Cancer: results and interpretation

Reading a synthetic survival run

The example illustrates a Cox time-to-event workflow using synthetic clinical and patch data. The Cancer documentation identifies Harrell’s C as a survival evaluation measure used in its examples; real-data HANCOCK examples also document Uno’s C.

A concordance measure summarises how well risk ordering agrees with observed event-time information. It is not a probability of survival for an individual patient.

Questions for participants

• Which input modalities were available?

• How are events and censored observations represented?

• What is the difference between risk ranking and clinical calibration?

• Why must a synthetic-data score not be reported as clinical predictive performance?

Next steps

The real-data HANCOCK examples require access to the dataset, the expected local directory structure, and further compute. Consult docs/quickstart.md in the Cancer repository before attempting them.

Cancer Getting Started · Getting started · Hands-on activity · Quick Start.