Cancer: hands-on activity

Cancer: hands-on activity#

Note

This is a proposed session activity using the synthetic survival example.

1. Run and observe#

From the Cancer repository root:

uv run python examples/synthetic/synthetic_survival.py

Record:

  • Which data types are combined

  • Which outcome is predicted

  • Which evaluation metric is reported

2. Follow the pipeline#

Open examples/synthetic/synthetic_survival.py and find:

  1. Creation of the synthetic clinical table and patch features.

  2. Generation of survival times and censoring indicators.

  3. Construction of the model/pipeline.

  4. Training and evaluation.

3. Discuss the research question#

  • Why is censoring important for time-to-event outcomes?

  • Why should records from one patient remain in the same data split?

  • What information might pathology patches add to clinical covariates?

Optional advanced reading#

The Cancer quickstart documents HANCOCK examples, including examples/hancock/survival_intermediate.py, from_config.py, cross_validation.py and classification_late.py. These require a local HANCOCK dataset and are not part of the no-data starter exercise.

Getting started · Results and interpretation · Participation guide.