Researchers used machine learning to evaluate bone density scans for calcification in the aorta, the main artery in the body. They show that their method could be used to predict future cardiovascular and other diseases, even before symptoms appear.
Just as calcification, or calcium deposits, can be problematic in the lining of blood vessels in the heart, so can calcification in the aorta, the largest artery in the body. From the heart, it branches up, supplying the brain and arms, and down into the abdomen, where it divides into smaller arteries supplying each leg.
Abdominal aortic calcification (AAC), calcification of the portion of the aorta that runs through the abdomen, can predict the development of cardiovascular diseases such as heart attack and stroke, and determine the risk of death. previous study It has also been found to be a reliable marker of dementia in later life. AAC is visible on bone density scans commonly used to detect lumbar osteoporosis, but it takes time for trained professionals to analyze these images.
AAC is usually quantified by trained imaging specialists using the 24-point scoring system AAC-24. A score of 0 represents no calcification, and a score of 24 represents the most severe degree of AAC. Now, researchers at Edith Cowan University in Australia have turned to machine learning to speed up the calcification assessment and scoring process.
The researchers fed 5,012 spine images taken by four different models of bone density machines into their machine learning model. Although other algorithms have been developed to assess AAC from these types of images, the researchers say this study is the largest and the first to test it in a real-world setting using images taken during routine bone density testing Research.
They then evaluated the model’s performance in accurately classifying images into hypocalcified, moderately calcified, and hypercalcified categories based on AAC-24 scores. To check the accuracy, the machine learning-based AAC scores were compared with those given by human experts. Experts and software arrive at the same decision 80% of the time. Three percent of people with high AAC scores were incorrectly diagnosed by the software as having low scores.
“This is notable because these individuals had the broadest spectrum of disease and the highest risk for fatal and nonfatal cardiovascular events and all-cause mortality,” Lewis said. “While the accuracy of the software could still be improved compared to human readouts, But these results are from version 1.0 of our algorithm, and we’ve improved the results considerably in the latest version.”
The researchers say their machine learning algorithm can analyze bone density scans at a rate of about 60,000 images per day. Considering that it takes experts an average of 5 to 15 minutes to analyze a single image, this is a huge improvement.
“The quick and easy access to these images and automated scoring at the time of bone density testing may lead to new approaches to early cardiovascular disease detection and disease monitoring in routine clinical practice in the future,” said corresponding author Joshua Lewis. Said research.
The researchers say their screening method could be used to detect diseases before symptoms appear.
“Automated assessment of the presence and extent of AAC with an accuracy similar to that of expert imaging opens up the possibility of mass screening for cardiovascular disease and other conditions — even before someone shows any symptoms,” Lewis said. “This will allow at-risk people to make necessary lifestyle changes earlier and put them in better shape and healthier later in life.”
The study was published in the journal Electronic Biomedicine.
source: Edith Cowan University