Stitching overlapping FiberFind sub-volumes

[Komi]

Hi everyone,

I am working on analyzing a large 3D tomographic volume (approx. 10 mm x 40 mm x 2.5 mm) where the average fiber diameter is around 20 µm.

At my original 10 µm voxel size, the structure is 1039 x 250 x 2122 voxels. However, a 20 µm fiber is only 2 voxels wide at this resolution, which is insufficient for FiberFind-AI to accurately detect the fibers. To achieve the necessary minimum of 4 to 5 voxels across a fiber, I must rescale the resolution to 2.5 µm.

Rescaling the entire volume at once would result in approximately 35 billion voxels. This dramatically exceeds both the 1024³ academic license limit and, more importantly, the physical RAM capacity of standard computing workstations.

To solve this hardware limitation, I developed a Python macro that tiles the volume into 66 overlapping sub-volumes, rescales them locally to 2.5 µm, runs the FiberFind inference, and clears the memory for each cube. I now have 66 individual .gdr files and their associated geometric data folders (fibers.g32, centerlines).

My question: I know the "Combine Results" tool only overlays histograms visually. Does GeoDict provide any post-processing option, Python script syntax, or internal library to topologically stitch the geometric data of these overlapping sub-volumes? My goal is to extract the true global fiber length and the global connectivity/network structure.

Thank you in advance for your insights!

Best regards,