Figure 2.

Glomerular segmentation strategy. To perform glomerular segmentation, the trained segmentation model was applied (Figure 1) to a square grid of points spaced 100 pixels (px) apart to classify grid points as inside glomeruli (large red dots) or outside glomeruli (large black dots). Human annotation of glomeruli is represented by red lines. For pixels classified as glomeruli, the segmenter returned to those locations to sample more finely at 20 px to determine glomerular boundaries (small dots). A final mask is generated through nearest-neighbor interpolation. Original magnification, ×40.