Skip to main content
. Author manuscript; available in PMC: 2023 Apr 1.
Published in final edited form as: Comput Methods Programs Biomed. 2022 Aug 4;225:107058. doi: 10.1016/j.cmpb.2022.107058

Fig. 5.

Fig. 5.

Qualitative results of the weakly-supervised bolus segmentation (in yellow) obtained with the 5 CNN backbones on four sample frames. Ground truth boluses are highlighted in blue. Among the 5 CNN models, the one using the CNN3 backbone (fourth row) yielded the best results in terms of overlap between the predicted and the ground truth bolus locations.