Analyzing human decisions in IGRT of head-and-neck cancer patients to teach image registration algorithms what experts know

Table 1 IGRT correction vectors accepted by experts and proposed by competing automated registration methods

Correction type T_τ	Correction vectors Mean ± SD [mm]			Pearson’s correlation coefficient PCC(T_τ, T_RT)			Absolute differences Δ_τ-RT = \|T_τ ^Fi-T_RT ^Fi\| Mean ± SD [mm]
	x	y	z	x	y	z	x	y	z
T_RT	0.1 ± 3.0	0.2 ± 2.9	−0.3 ± 2.6	1	1	1	0	0	0
T_BONY	0.3 ± 2.2	−0.5 ± 1.2	−0.1 ± 2.6	0.59	0.38	0.75	1.7 ± 1.7	2.1 ± 1.9	1.4 ± 1.2
T_tPTV	0.3 ± 2.2	0.5 ± 2.6	0.1 ± 2.7	0.62	0.63	0.49	1.4 ± 1.8	1.5 ± 1.9	2.0 ± 1.8
T_pPTV	0.3 ± 2.7	0.6 ± 2.2	−0.2 ± 2.5	0.89	0.79	0.70	0.9 ± 1.0	1.2 ± 1.4	1.5 ± 1.3
T_KNOWLEDGE	0.1 ± 2.5	−0.3 ± 2.5	−0.1 ± 2.6	0.80	0.73	0.75	1.2 ± 1.2	1.4 ± 1.6	1.4 ± 1.2

Fi fraction i, x right-left direction, y anterior-posterior direction, z cranio-caudal direction, SD standard deviation

ISSN: 1748-717X