DL assesses aortic valve calcium from routine, non-gated CT exams

Tuesday, December 1 | 1:30 p.m.- 2:30 p.m. | T6-SSCA05-7 | Room E353C

Automated deep-learning (DL) allows for opportunistic assessment of aortic valve calcium (AVC) on standard chest CT, according to this scientific presentation.

Research findings to be shared Tuesday afternoon suggest that using DL for this indication "could identify high-risk patients for life-saving intervention using existing imaging data, requiring no extra scan time and no additional radiation exposure," according to presenter and medical student Nithin Karthikeyan of the University of Sheffield in the U.K. and colleagues.

The researchers evaluated 693 consecutive patients from the ASPIRE registry [Assessing the Spectrum of Pulmonary Hypertension Identified at a REferral Centre] who underwent routine thoracic CT; the cohort included a mix of contrast-enhanced (n=438) and non-contrast (n=255) scans. They used DL to perform automated segmentation; the algorithm processed scans with a mean throughput of 19.9 seconds.

Overall, the group reported that AI-derived AVC volume associated with Vmax (p < 0.001) and yielded an area under the curve (AUC) of 0.98. It yielded a sensitivity of 100%, a specificity of 89.9%, and a negative predictive value of 100%.

"By providing a low-cost screening modality during routine clinical care, this system could capture moderate-to-severe aortic stenosis in populations that currently bypass diagnostic evaluation, without increasing the downstream burden of false-positive referrals," Karthikeyan's team concluded.

 

Page 1 of 2
Next Page