TABLE 2

MSE Between Noisy and CNN-Filtered Cylindric Phantom Images with Noiseless Image as Reference

	Total counts in millions
Image type	1		0.9		0.8		0.7		0.6		0.5		0.4		0.3		0.2		0.1
Image type	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Original	61.57	1.64	68.01	1.46	76.85	1.92	87.17	1.74	102.04	1.70	121.82	2.52	153.81	3.03	204.72	4.98	305.88	8.41	610.56	14.50
B-CNN	11.10	0.51	11.65	0.49	12.45	0.66	13.17	0.49	14.42	0.61	16.01	0.97	18.11	0.94	20.86	1.38	25.61	1.50	39.68	2.47
M-CNN	6.57	0.34	6.95	0.39	7.54	0.43	8.16	0.48	9.06	0.52	10.39	0.83	12.43	0.93	15.11	0.83	20.71	1.25	36.36	2.17
C-CNN	5.84	0.35	6.22	0.44	6.81	0.46	7.53	0.47	8.36	0.51	9.62	0.78	11.55	0.86	14.25	0.71	20.40	1.12	40.08	2.78
Gaus3	59.42	1.58	65.64	1.41	74.18	1.86	84.14	1.68	98.50	1.64	117.58	2.44	148.46	2.92	197.58	4.79	295.22	8.14	589.30	13.96
Gaus5	27.19	0.84	29.79	0.77	33.34	0.91	37.37	0.76	43.42	0.92	51.09	1.38	63.82	1.50	83.86	2.16	123.75	3.76	245.17	6.14
Gaus7	22.12	0.70	23.12	0.73	24.66	0.72	26.27	0.68	28.92	0.85	31.84	1.15	37.26	1.31	45.31	1.64	61.48	2.10	112.09	3.99
Gaus9	31.19	0.78	31.67	0.81	32.67	0.78	33.63	0.83	35.32	0.93	36.90	1.15	40.30	1.38	45.02	1.66	54.58	1.74	85.06	3.35
Gaus11	48.87	0.91	49.06	0.92	49.81	0.90	50.46	1.03	51.68	1.06	52.57	1.20	54.93	1.51	58.01	1.79	64.29	1.71	84.41	3.07
Gaus13	69.29	1.01	69.33	1.01	69.95	0.98	70.45	1.16	71.41	1.19	71.94	1.25	73.75	1.62	75.98	1.92	80.60	1.78	95.22	3.01
Med9	17.58	0.86	19.08	0.91	20.92	0.86	23.24	0.99	26.20	0.98	29.97	1.40	36.50	1.55	46.13	2.26	65.22	2.52	123.37	5.07
Med25	32.22	1.55	33.15	1.50	34.05	1.48	35.23	1.48	36.43	2.19	38.38	1.66	41.37	2.14	46.94	2.87	56.36	2.98	88.63	4.29
Med49	72.97	2.43	73.62	1.92	73.81	2.45	75.03	1.94	74.95	2.76	76.09	2.85	78.27	3.15	82.50	3.58	89.85	4.23	116.64	5.18
Med81	117.12	2.68	117.55	2.44	117.19	2.55	117.62	2.42	118.11	2.71	118.92	2.68	120.67	3.12	124.07	3.51	131.24	3.51	159.07	5.13

B-CNN = bone CNN; M-CNN = mix CNN; C-CNN = cylinder CNN; Gaus = Gaussian filter; Med = median filter.
Values represent mean MSE of 40 noise realizations.