Segmentation of renal tubules and urine collection system from HiP‑CT data

Develop and validate a robust segmentation methodology that reliably separates true anatomical structures—specifically the renal tubules and urine collection system—from imaging noise in hierarchical phase‑contrast tomography (HiP‑CT) datasets of human kidneys, overcoming the limitations of thresholding and marching‑cubes approaches.

Background

Using HiP‑CT datasets at 25‑micron resolution, the authors attempted threshold‑based segmentation via marching cubes and observed substantial noise and ambiguity in the resulting isosurfaces. In some regions, structures resembling renal tubules and ureters appeared, but differentiating signal from noise proved difficult.

They explicitly note the challenge of separating anatomical signal from imaging noise and outline next steps to establish a more reliable segmentation pipeline, indicating a clear unresolved methodological need.

References

Separating the signal from the noise in these images remains an ongoing task.

— Proceedings Virtual Imaging Trials in Medicine 2024  (2405.05359 - Abadi et al., 2024) in VITM 2024 — The Virtual Kidney (Session: Computational organ modeling)