Table III.

Runtime and Memory Evaluation of Different System Configurations and Target Volume Resolutions.

	input: 3 stacks, 255 slices × ~ 150 × 150 × 80								input: 6 stacks, 510 slices × ~ 150 × 150 × 80
	1xCPU		12xCPU		1xGPU		2xGPU		1xCPU		12xCPU		1xGPU		2xGPU
target voxel size [mm]	1.0	0.5	1.0	0.5	1.0	0.5	1.0	0.5	1.0	0.5	1.0	0.5	1.0	0.5	1.0	0.5
a motion measurement [s]	4.36	4.36	4.36	4.36	0.76	0.76	0.76	0.76	8.72	8.72	8.72	8.72	1.46	1.46	1.46	1.46
b PSF volume update [s]	109.42	722.14	22.4	140.65	0.75	5.37	0.63	3.61	216.20	1458.40	42.33	278.64	1.48	8.94	0.78	4.74
c slice-to-volume reg. [s]	227.16	1841.09	32.12	249.99	31.95	246.90	23.68	202.34	468.97	3564.43	65.67	505.07	39.08	114.39	26.83	77.68
d update RS parameter [s]	6.51	35.38	1.86	9.68	0.31	5.94	0.29	1.32	12.99	69.01	3.71	18.733	1.56	3.32	0.61	2.83
e bias estimation [s]	13.13	49.79	2.01	6.99	0.07	0.07	0.08	0.08	25.99	25.23	3.93	3.91	0.43	0.44	0.15	0.15
f super-resolution [s]	8.23	72.62	1.91	14.92	0.68	5.94	0.48	4.64	12.26	90.59	2.66	19.47	1.36	12.38	1.12	10.52
g bias correction [s]	14.03	199.08	2.55	0.69	3.47	33.76	0.65	3.22	16.38	191.29	2.98	35.47	1.38	6.98	1.31	6.51
Total [min]	76.38	609.49	13.57	100.76	5.70	45.63	4.38	34.42	142.22	951.95	24.36	162.83	8.98	33.81	6.10	26.02
Total runtime approximation [s]: total = a + m ⋅ (b + c + n ⋅ (d + e + f + g) + ñ ⋅ (d + e + f + g))
Memory footprint [GB]	> 10	> 12	> 12	> 20	< 0.5	< 0.5	< 0.5	< 0.5	> 16	> 24	> 16	> 24	< 1.0	< 1.0	< 1.0	< 1.0