Radiomic Features Are Superior to Conventional Quantitative Computed Tomographic Metrics to Identify Coronary Plaques With Napkin-Ring Sign

Abstract

Background: Napkin-ring sign (NRS) is an independent prognostic imaging marker of major adverse cardiac events. However, identification of NRS is challenging because of its qualitative nature. Radiomics is the process of extracting thousands of quantitative parameters from medical images to create big-data data sets that can identify distinct patterns in radiological images. Therefore, we sought to determine whether radiomic analysis improves the identification of NRS plaques.

Methods and results: From 2674 patients referred to coronary computed tomographic angiography caused by stable chest pain, expert readers identified 30 patients with NRS plaques and matched these with 30 non-NRS plaques with similar degree of calcification, luminal obstruction, localization, and imaging parameters. All plaques were segmented manually, and image data information was analyzed using Radiomics Image Analysis package for the presence of 8 conventional and 4440 radiomic parameters. We used the permutation test of symmetry to assess differences between NRS and non-NRS plaques, whereas we calculated receiver-operating characteristics' area under the curve values to evaluate diagnostic accuracy. Bonferroni-corrected P<0.0012 was considered significant. None of the conventional quantitative parameters but 20.6% (916/4440) of radiomic features were significantly different between NRS and non-NRS plaques. Almost half of these (418/916) reached an area under the curve value >0.80. Short- and long-run low gray-level emphasis and surface ratio of high attenuation voxels to total surface had the highest area under the curve values (0.918; 0.894 and 0.890, respectively).

Conclusions: A large number of radiomic features are different between NRS and non-NRS plaques and exhibit excellent discriminatory value.

Keywords: angiography; atherosclerosis; chest pain; coronary artery disease; multidetector computed tomography.

Figure 1.

Representative images of plaques with or without the napkin-ring sign (NRS). Volume-rendered and cross-sectional images of plaques with NRS in the top (A, C, and E) and their corresponding matched plaques in the bottom (B, D, and E) are shown. Green dashed lines indicate the location of cross-sectional planes. Colors indicate different computed tomographic attenuation values. NCP indicates noncalcified plaque.

Figure 2.

Manhattan plot of all 4440 calculated P values. The Manhattan plot shows all 4440 calculated P values comparing napkin-ring sign (NRS) vs non-NRS plaques and their distribution among the different classes of radiomic parameters. Radiomic features are lined up on the x axis, whereas the -log₂(P) values are plotted on the y axis. The red horizontal line indicates the Bonferroni-corrected P value of 0.0012. Radiomic parameters above the red line were considered statistically significant.

Figure 3.

Heatmap and clustering dendrogram of all 4440 calculated radiomic parameters. Each parameter was compared with all other parameters using linear regression analysis. Features were clustered based on R² values of the corresponding regression models and plotted along both axes. R² values <0.5 are black, whereas greater values are shown in red with increasing intensity. The 1−R² values was used as a distance measure between parameters and used for hierarchical clustering. The resulting clustering dendrogram can be seen on the right of the image. Cluster analysis indicated that the optimal number of clusters is 44 based on our radiomics data set.

Figure 4.

Stratified 5-fold cross-validated receiver-operating characteristic (ROC) curves of the best radiomic and conventional quantitative parameters. Radiomic parameters (blue) have higher discriminatory power to identify plaques with napkin-ring sign compared with conventional quantitative metrics (green). Detailed performance measures can be found in Table 4.