Evaluating the performance of convolutional neural networks with direct acyclic graph architectures in automatic segmentation of breast lesion in US images

Table 1 Mean values of the metrics for the dataset A, using the validation set and cropped images resized to 160 × 160 pixels

CNN	Global Accuracy	Accuracy	IoU	Weighted IoU	BF Score	Dice Coefficient
CNN1
Mean	0.904	0.916	0.766	0.843	0.472	0.776
Standard Deviation	0.045	0.047	0.095	0.063	0.085	0.119
CNN2
Mean	0.895	0.917	0.759	0.835	0.479	0.770
Standard Deviation	0.063	0.046	0.117	0.081	0.112	0.141
CNN3
Mean	0.935	0.919	0.819	0.886	0.553	0.829
Standard Deviation	0.030	0.050	0.076	0.050	0.103	0.009

ISSN: 1471-2342