Fig. 4

Improvement of calcification segmentation for different eppochs: Demonstratinghow the segmentation performance by the network improved during training, Fig. 4a gives an example of a ground-truth annotation that was used to train the network. A clear improvement in segmentation performance can be seen in Fig. 4b-d, where in the beginning stages of training (Fig. 4b), the network did not include all calcifications. However, a more extensive and smoothly coverage of the calcifications can been seen as the data is further processed, Fig. 4c and d. a ground truth anotation. b DSC=0.706, 24 epochs. c DSC=0.721, 46 epochs. d DSC=0.740, 50 epochs