Table 7.

Multi-organ lesion detection (MOLD) task results on the extracted multi-organ DeepLesion dataset. AS (%) at various FPs per image and mAP (%) were used as the evaluation metrics. $std$ denotes the standard deviation value.

Evaluation Metric	AS@0.5 ± std	AS@1 ± std	AS@2 ± std	AS@4 ± std	AS@8 ± std	AS@16 ± std	mAP ± std
Faster R-CNN [25]	37.01 ± 0.49	47.38 ± 0.44	56.73 ± 0.89	64.69 ± 0.33	70.97 ± 0.52	75.92 ± 1.43	32.20 ± 0.42
Original R-FCN [21]	31.52 ± 0.24	41.13 ± 0.33	49.69 ± 0.86	56.38 ± 1.16	63.21 ± 1.03	67.98 ± 1.05	25.74 ± 0.48
3DCE, 9 slices [1]	36.77 ± 1.28	47.01 ± 2.39	56.80 ± 2.46	65.08 ± 2.34	70.83 ± 2.00	75.65 ± 1.82	32.43 ± 1.66
Dense DAL 3DCE R-FCN, 9 slices [3]	42.09 ± 0.80	54.80 ± 1.44	65.17 ± 0.72	71.33 ± 1.28	76.79 ± 1.13	80.39 ± 0.81	36.76 ± 1.47
Dense DKS 3DCE R-FCN, 9 slices [37]	34.00 ± 0.58	42.84 ± 1.58	53.50 ± 0.63	62.48 ± 0.09	67.90 ± 0.09	74.36 ± 1.00	29.76 ± 0.28
MP3D, 9 slices [7]	44.32 ± 1.11	53.30 ± 0.89	61.69 ± 0.60	68.16 ± 0.90	74.94 ± 0.08	80.07 ± 0.26	43.58 ± 1.01
DA-SHT Dense 3DCE R-FCN, 9 slices (ours)	44.37 ± 0.12	58.05 ± 1.13	67.29 ± 0.12	73.88 ± 0.17	78.97 ± 0.14	82.11 ± 0.12	40.39 ± 0.41