Object-Specific Four-Path Network for Stroke Risk Stratification of Carotid Arteries in Ultrasound Images

Table 4

Sensitivity, specificity, precision, and -score comparisons between the baseline VGG16, FP-VGG16, and OSFP-Net. The best results are highlighted in bold. The listed metrics were obtained on the test dataset.

Fold	Method	Metrics (%)
		SEN	SPE	PRE	-score
1	VGG16	79.5	83.0	66.0	72.1
	FP-VGG16	87.5	100.0	100.0	93.3
	OSFP-Net	95.8	95.2	92.0	93.9
2	VGG16	74.0	85.5	75.5	74.7
	FP-VGG16	80.0	95.7	88.9	84.2
	OSFP-Net	95.0	95.7	90.5	92.7
3	VGG16	81.6	90.5	83.3	82.5
	FP-VGG16	80.0	100.0	100.0	88.9
	OSFP-Net	90.0	100.0	100.0	94.7
4	VGG16	84.7	89.2	86.2	85.5
	FP-VGG16	95.7	90.7	84.6	89.8
	OSFP-Net	100.0	100.0	100.0	100.0
5	VGG16	88.9	91.8	80.0	84.2
	FP-VGG16	93.1	100.0	100.0	96.4
	OSFP-Net	100.0	97.3	96.7	98.3
Avg. ±Std.	VGG16
	FP-VGG16
	OSFP-Net	96.2 ± 3.7	97.6 ± 2.0	95.8 ± 4.0	95.9 ±  2.8