HiP-CT Enables Multiscale Segmentation With Deep Learning
Zhou et al., published February 2, 2026, present a deep learning pipeline that leverages Hierarchical Phase-Contrast Tomography (HiP-CT) multiscale scans to segment small functional units across whole organs. Trained on high-resolution VOIs and using pseudo-labels to extend predictions to ca. 25 /voxel whole-kidney scans, nnUNet achieved a test Dice of 0.906 and detected 1,019,890 and 231,179 glomeruli in two donors. The pipeline enables comprehensive 3D morphological and spatial analyses at organ scale.
Key Points
- 1Trained nnUNet achieved 0.906 average Dice on high-resolution HiP-CT glomeruli segmentation
- 2Multiscale pseudo-labeling propagates high-resolution predictions to low-resolution organ scans, enabling whole-organ unit quantification
- 3Permits population-scale morphological and spatial analyses of glomeruli distributions for pathology and anatomy studies
Scoring Rationale
Significant methodological advance with validated whole-organ glomeruli mapping; peer-reviewed publication and open code enhance reproducibility and adoption.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
