Identify minimal OSIPI DCE Challenge subset for first dashboard example

[ctwhome]

Context

Following suggestion in #4, we should identify the smallest useful subset of the OSIPI DCE Challenge data to use as the first concrete dashboard example.

The goal is not to support the full challenge dataset immediately. The goal is to choose one small, defensible DCE example that can drive the first dashboard workflow.

Proposed first subset

Start with the OSIPI DCE Challenge synthetic data, preferably one minimal case such as:

• synthetic data rather than clinical data
• one patient/case
• one visit/timepoint, if applicable

Working hypothesis for discussion:

Use one synthetic DCE Challenge case as the first dashboard example, then add clinical/public datasets later.

Why synthetic first

Synthetic/reference data is a better starting point for dashboard development because it is:

• directly tied to OSIPI
• DCE-specific
• safer and simpler than clinical data
• likely to have known/reference outputs
• suitable for validation and regression testing
• small enough to make a focused dashboard path

Clinical data can be added later as the realistic demo once the dashboard flow works.

Initial local development fixture

For UI development, we can initially use the existing OSIPY fixture data instead of downloading/handling the full challenge data:

/Users/ctw/Sites/osipy/tests/fixtures/osipi_dro/dce/

Relevant files include examples such as:

• Ktrans.nii
• kep.nii
• vp.nii
• R10.nii
• AIF.npz
• Ea.npz

This can support an early dashboard flow with fixture/reference data while we decide the canonical challenge subset.

Candidate dashboard flow

1. Select use case: OSIPI DCE Challenge - Synthetic example
2. Show dataset/paper/source metadata
3. Let the user select a local data folder
4. Detect expected files/inputs
5. Generate or preview OSIPY YAML/config
6. Validate the configuration
7. Run or mock-run the DCE workflow
8. Inspect output maps, starting with Ktrans
9. Later: add comparison/scoring against reference outputs

Questions for Thursday

• Which exact synthetic DCE Challenge subset should be first?
  • one synthetic patient/case?
  • one visit/timepoint?
  • both synthetic cases/visits?
• Should the first output focus only on Ktrans, or also include vp, kep, R10, and AIF?
• Should challenge scoring be part of the first dashboard demo, or a follow-up?
• Should OSF be treated as the canonical source for the challenge data?
• What is the smallest success criterion for the first dashboard example?
  • config validates?
  • OSIPY runs?
  • output map appears?
  • scoring works?

Acceptance criteria

• A recommended first DCE Challenge subset is selected and documented.
• The subset has a clear source link and expected local folder structure.
• Required input files and expected output/reference files are listed.
• The initial dashboard path for this subset is sketched.
• Follow-up implementation issues can be created from this decision.

Related: #4