Skip to main content
. 2020 Dec 17;10(12):1691. doi: 10.3390/biom10121691

Table A1.

The performance of the segmentation of the heart, lung, and thoracic wall using U-net and DeepLabv3+.

CNN Structure mIoU mDice mPrecision mRecall
Mean SD Mean SD Mean SD Mean SD
U-net Heart 0.686 0.042 0.798 0.036 0.849 0.028 0.784 0.069
Lung 0.712 0.015 0.828 0.011 0.832 0.020 0.832 0.014
Thoracic wall 0.448 0.017 0.610 0.016 0.679 0.033 0.568 0.032
DeepLabv3+ Heart 0.734 0.012 0.842 0.009 0.864 0.008 0.833 0.008
Lung 0.684 0.010 0.808 0.008 0.833 0.008 0.793 0.014
Thoracic wall 0.417 0.007 0.582 0.007 0.675 0.009 0.525 0.013

SD = standard deviation.