2022 · Computer Vision
3D Mouse Lung Segmentation
Experimental light-sheet microscopy volumes segmented and rendered in ITK-SNAP and 3D Slicer.
- Light-sheet volume
- Watershed and K-means
- 3D reconstruction


Research projectGitHub ↗
Problem
Experimental mouse lung volumes contain noise and uneven illumination, making a single intensity threshold unreliable across slices.
My role
Compared segmentation methods and built a workflow for reviewing the resulting 3D surfaces.
Approach
- Compared thresholding, denoising, clustering, and morphological operations on a light-sheet microscopy volume.
- Combined marker-based watershed with K-means to classify regions and assemble a mask across slices.
- Exported the segmented NRRD volume for checking the geometry in ITK-SNAP and 3D Slicer.
Results
- Produced an exportable 3D segmentation mask and inspected the reconstructed geometry from multiple views.
Stack
- Python
- NumPy
- Watershed
- K-means
- ITK-SNAP
- 3D Slicer