Table 6.

Contribution of individual characteristics.

Characteristics	AUC	Accuracy	Sensitivity	Specificity	p
All	0.78	0.79	0.84	0.61	-
No lumen	0.72	0.71	0.76	0.56	0.039
No calcium	0.73	0.54	0.52	0.61	0.152
No attenuation	0.76	0.76	0.81	0.61	0.191
No tree	0.76	0.74	0.79	0.56	0.380

Rows correspond to the proposed approach, networks trained on artery characteristic [with the characteristics from the artery tree (bifurcation and side-branch)], and a network trained on all characteristics apart from the tree characteristics (no tree). Among the separate artery characteristics, the network trained on the lumen area performed best. Including all characteristics in the proposed approach lead to the best performance. Excluding the tree characteristics resulted in a slight decrease in performance. p-values indicate the statistical significance of AUC improvements of the model using all characteristics over the model using the respective characteristic.