Secondary Logo

Journal Logo

Institutional members access full text with Ovid®

Semiautomated Characterization of Carotid Artery Plaque Features From Computed Tomography Angiography to Predict Atherosclerotic Cardiovascular Disease Risk Score

Zhu, Guangming, MD, PhD*; Li, Ying, MD, PhD*†; Ding, Victoria, MS; Jiang, Bin, MD, PhD*; Ball, Robyn L., PhD; Rodriguez, Fatima, MD, MPH§; Fleischmann, Dominik, MD; Desai, Manisha, PhD; Saloner, David, PhD; Gupta, Ajay, MD, MS#; Saba, Luca, MD**; Hom, Jason, MD††; Wintermark, Max, MD, MAS*

Journal of Computer Assisted Tomography: May/June 2019 - Volume 43 - Issue 3 - p 452–459
doi: 10.1097/RCT.0000000000000862

Purpose To investigate whether selected carotid computed tomography angiography (CTA) quantitative features can predict 10-year atherosclerotic cardiovascular disease (ASCVD) risk scores.

Methods One hundred seventeen patients with calculated ASCVD risk scores were considered. A semiautomated imaging analysis software was used to segment and quantify plaque features. Eighty patients were randomly selected to build models using 14 imaging variables and the calculated ASCVD risk score as the end point (continuous and binarized). The remaining 37 patients were used as the test set to generate predicted ASCVD scores. The predicted and observed ASCVD risk scores were compared to assess properties of the predictive model.

Results Nine of 14 CTA imaging variables were included in a model that considered the plaque features in a continuous fashion (model 1) and 6 in a model that considered the plaque features dichotomized (model 2). The predicted ASCVD risk scores were 18.87% ± 13.26% and 18.39% ± 11.6%, respectively. There were strong correlations between the observed ASCVD and the predicted ASCVDs, with r = 0.736 for model 1 and r = 0.657 for model 2. The mean biases between observed ASCVD and predicted ASCVDs were −1.954% ± 10.88% and −1.466% ± 12.04%, respectively.

Conclusions Selected quantitative imaging carotid features extracted from the semiautomated carotid artery analysis can predict the ASCVD risk scores.

From the *Neuroradiology Section, Department of Radiology, Stanford University School of Medicine, Palo Alto, CA;

Department of Neurology, PLA Army General Hospital, Beijing, China;

Department of Medicine, Quantitative Sciences Unit, Stanford University;

§Division of Cardiovascular Medicine, Stanford University; and

Cardiovascular Imaging Section, Department of Radiology, Stanford University School of Medicine, Palo Alto; and

Department of Radiology, University of California San Francisco, San Francisco, CA;

#Department of Radiology, Weill Cornell Medicine, New York, NY;

**Dipartimento di Radiologia, Azienda Ospedaliero Universitaria di Cagliari, Italy; and

††Department of Medicine, Stanford University School of Medicine, Palo Alto, CA.

Received for publication October 29, 2018; accepted February 3, 2019.

Correspondence to: Max Wintermark, MD, MAS, Neuroradiology Section, Department of Radiology, Stanford University School of Medicine, 300 Pasteur Dr, Grant S047, Stanford, CA 94305 (e-mail:

The authors declare no conflict of interest.

J.H. and M.W. are co–senior authors.

Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.