Table 13.

Experimental results after adding different levels of Gaussian noise to the predicted depth scores generated by the depth regressor in the first stage. Multi-organ lesion detection (MOLD) task results on the extracted multi-organ DeepLesion dataset have been shown using the proposed DA-SHT Dense 3DCE R-FCN, 9 slices model. AS (%) at various FPs per image and mAP (%) were used as the evaluation metrics. $std$ denotes the standard deviation value.

Gaussian Noise	AS@0.5 ± std	AS@1 ± std	AS@2 ± std	AS@4 ± std	AS@8 ± std	AS@16 ± std	mAP ± std
mean = 0, standard deviation = 0.01	46.24 ± 0.67	56.84 ± 0.52	67.31 ± 0.47	74.54 ± 0.16	78.78 ± 0.43	81.56 ± 0.29	40.69 ± 0.50
mean = 0, standard deviation = 0.02	44.56 ± 0.57	56.03 ± 0.27	67.37 ± 0.26	73.77 ± 0.18	78.89 ± 0.23	81.09 ± 0.09	40.39 ± 0.50
mean = 0, standard deviation = 0.1	43.81 ± 0.12	56.30 ± 0.27	66.51 ± 0.56	73.86 ± 0.75	77.77 ± 0.22	81.49 ± 0.36	39.90 ± 0.62
mean = 0, standard deviation = 0.2	43.80 ± 1.68	56.43 ± 1.00	67.24 ± 0.28	73.23 ± 0.44	78.40 ± 0.50	81.20 ± 0.59	39.53 ± 1.09