Table 3.

Comparison of lung and lesion segmentation results on the test dataset

Model	Lung
	Dice (%)	Hausdorff (mm)	VOE (%)	RVD (%)	ASD (mm)	RMSD (mm)	Accuracy (%)
FCN	88.12 ±3.18	42.47 ±27.30	− 9.61 ±16.97	− 7.81 ±16.85	18.53 ±6.87	1.35 ±0.61	93.15 ±8.78
SegNet	95.47 ± 2.84	37.59 ±47.08	− 5.16 ±6.76	− 4.83 ±6.29	13.53 ±10.58	0.91 ±0.56	96.43 ±1.35
Li et al. [23]	95.85 ± 2.11	20.30 ±7.45	3.39 ±6.46	− 3.98 ±5.77	8.55 ±6.76	0.63 ±0.47	95.38 ±2.03
Wang et al. [42]	95.72 ± 2.26	20.55 ±8.11	− 5.15 ±3.82	− 4.85 ±3.55	9.61 ±7.98	0.79 ±0.60	96.06 ±0.35
Müller et al. [43]	95.96 ±1.92	10.09 ±2.74	− 5.07 ±6.47	− 4.75 ±6.29	4.75 ±6.28	0.34 ±0.35	95.15 ±2.36
H-DenseUNet	96.38 ±1.64	11.14 ±3.15	2.15 ±1.12	2.19 ±1.16	3.48 ±2.26	0.32 ±0.26	96.63 ±0.03
TSH-UNet	97.00 ±1.26	8.95 ±2.23	− 0.44 ±2.44	− 0.41 ±2.41	2.73 ±2.10	0.27 ±0.28	97.23 ±0.93
Model	Lesion
	Dice (%)	Hausdorff (mm)	VOE (%)	RVD (%)	ASD (mm)	RMSD (mm)	Accuracy (%)
FCN	73.00 ±7.86	23.67 ±8.64	29.71 ±25.11	40.14 ±36.03	11.25 ±0.60	0.71 ±0.40	64.48 ±12.52
SegNet	81.98 ±5.65	17.69 ±5.02	10.62 ±16.34	12.91 ±19.72	8.54 ±5.09	0.70 ±0.49	78.44 ±9.09
Li et al. [23]	84.88 ±2.55	17.90 ±6.69	10.71 ±10.95	12.04 ±12.52	7.92 ±4.26	0.63 ±0.38	80.84 ±5.74
Wang et al. [42]	84.44 ±2.75	18.28 ±5.89	13.19 ±9.57	14.71 ±11.56	7.19 ±4.19	0.58 ±0.39	79.44 ±5.35
Müller et al. [43]	88.65 ±3.25	10.54 ±2.72	− 6.16 ±8.71	− 5.62 ±8.32	3.97 ±1.56	0.31 ±0.18	83.41 ±5.17
H-DenseUNet	83.80 ±2.88	15.70 ±5.11	16.56 ±13.13	19.15 ±15.30	6.66 ±2.85	0.49 ±0.26	77.81 ±7.01
TSH-UNet	89.22 ±1.93	8.47 ±2.55	12.24 ±8.18	7.33 ±3.63	3.64 ±1.52	0.29 ±0.15	84.58 ±3.48