Adaptive Whole-Tissue Measurement for QuPath
A reproducible exact/adaptive whole-tissue measurement system for QuPath pixel classifiers, developed from a production pathology workflow and generalized into a parallel, resumable measurement architecture with statistically controlled early termination.
Whole-tissue exact measurement
The validated reference workflow exhaustively partitions annotated tissue, measures classifier-positive and classifier-negative area, and pools those measurements into a whole-tissue result.
Parallel, resumable execution
QuPath remains the measurement kernel. Nextflow owns image-level dispatch, isolation, retry and provenance; Make exposes simple human-facing verbs for exact measurement, estimation, verification and resume.
Adaptive stopping, after parity
The adaptive path will use spatially balanced sampling with a finite-population ratio estimator and LLN/CLT-derived precision criteria. It remains disabled until exact numerical parity is demonstrated.
Current status. The exhaustive kernel has been promoted into the new workflow surface. The next hard gate is runtime parity against the original production implementation. Adaptive stationary estimation is intentionally unavailable until that comparison passes.