Fat measurement from AI abdominal CT predicts heart attack and stroke

Body mass index (BMI) is traditionally used to determine a person’s body fat percentage, but analyzing abdominal CT images using automated deep learning tools allows for more accurate body composition measurements. Predict major cardiovascular events. the study.

At a presentation during the December 2 session of the Radiological Society of North America (RSNA), researchers at the University of California, San Francisco (UCSF) evaluated abdominal fat and artificially performed a CT scan to assess abdominal fat. We discussed the advantages of using intelligence. Patients can face a heart attack or stroke within 5 years.

This research received the RSNA2020 Trainee Research Award.

Kirti Magudia, MD, Ph.D., Abdominal Imaging and Ultrasound Fellow, University of California, San Francisco. States that at a basic level, BMI as a measurement is flawed. This is because we only consider weight and height. As a result, the two people have very different muscle-to-fat ratios and can have very different health results, but they still have the same BMI.

However, using a single CT slice of the abdomen, the provider can get a much more accurate image of the patient’s body composition. This image shows the amount of subcutaneous fat, visceral fat area, and skeletal muscle. However, the problem is that making these measurements manually is time consuming and costly.

“Routine abdominal CT scans provide a more detailed way to look at body composition, but we don’t currently use it,” Magudia said.

To make these CT scans easier and more economical to use, Magudia and an interdisciplinary team have fully used deep learning to accurately identify body composition metrics from these abdominal images. Created an automated method.

The team used 33,182 abdominal CT scans performed on 23,136 outpatients at Partners Healthcare in Boston in 2012 to 12,128 people who were not diagnosed with major cardiovascular or cancer at the time of imaging. Identified the patient. Using an L3CT slice from the third lumbar spine, calculate the weight composition of each patient (57% female) and group based on normalized values ​​of subcutaneous fat area, visceral fat area, and visceral fat area. Was divided into four quartiles. Skeletal muscle area.

The team retrospectively investigated scans of these 12,128 patients and found that within five years of abdominal CT scans, there were 1,560 heart attacks and 938 strokes within the group. Their statistical analysis revealed that the visceral fat area was independently associated with both of these events, but BMI was not. Patients with the highest visceral fat area were most likely to have a heart attack and patients with the lowest visceral fat area were protected from stroke, even if the team adjusted for known cardiovascular risk factors, Magdia said. ..

“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 said.

According to Magdia, based on the work of the team, providers will be confident with the large-scale use of fully automated and normalized body composition analysis.

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

Click for more information on RSNA. here..