Fig. 2 | Scientific Reports

Fig. 2

From: A deep learning approach for automatic 3D segmentation of hip cartilage and labrum from direct hip MR arthrography

Fig. 2

Manual segmentation workflow of internal dataset. Manual segmentation workflow for hip cartilage and labrum on the internal data set. (a) Raw image (Inv2) of the 3D T1 magnetization prepared 2 rapid acquisition gradient echoes (MP2RAGE) sequence were used for manual segmentation of hip cartilage (blue) and labrum (red). (b) Corresponding manual and automatic 3D morphologic models. (c) Masks of cartilage and labrum were then applied to the co-registered T1 map for (d) 3D visualization of post-contrast T1 relaxation time (dGEMRIC) using a voxel wise color-graded scale. Dice Similarity Coefficient between manual and automatic segmentation of this particular example was 0.95 for cartilage and 0.87 for labrum. 3D models were visualized using a custom-made software application for automatic segmentation and visualization8.

Back to article page