Skip to main content
. Author manuscript; available in PMC: 2020 Jul 10.
Published in final edited form as: IEEE Access. 2020 May 29;8:101550–101568. doi: 10.1109/access.2020.2998537

TABLE III.

Quantitative comparison of FC-Dense ContextNet (independent label probability estimation) with attention loss and overlap loss in dentate and interposed nuclei segmentation

Target Dentate Interposed
Metric CMD (mm) MSD (mm2) DC Volume (mm3) CMD (mm) MSD (mm2) DC Volume (mm3)
FC-Dense ContextNet (independent label estimation) 0.482±0.29 0.375±0.10 0.875±0.04 739±167 (668±159) 1.115±0.89 0.489±0.31 0.673±0.14 59±21 (53±24)
+ attention loss 0.549±0.42 0.392±0.08 0.867±0.04 770±165 (668±159) 1.040±0.75 0.472±0.24 0.680±0.12 59±23 (53±24)
+ overlap loss 0.555±0.37 0.377±0.10 0.874±0.04 735±172 (668±159) 1.047±0.82 0.467±0.21 0.687±0.13 58±21 (53±24)
+ attention loss and overlap loss (DCN-Net) 0.514±0.35 0.380±0.11 0.873±0.05 736±165 (668±159) 1.085±0.92 0.514±0.35 0.682±0.16 53±18 (53±24)
*

Bold indicates p<0.001 for paired t-tests with FC-Dense ContextNet. ( ) is ground truth volume.