Table 1.

Landmark-based comparison between automatic registration predictions and expert neuroanatomists’ manual annotations of these landmarks.

	Intra-animal registrations (n = 20)			Inter-animal registrations (n = 20)
	AR vs Expert1	AR vs Expert2	Expert1 vs Expert2	AR vs Expert1	AR vs Expert2	Expert1 vs Expert2
Landmark 1 (µm)	158 ± 94	155 ± 90	184 ± 94	188 ± 83	266 ± 129	297 ± 114
Landmark 2 (µm)	78 ± 58	140 ± 90	191 ± 76	152 ± 78	207 ± 97	214 ± 87
Landmark 3 (µm)	154 ± 82	105 ± 70	253 ± 101	178 ± 81	246 ± 82	269 ± 113
Landmark 4 (µm)	88 ± 69	118 ± 74	233 ± 57	139 ± 66	163 ± 76	255 ± 75
All (µm)	120 ± 84	130 ± 82	215 ± 87	164 ± 78	220 ± 104	259 ± 102

20 registration operations were performed for both intra-animal and inter-animal acquisitions between paired acquisitions (inter- or intra-animals). For each pair, the automatic registered data was resampled in the reference dataset space and two expert neuroanatomists were asked to annotate four landmarks within the two datasets. Individual landmark 3D distance shifts between registration prediction and expert annotation were averaged over the 20 estimations, as well as the overall shift. Automatic registration (AR) was compared to individual expert annotation, and the two experts’ annotations were compared to each other. Automatic registration predictions were globally shifted by 120 ∓ 84 μm relative to the first expert annotation and by 130 ∓ 82 μm relative to the second, whereas inter-annotator shift was globally estimated to 215 ∓ 87 μm for inter-animal dataset registration. The same shifts are estimated to respectively be 164 ∓ 78 μm, 220 ∓ 104 μm and 259 ∓ 102 μm for inter-animal dataset registration. The variance was computed as standard deviation.