TABLE 1.

Segmentation performance of deep learning methods on ICH dataset.

Method	Parameters (1 × 106)	ICH lesions (95% CI)	IPH (95% CI)	IVH (95% CI)	EDH (95% CI)	SDH (95% CI)	SAH (95% CI)
U-Net	2.47	0.624 (0.587, 0.661)	0.688 (0.638, 0.738)	0.518 (0.457, 0.580)	0.222 (0.083, 0.361)	0.321 (0.239, 0.404)	0.245 (0.206, 0.284)
DU-Net	4.83	0.611 (0.573, 0.649)	0.674 (0.622, 0.725)	0.496 (0.432, 0.560)	0.119 (0.014, 0.224)	0.256 (0.171, 0.341)	0.226 (0.186, 0.265)
SEU-Net	2.47	0.629 (0.592, 0.666)	0.691 (0.642, 0.741)	0.538 (0.478, 0.598)	0.253 (0.101, 0.405)	0.381 (0.304, 0.459)	0.256 (0.216, 0.295)
DA-Net	26.97	0.669 (0.636, 0.702)	0.739 (0.695, 0.782)	0.605 (0.550, 0.660)	0.244 (0.112, 0.375)	0.434 (0.348, 0.520)	0.269 (0.227, 0.311)
HR-Net	17.12	0.686 (0.656, 0.717)	0.758 (0.715, 0.802)	0.654 (0.603, 0.706)	0.205 (0.046, 0.364)	0.531 (0.458, 0.606)	0.317 (0.306, 0.389)
Ours	43.06	0.716 (0.685, 0.747)	0.784 (0.745, 0.824)	0.680 (0.631, 0.730)	0.359 (0.173, 0.545)	0.534 (0.455, 0.613)	0.337 (0.293,0.382)