Figure 3 | Scientific Reports

Figure 3

From: Automatic segmentation of trabecular and cortical compartments in HR-pQCT images using an embedding-predicting U-Net and morphological post-processing

Figure 3The alternative text for this image may have been generated using AI.

Sample visualization of the two images in the test dataset with the most extreme over- and under-estimates of Ct.Po (predicted value less reference value). Both are distal tibia images. The largest over-estimate of Ct.Po was + 1.48% (left column: a,b,e,f) while the largest under-estimate of Ct.Po was -1.48% (right column: c,d,g,h). (a–d) 3D surface renderings are shown of reference (a,c) and predicted (b,d) cortical compartment segmentations. (e–h) The image is shown in all panes in grayscale, overlaid with colored masks corresponding to the reference (e,g—green), the prediction (f,h—blue), and the voxels for which there is disagreement (f,h—red).

Back to article page