Background: Motion artifacts afect the images of coronary calcifed plaques. This study utilized convolutional neural<br/>networks (CNNs) to classify the motion-contaminated images of moving coronary calcifed plaques and to determine<br/>the infuential factors for the classifcation performance.<br/>Methods: Two artifcial coronary arteries containing four artifcial plaques of diferent densities were placed on<br/>a robotic arm in an anthropomorphic thorax phantom. Each artery moved linearly at velocities ranging from 0 to<br/>60 mm/s. CT examinations were performed with four state-of-the-art CT systems. All images were reconstructed with<br/>fltered back projection and at least three levels of iterative reconstruction. Each examination was performed at 100%,<br/>80% and 40% radiation dose. Three deep CNN architectures were used for training the classifcation models. A fvefold cross-validation procedure was applied to validate the models.<br/>Results: The accuracy of the CNN classifcation was 90.2±3.1%, 90.6±3.5%, and 90.1±3.2% for the artifcial plaques<br/>using Inception v3, ResNet101 and DenseNet201 CNN architectures, respectively. In the multivariate analysis, higher<br/>density and increasing velocity were signifcantly associated with higher classifcation accuracy (all P&lt;0.001). The classifcation accuracy in all three CNN architectures was not afected by CT system, radiation dose or image reconstruction method (all P&gt;0.05).<br/>Conclusions: The CNN achieved a high accuracy of 90% when classifying the motion-contaminated images into<br/>the actual category, regardless of diferent vendors, velocities, radiation doses, and reconstruction algorithms, which<br/>indicates the potential value of using a CNN to correct calcium scores.

doi.org/10.1186/s12880-021-00680-7, hdl.handle.net/1765/137104
BMC Medical Imaging
Erasmus MC: University Medical Center Rotterdam

Magdalena Dobrolińska, N.R. (Niels) van der Werf, Marcel J.W. Greuter, B. (Beibei) Jiang, R. H.J.A. Slart, & Xueqian Xie. (2021). Classification of moving coronary calcified plaques based on motion artifacts using convolutional neural networks: a robotic simulating study on influential factors. BMC Medical Imaging, 21(1). doi:10.1186/s12880-021-00680-7