Table 6.

The statistical analysis of KL-grade-based KOA classification and MOAKS score prediction using trabecular parameters from different regression methods.

Methods	Metrics	Radiomic-based (CT)	CNN (CT)	CT-SMA (CT)	CNN (MR)
KL-grade-based KOA classification	Precision	0.689 ± 0.078	0.745 ± 0.041	0.863 ± 0.053	0.887 ± 0.061
	Recall	0.773 ± 0.045	0.794 ± 0.047	0.903 ± 0.068	0.935 ± 0.037
	F1 score	0.738 ± 0.057	0.763 ± 0.040	0.879 ± 0.056	0.910 ± 0.044
	AUC	0.742 ± 0.032	0.778 ± 0.044	0.883 ± 0.029	0.914 ± 0.048
MOAKS score prediction	Mean Average Error	1.438 ± 0.153	1.059 ± 0.072	0.893 ± 0.041	0.876 ± 0.038
	R-Square	0.723 ± 0.129	0.790 ± 0.064	0.881 ± 0.037	0.903 ± 0.030

The classification is implemented using regressed trabecular parameters on image patches as features. Here we compare different methods to obtain trabecular parameters and the employed aggregation for different methods is Transformer layers. We compare radiomic-based method, CNN-based method and our proposed CT-SMA on CT images. Results of CNN-based method on MR images are provided in the right column for references.

Bold values indicate the best performance among the CT-based methods.