A professor at UW-Madison is developing new methods for reducing errors in CT scans, MRIs and other imaging techniques.
Guang-Hong Chen is a medical physics specialist at the Wisconsin Institutes for Medical Research with a number of patents and technologies up for licensing. His work centers on ways to enhance internal medicine by improving how doctors see inside patients.
One widely used tool is known as computerized tomography, or CT imaging. X-ray images are taken of the patient from various angles, and a computer stitches the images together to create cross-sections of the patient’s body. CT scans provide more detailed information than normal X-rays, but the reconstructed images can contain errors known as artifacts.
To avoid these errors, Chen and collaborator Jiang Hsieh created a new way to break down scanned images, separating sections that contain artifacts from others that are compromised. By recombining the sections of the image using a set of image references, they were able to reduce artifacts in the scans they tested.
The researchers say their method could lead to lower hardware costs by offsetting the expense of calibrating the system to account for potential sources of artifacts. An info sheet from the Wisconsin Alumni Research Foundation shows researchers tested 20 images with industry-specific artifacts in developing this method for cleaning up CT images.
While some of Chen’s research focuses on CT imaging and other specific scans, he’s also created a way to improve image reconstruction across imaging types including MRI, PET scans and SPECT.
Magnetic resonance imaging — commonly referred to as MRI — uses magnets and radio waves to look at organs and other internal structures. PET scans rely on radiation and can detect diseases before other types of imaging, while SPECT is another nuclear imaging test that uses radioactive substances and special cameras to create complex internal pictures.
In all of these scans, errors can arise when the subject being scanned shifts or breathes, or when the beam interacts abnormally with denser structures such as bone.
In his research, Chen has drawn a connection between artifacts and data inconsistency and outlined a way to leverage that understanding. This method to reduce artifacts is based on a data inconsistency metric, or DIM.
By examining data inconsistency levels at individual data points, Chen was able to select an “optimal data set” with a low inconsistency and reconstruct images with “minimal artifact contamination.”
An info sheet shows Chen’s work could be incorporated into medical image reconstruction software used for cardiac and respiratory procedures, where movement is largely unavoidable. He has constructed patient data sets for patients with metal implants, as well as others with motion artifacts and other issues.
Chen has more than a dozen other technologies available for licensing at WARF’s website.
See the full list: http://www.discoveryportal.org/faculty.aspx?id=2329
