AI abdominal fat measurement predicts heart attack and stroke

Oak Brook, Illinois-Automatic deep learning analysis of abdominal CT images produces more accurate measurements of body composition, with major cardiovascular events such as heart attack and stroke being the overall body weight and body mass index (BMI). Better to predict than. A study presented today at the Annual Meeting of the Radiological Society of North America (RSNA).

“The established cardiovascular risk model relies on factors such as body weight and BMI, which are a rough surrogate for body composition,” said Kirti Magudia, MD, Ph., Abdominal Imaging and Ultrasound Fellow at the University of California, San Francisco. .D. Says. “It is well established that people with the same BMI can have significantly different muscle to fat ratios. These differences are important for a variety of health consequences.”

Unlike height- and weight-based BMI, a single axial CT slice of the abdomen visualizes the volume of the subcutaneous fat area, visceral fat area, and skeletal muscle area. However, manually measuring these individual areas can be time consuming and costly.

As a radiologist resident at Brigham and Women’s Hospital in Boston, Dr. Magdia is part of a team of interdisciplinary researchers, including radiologists, data scientists, and biostatisticians, who study deep learning, a type of artificial intelligence. We have developed a fully automated method using (AI)-determine body composition metrics from abdominal CT images.

“Regular abdominal CT scans provide a more detailed way to examine body composition, but we are not currently using it,” said Dr. Magudia.

This study cohort was derived from 33,182 abdominal CT outpatient examinations performed in 2012 at Partners Healthcare in Boston on 23,136 patients. The researchers identified 12,128 patients who had not been diagnosed with major cardiovascular and cancer at the time of imaging. The average age of the patients was 52 years and 57% of the patients were female.

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

This retrospective study identified one of these 12,128 patients who had a myocardial infarction (heart attack) or stroke within 5 years of an index abdominal CT scan. The researchers found that the group had 1,560 myocardial infarctions and 938 strokes.

Statistical analysis showed that the visceral fat area was independently associated with future heart attacks and strokes. BMI was not associated with heart attack or stroke.

“The group of patients with the highest percentage of visceral fat area was more likely to have a heart attack, even when adjusted for known cardiovascular risk factors,” said Dr. Magdia. “The group of patients with the least visceral fat area was protected from stroke in the years following the abdominal CT scan.”

“These results show that the accurate measurements of body muscle and fat compartments achieved by CT are superior to traditional biomarkers for predicting the risk of cardiovascular outcomes.” She added.

According to Dr. Magudia, this study shows that fully automated and normalized body composition analysis can be applied to large research projects.

“This study shows the potential of AI systems to add value to clinical care by extracting new information from existing image data,” said Dr. Magudia. “With the introduction of AI systems, radiologists, cardiologists and primary care physicians will be able to provide better care to their patients while minimizing the incremental cost to their healthcare system.”

This treatise has received the RSNA2020 Trainee Research Award.

Co-authors are Christopher P. Bridge and D. Phil. , Camden P. Bay, Ph.D. , Florian J. Fintelmann, MD, Anababic, Ph.D. , Catherine P. Andriol, Ph.D. , Brian. M. Wolpin, MD, and Michael H. Rosenthal, MD, Ph.D.