Don't believe it, and don't dismiss it either. Use independent data and test it directly.
Pick a cancer or non-cancer whole-slide image, keeping the sample and diagnosis information. Frozen sections are a known exception — FFPE (paraffin-embedded) samples are recommended.
Download and install Community Edition, and use it to extract a sufficient number of nuclei (≥10,000 recommended) and generate the RSi distribution.
Record Tail Closed, Tail Runaway, borderline, or indeterminate — and note the sample source and tissue preparation method.
Independent data shows the same correspondence.
The phenomenon only appears under specific tissue or experimental conditions.
Cancer without Tail Runaway, or non-cancer with clear Tail Runaway.
Measure at least 10,000 nuclei, rather than relying on a small set of representative objects.
Measure only one RSi per nucleus, forming a viewable, comparable statistical distribution.
In current data, cancer tissue mostly presents as Tail Runaway; non-cancer tissue as Tail Closed.
Download Community Edition and reproduce it independently with your own SVS data.