The use of abdominal computed tomography (CT) is increasingly common in Western medicine. While CT scans are useful for specific clinical purposes – for example, determining the cause of a patient’s acute abdominal pain – they also provide a wealth of supplementary data that often goes unused. CT scans can quantify bone mineral density, visceral and subcutaneous fat, skeletal muscle, and liver fat, among other organ and tissue metrics.
In a process called “opportunistic CT screening,” these rich supplementary data are leveraged to evaluate a patient’s overall health and wellness. From the patient perspective, this adds significant prognostic value without requiring additional imaging time or radiation exposure.
Still, the labor-intensive nature of manual (or even semi-automated) body composition measurements has prevented opportunistic screening from being integrated into clinical practice. Artificial intelligence (AI) might offer a solution by allowing the analysis of biomarker information to be automated.
“If we can confirm high clinical value for these CT-based measures of body composition, then having an automated means will provide rapid and objective assessment,” explains Perry Pickhardt, MD. “Manual measures are too time-consuming for routine clinical use. The automation also allows for large-scale research.”
One example of this type of “large-scale research” is the Opportunistic Screening Consortium in Abdominal Radiology (OSCAR) project. Founded by Dr. Pickhardt, John Garrett, PhD, and Ronald Summers, MD, PhD, the group aims to develop and validate a process for the automated analysis of biomarker information that can translate to routine clinical use.
The OSCAR project has used AI tools to process every abdominal CT performed at UW Madison since 2004. That volume of cases would typically take six to eight months to process: with AI, it can be done in a day. In addition to analyzing UW’s abdominal CT cohort, the group is also replicating their workflow at more than a dozen additional clinics and hospitals, including Harvard, Northwestern, Georgetown, Memorial Sloan Kettering Cancer Center, and international sites in Canada, Australia, and Sweden.
Applying these tools at multiple centers will help address variations related to patient demographics and different technical environments, as well as explore the prognostic value of these combined body composition measures for predicting future adverse events. This large-scale study aims to develop a generalizable, vendor-neutral CT solution that can translate to routine clinical use.
“We are hoping to generate predictive risk models based on these opportunistic CT body composition measures, which have heretofore been generally ignored,” says Dr. Pickhardt. “If we can confirm their value, this could ultimately lead to a low-dose CT ‘virtual physical exam’ that is more predictive for cardiometabolic risk than any currently available method.”
By leveraging AI, this team of researchers is developing a prognostic tool that has the potential to improve care for patients at UW and beyond.