Semi-automated segmentation of pre-operative low grade gliomas in magnetic resonance imaging

Cancer Imaging

Table 2 Performance of all segmentations compared to STAPLE true segmentation

	\( \widehat{p} \)	\( \widehat{q} \)	PPV	NPV	DI	JI
A	0.979	0.999	0.965	0.999	0.97 ± 0.05	0.94 ± 0.08
L	0.883	0.998	0.951	0.996	0.91 ± 0.06	0.84 ± 0.09
J	0.967	0.996	0.882	0.999	0.92 ± 0.08	0.86 ± 0.11
Op1	0.922	0.996	0.893	0.997	0.90 ± 0.06	0.82 ± 0.09
Seg3D	0.907	0.996	0.879	0.997	0.89 ± 0.06	0.80 ± 0.09
Op2	0.916	0.997	0.900	0.997	0.90 ± 0.05	0.82 ± 0.08
Op1’	0.914	0.996	0.896	0.997	0.90 ± 0.06	0.82 ± 0.09
T2W	0.920	0.996	0.893	0.997	0.90 ± 0.06	0.82 ± 0.09
EM	0.830	0.996	0.906	0.994	0.84 ± 0.17	0.75 ± 0.18

A, L, J: Experts’ manual segmentations. Op1: First operator’s segmentation. Seg3D: Segmentation in 3D. Op2: Second operator’s segmentation. Op1’: Second segmentation of the first operator. T2W: Segmentation using only T2W images. EM: Segmentation using EM

ISSN: 1470-7330