Biological state is already visible.
Much of it simply hasn't been measured.
NucleoScope turns biological images into quantitative measurements of state.
Our first observed readout: Tail Closed / Tail Runaway in H&E WSI.
Pathology has long relied heavily on qualitative interpretation of nuclear morphology. NucleoScope asks whether a reproducible quantitative nuclear-state measurement can add an objective readout. In the data analyzed so far, this readout has not required separate rules for different species, cancer types, or tissue sites — a pattern still open to independent testing.
Our long-term goal is to explore whether nuclear-state measurement can play a thermometer-like role: a reproducible quantitative reading independent of the observer.
This separation matters: even if a future explanation of why this pattern exists is revised or overturned, that would not, by itself, overturn the reproducible measurement or the observed statistical correspondence.
We measure the statistical organization of large nuclear populations.
Cancer‑associated versus non‑cancer WSI repeatedly corresponds to one statistical feature of the RSi distribution.
Why do apparently different pathological conditions converge to a similar statistical morphology pattern?
Not addressed here. The biological mechanism remains an open question for future work.
Tail Runaway appears to persist across sufficiently large local nuclear ensembles within a cancer‑associated WSI, without using pathologist‑defined tumor regions to select those ensembles.
3.79M nuclei · 378 consecutive 10k‑nucleus blocks · 1,000 random neighborhoods
Tail Runaway is reproducible across independently sampled local regions within a cancer‑associated slide. The readout does not require histologic ROI labels as an input.
Histological tumor boundaries need not be nuclear‑state boundaries.
Pathology describes tissue identity and spatial classification; NucleoScope measures nuclear structural‑state distributions. These are different questions — they need not produce the same boundaries.
Formal comparison with pathologist‑annotated tumor and non‑tumor ROIs remains to be performed.
Extracts large‑scale nuclear structural variables from SVS images into a statistical distribution and spectrum readout — independent of any pathologist's label, reproducible on the same image.
In the datasets analyzed so far: Cancer‑associated WSI → Tail Runaway / Non‑cancer WSI → Tail Closed, across species and tumor types. Known exception: frozen sections.
You're invited to use your own SVS data and the free software to reproduce, revise, or falsify this observation.
Both tools can derive per‑nucleus measurements from WSI. QuPath emphasizes flexible object‑level analysis; NucleoScope applies a predefined population‑level measurement and spectrum readout.
AI pathology systems learn predictive relationships from training data. NucleoScope takes a different approach: it applies the same predefined measurement rule to each WSI and examines the resulting nuclear‑state distribution.
Generalization depends on training distribution and task‑specific labels.
The same predefined measurement rule is applied regardless of species, cancer type, or tissue site. Whether the observed Tail correspondence generalizes across these contexts remains an empirical question.
Why download NucleoScope?
SVS → Tail Status. Not just a CSV, not just a distribution — directly, 🟢 Tail Closed / 🔴 Tail Runaway.
In the data analyzed so far — human, rat, dog, 33 TCGA cancer types, multiple tissue sites — the same RSi / Tail readout process applies. No need to build separate rules for each cancer type.
Download the software → use public SVS data → run it yourself → judge whether the result holds. Anyone can verify it independently.
One frozen diagnostic WSI from each of the 46 TCGA-OT cancer classes was measured with the same nuclear-state pipeline. All 46 showed Tail Runaway. No exception found so far.
Inspect the exact slides, download the individual per-nucleus CSV files, or choose another public TCGA slide and try to find an exception.