SQAN 2.0 supports medical imaging researchers and technologists, and near-real time quality control

This is the final installment of a three-part series on medical and neuroimaging support by UITS Research Technologies features SQAN 2.0. The first story on XNAT is available here, and the second, on ConnPipe, is here.

At the Indiana University School of Medicine Center for Neuroimaging (CfN), Dr. Karmen Yoder says researchers using positron emission tomography, or PET scans, are always aware of potential mistakes. “We collect many scans of each patient over 1-3 hours, and sometimes things go wrong, like a power outage,” said Yoder, Professor of Radiology and Imaging Sciences.

These mistakes are difficult for researchers to spot using existing data extraction methods, which can affect the quality of a study’s data. In response, the Scalable Compute Archive (IU SCA) team developed SQAN, or Scalable Quality Assurance for Neuroimaging.  “Partnering with SCA, we are using SQAN to better understand the timing of the different acquisitions and we can go back and retroactively apply that data to our data processing stream,” said Yoder. “SQAN was able to get into the files of all these different scans and was able to pull out the start and stop times of all the individual scans and give that back to us. The type of data we work with, you can’t just read an image,” she added.

SQAN provides a critical quality assurance for Yoder’s research because each step of PET neuroimaging lays the foundation for the data’s precision overall. PET can be described as a time-lapse of information about specific processes within the brain. Rather than presenting a static image, PET traces the injection of a very small amount of a radioactive drug (or tracer) on its journey throughout the brain. After patient scans undergo preprocessing, researchers then apply tracer kinetic modeling to observe how the radioactivity moves through tissue. At the CfN, Yoder’s lab currently focuses on the interaction of dopamine receptors. “Data modeling is critical to understanding the quantitative index for the studies we do,” said Yoder. “SQAN was one of the most meaningful quality control procedures we could have,” she continued, “because if we don’t have that temporal information, we’re lost.”

“There is a lot of detail that goes into doing neuroimaging,” said Rob Bryant, a lead technologist with IU’s Department of Radiology and Imaging Sciences. “We focus a great deal on making sure the quality of the images is pristine,” continued Bryant, who has worked with the SCA to generate templates of ideal patient scans. Researchers use these templates as a baseline for human patient scans. “SQAN helps eliminate any unseen errors early in the research imaging process allowing the technologist to focus more on overall image quality and reproducibility,” he said.

Following the initial release of SQAN 2.0 in early 2020, the SCA team has updated the portal code to use current user interface design best practices and frameworks such as Vue.js and Vuetify. “Designing a user interface that is capable of distilling the information stored in millions of metadata fields presented an interesting challenge, and we’re excited that our solution has reduced the workload on medical researchers and technologists,” said Michael D. Young, Lead Developer, SCA.

“The challenge with PET data is that the scanner essentially buries critical data points important to researchers who rely on complex modeling. We are pleased to improve research quality by bringing these details to light in a user-friendly way,” said Arvind Gopu, manager, SCA. “Also, thanks to persistence and hard work by our project team, I am most thrilled to offer near-real time QC for technologists right after they have scanned a subject,” he added. 

A demonstration of SQAN can be found at https://sqan.sca.iu.edu, and the source code can be found at http://github.com/IUSCA/SQAN.