Engineering Professor Awarded $2.3 Million to Improve CT Scan Images
Hengyong Yu Is Developing Computer Program for New Generation of CT Scanners
By Edwin L. Aguirre
Prof. Hengyong Yu of the Department of Electrical and Computer Engineering has been awarded a four-year, $2.3 million grant by the National Institutes of Health’s National Institute of Biomedical Imaging and Bioengineering to help improve the image quality and resolution of photon-counting computed tomography (CT) scans.
“This technique fully utilizes the energy spectrum of the X-ray source and the energy-discriminating capability of the photon-counting detector,” says Yu, who is the project’s principal investigator. “This novel technology generates multienergy images at high spatial resolution, so it outperforms conventional CT imaging in characterizing soft tissues and contrast agents as well as pharmaceutical drugs.”
Each year, more than 80 million CT scans are performed in the United States to detect diseases such as cancer and internal injuries such as fractured bones. The scans combine a series of X-ray images and process them on a computer to create detailed, cross-sectional views of the patient’s bones, soft tissues and blood vessels.
According to Yu, photon-counting CT offers not only fast, noise-free imaging but also lower dose of ionizing radiation compared with traditional CT scanners.
“This opens a new door to huge opportunities for imaging at functional, cellular and even molecular level, using novel contrast agents such as gold and bismuth nanoparticles,” he says.
Yu’s co-investigators are Profs. Yan Luo (electrical and computer engineering) and Yu Cao (computer science). Doctoral students Shuo Han and Bahareh Morovati are assisting Yu in lab research, and two additional Ph.D. students and a postdoctoral fellow will be joining the project.
External collaborators include Dr. Anthony Butler of MARS Bioimaging Ltd., a medical imaging company based in Christchurch, New Zealand, and Prof. Ge Wang of Rensselaer Polytechnic Institute.
The Power of CT Image Reconstruction
When an X-ray beam passes through a patient’s body, the resulting two-dimensional view shows all the bones, tissues and organs along the X-ray’s path as overlapping images. A CT scan takes a series of X-rays along different directions, and the CT’s computer can analyze and reconstruct the image to provide radiologists and physicians with a clear, nonoverlapping internal view of the body.
“With a photon-counting detector, when an X-ray passes through a patient’s body, multiple two-dimensional views can be obtained simultaneously for different X-ray energy channels,” notes Yu. “Similarly, images can be reconstructed for different energy channels, resulting in a colorful, 3D internal view of the body.”
Despite the great utility and huge potential of photon-counting CT, Yu says there are well-recognized technical challenges that researchers need to overcome in image reconstruction. For example, to improve the photon-counting CT’s time resolution for dynamic, contrast-enhanced imaging, he says one solution is to lessen the time it takes to acquire data by reducing the number of X-rays taken and/or the exposure time for each X-ray. However, this can result in having too few images or low-dose X-ray images to work with, which can lead to problems in image reconstruction.
Yu’s goal is to develop deep-learning algorithms and artificial neural networks that can automatically extract and analyze information or find hidden patterns from large amounts of data to reconstruct the images, despite having too few or low-dose X-ray views.
For this project, Yu is using a new photon-counting CT system that MARS Bioimaging had developed based on the Medipix detector licensed out of CERN, the European Organization for Nuclear Research. The detector can capture and process information from individual X-ray photons, producing 3D color images at very high resolution.
“We will demonstrate the clinical applications of the software we developed on a MARS photon-counting CT scanner,” Yu says. “Initially, the scanner will be used mainly to image hands and feet because of the limited detector size of the system. However, our software can be applied to other photon-counting CT scanners as well as other parts of the human body.”