Fig. 3
From: A neural network with a human learning paradigm for breast fibroadenoma segmentation in sonography
Comparison of a the breast ultrasonic image, b the ground truth masks, and the predicted masks using c baseline, d C3ECA and C3CBAM, e C3ECA and probSparse, f BottleneckCSP and LogSparse, g Focus and LogSparse, and h C3ECA and LogSparse for three representative cases of breast fibroadenoma I (top row): a conventional fibroadenoma; II (middle row): a fibroadenoma with an overall longer cross-section; III (bottom row): a fibroadenoma with an overall larger area