A routine osteoporosis screening bone density test can also detect increased risk for a heart attack because of the presence of calcium in the aorta. But reading these images requires expertise and can be time-consuming.
Now, research from a multi-institution collaboration, including Harvard Medical School and Hebrew SeniorLife, reports that this calcification test score can be calculated quickly by using machine learning, without the need for a person to grade the scans.
This finding is published in the journal eBiomedicine.
“This development paves the way for use in routine clinical settings with little or no time to generate the useful calcification score that predicts heart attacks,” said Douglas Kiel, HMS professor of medicine and director of the Musculoskeletal Research Center at Hebrew SeniorLife and an author on the paper.
Trusted health and wellness information
From Harvard Health Publishing
The scoring of abdominal aortic calcification (AAC) from bone density machine images is laborious and requires careful training. As a result, AAC scoring is not routinely performed when these images are acquired in clinical practice. This study developed, validated, and tested machine-learning algorithms for AAC assessment, called ML-AAC-24, and evaluated it in a real-world setting using a registry study of 8,565 older men and women. Greater ML-AAC-24 scores were associated with substantially higher cardiovascular disease risk and poorer long-term prognosis.
“During DXA scans obtained for bone-mineral density testing, vascular calcification of the aorta can be seen and quantified. This study developed a machine-learning algorithm to automatically determine the severity of the calcification that corresponds closely with the manual reading that is far more time-consuming to perform,” said co-first author Naeha Sharif of Edith Cowan University in Australia.
Authorship, funding, disclosures
The study was supported by an Australian National Health and Medical Research Council Ideas Grant and the University of Manitoba Rady Innovation Fund. Kiel’s time was supported by a grant from the U.S. National Institute of Arthritis and Musculoskeletal and Skin Diseases (R01 AR 41398).
