Fig. 14
From: Generalizable deep learning framework for 3D medical image segmentation using limited training data
Per-patient performance on orbital segmentation. Left panel shows results for the class orbita, and right panel shows the results for the class eyes. Note that low scores are not consistent between patients, such as for patient #3 there is a low score for the class orbita, whereas for patient #6 there is a low score for the class eyes