AI Used to Predict Risk of Thoracic Aortic Aneurysm

AI Used to Predict Risk of Thoracic Aortic Aneurysm
This image shows the heart in a patient with ascending aortic aneurysm. 3D CT reconstruction is a modern computed technique that can built a volumetric (3D) image of organs and structures of the human body.

A team led by investigators at Massachusetts General Hospital (MGH) recently reported in Nature Genetics that they used deep learning—a form or artificial intelligence—to uncover insights into the genetic basis for variation in the aorta’s size which can identify people at risk individuals for thoracic aortic aneurysms.

Aortic aneurysms are balloon-like enlargements of the aorta which can tear or rupture causing sudden cardiac death. Typically, patients show no signs or symptoms before the aorta, which carries blood from the heart to the rest of the body, fails. The recent findings leveraging the new AI tool may point to new preventive and therapeutic targets.

“We trained a deep learning model to evaluate the dimensions of the ascending and descending thoracic aorta in 4.6 million cardiac magnetic resonance images from the UK Biobank,” write the investigators.

“We then conducted genome-wide association studies in 39,688 individuals, identifying 82 loci associated with ascending and 47 with descending thoracic aortic diameter, of which 14 loci overlapped. Transcriptome-wide analyses, rare-variant burden tests and human aortic single nucleus RNA sequencing prioritized genes including SVIL, which was strongly associated with descending aortic diameter.

“A polygenic score for ascending aortic diameter was associated with thoracic aortic aneurysm in 385,621 UK Biobank participants (hazard ratio = 1.43 per s.d., confidence interval 1.32–1.54, P = 3.3 × 10−20). Our results illustrate the potential for rapidly defining quantitative traits with deep learning, an approach that can be broadly applied to biomedical images.”

“There were no aortic measurements provided by the UK Biobank, and we wanted to read the aortic diameter in all of the images collected,” explains lead author James Pirruccello, MD, a cardiologist at MGH and an instructor in medicine at Harvard Medical School. “That is hard for a human to do because it would take a long time, which motivated our use of deep learning models to do this process at a large scale.”

The researchers trained deep learning models to evaluate the dimensions of the ascending and descending sections of the aorta in 4.6 million cardiac images. They then analyzed the study participants’ genes to identify variations in 82 genetic loci linked to the diameter of the ascending aorta and 47 linked to the diameter of the descending aorta. Some of the loci were near genes with known associations with aortic disease.

“When we added up the genetic variants into what’s called a polygenic score, people with a higher score were more likely to be diagnosed with aortic aneurysm by a doctor,” says Pirruccello. “This suggests that, after further development and testing, such a score might one day be useful to help us identify people at high risk of an aneurysm. The genetic loci that we discovered also offer a useful starting point for trying to identify new drug targets for aortic enlargement.”

Pirruccello adds that the findings also provide supportive evidence that deep learning and other machine learning methods can help accelerate scientific analyses of complex biomedical data such as imaging results.