Louis-Armstrong

AI Abdominal Fat Data: Promising for CV Risk Assessment

CT imaging plus artificial intelligence (AI) may offer a better way to predict major cardiovascular (CV) events, a researcher said.

In a retrospective study of more than 23,000 patients in 2012, the level of visceral fat imaged in abdominal CT scans — based on a fully automated, deep learning tool to determine body composition metrics from the images — was independently associated with subsequent myocardial infarction (MI) and subsequent stroke, while BMI was not, according to Kirti Magudia, MD, PhD, of the University of California San Francisco. The research was done at Brigham and Women’s Hospital in Boston.

Also, participants who were free of heart disease at baseline CT, but who had the highest quartile of visceral fat area, had a 31% excess risk of having a heart attack within 5 years (hazard ratio 1.31, 95% CI 1.03-1.67, P<0.04), Magudia reported at the Radiological Society of North America (RSNA) virtual meeting.

Additionally, those who had the second highest, third highest, and highest quartiles of visceral fat area had a 37% to 49% excess risk of having a stroke within 5 years (P<0.04) versus the leanest participants in the study.

“BMI was not independently associated with MI or stroke in a model adjusted for all BC [body composition] metrics,” the authors noted.

“Established cardiovascular risk models rely on factors like weight and BMI that are crude surrogates of body composition,” Magudia explained. “It’s well established that people with the same BMI can have markedly different proportions of muscle and fat. These differences are important for a variety of health outcomes.”

“The group of patients with the highest proportion of visceral fat area were more likely to have a heart attack, even when adjusted for known cardiovascular risk factors,” she added. “The group of patients with the lowest amount of visceral fat area were protected against stroke in the years following the abdominal CT exam.”

“These results demonstrate that precise measures of body muscle and fat compartments achieved through CT outperform traditional biomarkers for predicting risk for cardiovascular outcomes,” she said.

A single axial CT slice of the abdomen can visualize the volume of subcutaneous fat area, visceral fat area, and skeletal muscle area, but manually measuring these individual areas is time intensive and costly.

“Abdominal CT scans that are routinely performed provide a more granular way of looking at body composition, but we’re not currently taking advantage of it,” Magudia said.

While the group’s fully automated BC analysis of abdominal CT exams is currently a research tool, “this analysis can be done easily…in fact, a consumer-level graphics processing unit would be sufficient to fully automate body fat composition analysis with the speed and scale necessary for a large hospital, harnessing latent value from routine imaging,” Magudia told MedPage Today.

The researchers, including radiologists, a data scientist, and a biostatistician, analyzed 33,182 abdominal CT outpatient exams performed on 23,136 patients at Partners Healthcare. They identified 12,128 patients (57% women; mean age 52; majority white) who were free of major CV events and cancer diagnoses at the time of imaging. After 5 years of follow-up, there were 1,560 MIs and 938 strokes, they said.

The researchers selected the L3 CT slice (from the third lumbar spine vertebra) and calculated body composition areas for each patient. Patients were then divided into four quartiles based on the normalized values of subcutaneous fat area, visceral fat area, and skeletal muscle area.

“This work shows the promise of artificial intelligence systems to add value to clinical care by extracting new information from existing imaging data,” Magudia said in a statement. “The deployment of artificial intelligence systems would allow radiologists, cardiologists and primary care doctors to provide better care to patients at minimal incremental cost to the health care system.”

Perry Pickhardt, MD, of the University of Wisconsin School of Medicine and Public Health in Madison, told MedPage Today that “Automated measures of body composition derived from CT scans are a relatively straightforward application of artificial intelligence that should be widely available to practicing radiologists in the near future.”

“Specifically, measures of intra-abdominal fat, muscle, bone mineral density, vascular calcification, and liver fat could soon be a routine opportunistic add-on to CT scans,” Pickhardt, who was not involved in the study, explained. “Importantly, these measures can provide valuable prognostic information, regardless of the original clinical indication for scanning.”

The study was awarded an RSNA 2020 Trainee Research Prize.

Last Updated December 03, 2020

Disclosures

Magudia and Pickhardt disclosed no relevant relationships with industry.

Source Article