Table 4. The accuracy of SVM with different kernel functions for the threshold and region based technique for the DDSM dataset.

Numbers in red indicate the best values between the several techniques.

SVM Kernel functions	Threshold + region based segmentation technique
	Accuracy	AUC	Sensitivity	Specificity	Precision	F1 score
Linear	80.5%	0.88	0.774	0.842	0.86	0.815
Quadratic	80.1%	0.87	0.772	0.833	0.85	0.809
Cubic	78.3%	0.85	0.764	0.797	0.81	0.786
Fine Gaussian	54%	0.7	0.51	0.833	0.99	0.673
Medium Gaussian	79.1%	0.86	0.756	0.820	0.84	0.796
Coarse Gaussian	77.2%	0.85	0.736	0.813	0.84	0.785