2026-11-12T00:26:11.176403+00:00
PLISM paired HistoQC field admission. Freeze made after verification of all native footprints and before computing or reading any mask.
Population: 13 serial human TMA H&E sections with pre-existing AT2-reference tissue sampling, 12 registered 128 µm fields per section. Include all 156 fixed fields including the three previously excluded by the earlier two-scan optical coverage criterion. Use native pixel-centre polygons from maps/. No new fields or replacement.
Whole-slide image processing: upstream HistoQC revision f35e518d875e3e3799f89a7aed4e766c5a9d35e4, default config (config.ini), complete scanned extent, one shared set of classifier models across all scans, one process, seed 20261111. Score only final mask_use. Use complete native reduced-resolution and associated-image payloads; abort any QC baseline request on range-acquired source files. Missing mask, incomplete scan, or unsupported geometry is a technical failure.
Project field pixel edges into each mask using its actual width/height over the verified full scan extent. Compute exact polygon overlap with positive mask pixels divided by polygon area. Study-defined field inclusion is >=0.50 of the physical field area (ties included), applied identically. Record continuous fractions for both scanners. Four possible decisions per field; primary is fraction changing inclusion per section, then arithmetic mean across the 13 sections. Report both directions per section and all three prior exclusions. No direction or minimum effect presupposed. These serial sections are correlated; descriptive results conditional on this array, sampling and scanners, no independent-patient inference or declaration of which scanner is right. No outcome data or private measurements.
