Two fully automated data-driven 3D whole-breast segmentation strategies in MRI for MR-based breast density using image registration and U-Net with a focus on reproducibility

Visual Computing for Industry, Biomedicine, and Art

Table 3 Comparison of test-retest reproducibility of MagDensity derived using different segmentation methods

	Registration	Deep learning	Dynamic programming	Manual intra-rater (rater 1)	Manual intra-rater (rater 2)	Manual inter-rater (rater 1 vs rater 2)
Mean Δ_2-1	0.236	0.292	0.390	0.282	0.369	1.116
Mean \|Δ_2-1\|	0.529	0.683	0.695	0.540	0.702	1.140
Max \|Δ_2-1\|	1.094	1.748	1.938	1.615	2.014	3.763
MSE	0.370	0.741	0.763	0.479	0.855	1.967
ICC (95%CI)	0.993 (0.982, 0.997)	0.983 (0.956, 0.993)	0.982 (0.949, 0.993)	0.988 (0.966, 0.995)	0.982 (0.952, 0.993)	0.955 (0.444, 0.989)