How is the uniformity acquisition traditionally evaluated? Using both a subjective visual assessment and a quantitative pixel value-based analysis metric called “integral uniformity,” which describes the variance in pixel values across the image.
“Although visual assessment is currently the gold standard, to overcome its subjectivity, people tend to rely heavily upon the results of the integral uniformity metric and allow its results to sway their decisions about the clinical quality of the image,” says Jeff Nelson, lead nuclear medicine physicist for Duke University Health System. “Unfortunately, the integral uniformity metric isn’t great for identifying all types of artifacts. For example, subtle or structured patterns—such as those from photomultiplier tubes—within uniformity images can indicate a clinically unacceptable system and may not be reflected in the integral uniformity score.”
So Nelson and colleagues launched a Medical Physics 3.0-based quality improvement project. “My goal was to design a new uniformity metric that better reflects the visual appearance of the image and can identify a wider variety of clinically unacceptable structures and artifacts within the uniformity image,” he explains.
The resulting new metric, Structured Noise Index (SNI), is based on analyzing image noise texture. “It uses frequency-based analysis methods derived from a two-dimensional noise power spectrum to identify and quantify any non-quantum content within the image,” notes Nelson. “And we’ve demonstrated that SNI aligns much closer with observer visual assessments than the traditional integral uniformity pixel-value-based metric.”
After rolling this metric into an automated uniformity quality assessment workflow system, Nelson and colleagues discovered that it is not only better at identifying artifacts requiring service but also the automated workflow improves communication among clinical and service personnel who receive automated email alerts when potential artifacts are identified.
