Deep learning model can rapidly quantify CAC using CT attenuation maps obtained from PET/CT scans
WEDNESDAY, Oct. 5, 2022 (HealthDay News) — Deep-learning (DL) positron emission tomography (PET) computed tomography (CT) attenuation correction (CTAC) scan scores may predict cardiovascular risk similarly to standard coronary artery calcium (CAC) scores, according to a study published online Sept. 14 in JACC: Cardiovascular Imaging.
Konrad Pieszko, M.D., Ph.D., from Cedars-Sinai Medical Center in Los Angeles, and colleagues adapted a novel DL model, developed for video applications, to rapidly quantify CAC from PET CTAC scans. The model was trained using 9,543 CT scans and was tested from an external cohort of 4,331 patients undergoing PET/CT imaging who had major adverse cardiac events (MACEs). MACE risk stratification was analyzed in four CAC score categories.
The researchers found that automatic DL scoring required less than six seconds per scan. Stepwise increases in the risk for MACE were seen in DL CTAC scores across the CAC score categories. The net reclassification improvement was not significant for standard CAC scores over DL CTAC scores. Similar negative predictive values were seen for MACE of zero CAC with standard and DL CTAC CAC scores (85 versus 83 percent).
“The addition of routine CAC scoring with CTAC scans could lead to improved diagnosis, risk stratification, and disease management and could influence lifestyle recommendations,” the authors write.
Cedars-Sinai has a pending patent application on the use of convolutional long short-term memory for multislice medical image segmentation; two authors disclosed financial ties to the medical device industry.
Copyright © 2022 HealthDay. All rights reserved.
