LCC-Net: A Lightweight Cross-Consistency Network for Semisupervised Cardiac MR Image Segmentation

Figure 2

Illustration of the input space perturbation used in our study. (a) The cardiac MR images. From left to right: original image , original image , the perturbed image , and the perturbed image . (b) Their corresponding label images. We evenly split an image into four blocks and then randomly exchange one or two blocks on the corresponding positions. The same operations are applied on their ground truth label images, as in (b). Best view in color.

(a)

(b)