Case Study
Advancing a Free Preventive Care Tool
About the Research
Quest Accelerates CT Scan Analysis by 6x and Advances a Free Preventive Care Tool
Dr. Amir Borhani is associate professor of radiology, with secondary appointment in the Department of Surgery, at Northwestern's Feinberg School of Medicine. He studies quantitative imaging, specifically looking at predicting the future risk of cardiovascular events and death by using CT-based imaging biomarkers.
Dr. Borhani is working to automatically extract information such as the density of the liver, density of the bone, quality of the muscles, and the size of certain organs from CT scans because these factors may predict the future risk of cardiovascular events and death. The hope is that if a patient is, for example, given a CT scan to determine the possibility of kidney stones, this method will automatically detect risks of future cardiovascular events and preventively flag it to doctors for their patients.
Research Challenge
This research is part of a collaborative effort spearheaded by the National Institutes of Health and the University of Wisconsin–Madison. These institutions developed an algorithm called OSCAR TOOLKIT to extract these imaging biomarkers. Northwestern is one of the 10 institutions in the international cohort validating this toolkit by collectively studying up to two million CT scans and correlating the imaging findings with clinical outcome.
To do this, Dr. Borhani needed to expand from processing several hundred CT scans to more than 150,000 in a HIPAA-compliant environment. After approaching Feinberg IT for assistance, Dr. Borhani quickly learned that setting up a custom server would cost nearly $50,000. Feinberg IT suggested he reach out to Northwestern IT’s Research Computing and Data Services (RCDS) team to see if Quest, the University’s high-performance computing cluster, could be used instead.
Solution
In his initial consultation with RCDS, it was determined that Quest had the necessary resources to process this data, but Quest is not HIPAA-compliant. To address the data-security needs, the team designed a workflow to de-identify the data securely, move the de-identified data from a Northwestern Medicine archive to Feinberg data storage (FSMResFiles), and then to Quest to be analyzed. While this sounds like a relatively straightforward process, because there are such large data sets, patient data, and multiple systems involved, it took more than a year to complete this research.
Impact
Without Quest, Dr. Borhani says he wouldn’t have been able to conduct this research due to budget constraints. Using Quest, RCDS calculated that Dr. Borhani was able to process the 150,000 CT scans six times faster due to the workflow improvements made by RCDS and Quest's processing resources.
RCDS also collaborated with the creators of the OSCAR TOOLKIT on how to make their pipeline easier to use on other shared computing systems like Quest. This collaboration ensured that other researchers with institutional high-performance compute resources can now learn from the work Dr. Borhani and RCDS did with the technical development and implementation of workflows for this research.
Additional Details
- Researchers: Amir Borhani
- RCDS Lead Consultant: Scott Coughlin
- RCDS Services Used: