imaging, and is scheduled to be unveiled on the 24th when the conference opens.