NucleoScope's job is to measure and to surface the phenomenon. Validation and explanation belong to the scientific community.
The same slide, read by different doctors, can get different verdicts — that's not any one doctor's fault. The field has long lacked a ruler that doesn't depend on personal experience and doesn't change with the observer. RSi is our attempt at that ruler: a measurable, reproducible objective structural-state variable. It isn't meant to replace pathologists — it's meant to give doctors and researchers a shared, objective reference.
"AI cancer predictor" apps are already common knowledge: show a model thousands of slides labeled by doctors, and it learns to predict "what a doctor would probably say." That's easy to grasp because it's imitation — like training a student to guess the teacher's grade by repeatedly checking answers.
NucleoScope's nuclear-state-spectrum approach takes a different path, which is why it's less intuitive on first hearing: it doesn't learn from doctors' judgments. It computes a number (RSi) directly from image structure, with no need to see how any doctor labeled anything first — more like building a new ruler that measures the slide itself, rather than learning how someone else scores it.
Yes — nucleus segmentation uses AI / deep learning. But "does it use AI" isn't what separates "measurement" from "subjective judgment"; reproducibility is. Run the same SVS image at any time, by anyone, and the RSi and Tail Runaway result comes out the same. AI in our pipeline is used for "finding where the nuclei are" (segmentation) — a task that doesn't require any cancer label to train. What actually produces the Tail Runaway readout is an independent structural computation and statistical distribution — not a model trained to predict cancer.
1｜Measurement: RSi is computed directly from image structure. The same SVS gives the same result no matter who runs it, independent of anyone's experience or state of mind.
2｜Empirical phenomenon: The correspondence between Tail Runaway and cancer was found by statistically comparing many independent measurements against known labels — it's a pattern found in data, not one person's gut feeling. A doctor's subjective experience can't be separated from "which doctor, what training, what kind of day they're having." "Measurement + empirical phenomenon," by contrast, can be reproduced by anyone using the same free software, at every step.
One person's access to data, compute, and time is limited. We tried to find counterexamples and, within the range analyzed so far, found one (frozen sections) — so we're publishing the measurement tool and inviting more independent teams to keep testing it.
The old workflow: SVS → CSV → your own analysis → your own judgment. That works, but every cut is manual — like swinging an 🪓 axe. The new workflow: SVS → Tail Status. That's a 🪚 chainsaw. The tree hasn't changed. The tool has.