Skip to main content
. 2022 Feb 14;12:782988. doi: 10.3389/fonc.2022.782988

Table 4.

Segmentation performance evaluation between benign and malignant tumors in DSC(%).

Methods Mask-RCNN U-Net
Benign Malignant Benign Malignant
Baseline 80.33 ± 4.997 68.22 ± 8.047 65.66 ± 2.962 63.60 ± 3.767
RefMeth_1 78.81 ± 5.435 71.51 ± 3.566 66.16 ± 4.266 63.84 ± 6.313
RefMeth_2 80.76 ± 3.115 71.85 ± 5.060 67.56 ± 3.148 62.20 ± 4.988
RefMeth_3 81.49 ± 5.465 74.80 ± 3.377 66.74 ± 6.118 59.08 ± 1.969
RefMeth_4 84.00 ± 4.938 77.48 ± 7.164 68.23 ± 5.617 64.73 ± 4.333
PropMeth_1 88.07 ± 1.909 81.16 ± 1.681 79.18 ± 3.699 74.22 ± 4.886
PropMeth_2 87.64 ± 2.100 81.13 ± 2.234 79.52 ± 4.384 77.13 ± 2.542
PropMeth_3 87.89 ± 1.759 84.58 ± 1.510 79.93 ± 3.434 76.25 ± 4.010

In addition, a Wilcoxon signed-rank test was conducted to test the statistical significance of the DSC results of different methods. A p-value less than 0.05 is considered as statistical significance. The results are shown in Table 5 .