Table 3.

Effect of widening and deepening the CNN‐SP architecture on performance for the hippocampus segmentation experiment. The number of 3 × 3 × 3 convolutional layers and their associated number of learnable filters are specified in parentheses (e.g., [8–32] specifies eight 3 × 3 × 3 convolutional layers each with 32 filters)

	Dice (%)	MHD (mm]
CNN‐SP [8–32]	91.5 [2.0]	0.23 [0.08]
CNN‐SP [8–64]	91.4 [2.3]	0.24 [0.12]
CNN‐SP [10–32]	91.6 [2.0]	0.23 [0.07]
CNN‐SP [12–32]	91.7 [2.1]	0.23 [0.08]
CNN‐SP [14–32]	92.0 [1.8]	0.22 [0.06]
CNN‐SP [16–32]	91.9 [1.9]	0.21 [0.07]

Note: Mean Dice and MHD values over both left and right hippocampi are reported, with SDs in parentheses.

Abbreviations: CNN‐SP, CNN with spatial priors; MHD, modified Hausdorff distance.