Table 6.

Overall CAD performance comparison using different training methods for 2.5-mm testing dataset.

Training method	Candidate detector sensitivity (before classification)	Number of features selected for classification	Overall CAD performance AUC (0 to 10 FPs)	95% Confidence AUC (0 to 10 FPs) interval	ANODE score
Aggregate	96.49	13	8.57	$8.45\pm 0.50$	0.705
Homogeneous thickness	96.49	12	8.68	$8.61\pm 0.56$	0.718
Common thickness	96.49	12	8.74	$\mathbf{8.66}\pm \mathbf{0.47}$	0.723

Note: Bold values represent the best performance among the compared methods.