Skip to content

NVAP 0.1.6 — Recover missed microglia cell assignments

Choose a tag to compare

@gero-studios gero-studios released this 21 Oct 23:41

NVAP 0.1.6 improves actual microglia cell and branch assignments, beyond the display-preservation fix in 0.1.5.

Fixed and improved

  • Unassigned signal regions are checked for credible cell bodies at the active threshold, so a brighter neighbouring cell does not prevent detection of a dimmer soma.
  • Branches can recover a cell owner through fainter measured signal paths. These paths establish ownership without adding below-mask bridge voxels to measured cells. Fragments with competing owners or no supporting signal path remain unassigned.
  • Smaller accepted cells are preserved when larger cells are present; the previous 15,000-voxel preference now affects ranking only.
  • Region growth preserves diagonal connections between Z planes.
  • The Microglia Viewer reports the percentage of visible signal assigned within the current slice trim. Remaining unassigned signal stays visible in green.

Verification on the saved cortex reference

At unchanged thresholds, assigned above-threshold signal increased from 68.2% to 92.9%, and accepted cell candidates increased from 11 to 13. The correction assigns 398,481 additional above-threshold voxels to actual cell labels. No previously assigned voxels became unassigned.

All 13 candidates produced per-cell soma and process measurements on the native stack. The test suite passed in both CPU and DirectML development environments (239 passed, 3 skipped in each).

The remaining 7.1% is retained visibly as unassigned signal. Assignment coverage is not a claim of biological ground-truth accuracy.

Windows downloads

  • NVAP-0.1.6-windows-cpu.exe — standalone Windows x64 CPU build.
  • NVAP-0.1.6-windows-directml.exe — standalone Windows x64 build with DirectML GPU support and CPU fallback.
  • SHA256SUMS.txt — checksums for both executables.

After updating, rerun Segmentation and Analysis to refresh cell labels and measurements. Choose Cell: All to see the complete signal and assignment coverage.