TABLE I:

Ablation study on U-Nets of varying depths alongside with the new variants of U-Nets proposed in this work. U-Net L^d refers to a U-Net with a depth of d (Fig. 1(a–d)). U-Net^e, UNet+, and UNet++ are the new variants of U-Net, which are depicted in Fig. 1(e–g). “DS” denotes deeply supervised training followed by average voting. Intersection over union (IoU) is used as the metric for comparison (mean±s.d. %).

Architecture	DS	Params	EM	Cell	Brain Tumor
U-Net L1	$\mathcal{X}$	0.1M	86.83±0.43	88.58±1.68	86.90±2.25
U-Net L2	$\mathcal{X}$	0.5M	87.59±0.34	89.39±1.64	88.71±1.45
U-Net L3	$\mathcal{X}$	1.9M	88.16±0.29	90.14±1.57	89.62±1.41
U-Net (L4)	$\mathcal{X}$	7.8M	88.30±0.24	88.73±1.64	89.21±1.55
U-Nete	✓	8.7M	88.33±0.23	90.72±1.51	90.19±0.83
UNet+	$\mathcal{X}$	8.7M	88.39±0.15	90.71±1.25	90.70±0.91
UNet+	✓	8.7M	88.89±0.12	91.18±1.13	91.15±0.65
UNet++	$\mathcal{X}$	9.0M	88.92±0.14	91.03±1.34	90.86±0.81
UNet++	✓	9.0M	89.33±0.10	91.21±0.98	91.21±0.68