Add OSIPI DCE Challenge as first dashboard data example

[ctwhome]

Context

Suggestion to use the OSIPI DCE challenge dataset as the first realistic data example for developing the OSIPY dashboard.

Relevant links

• Dashboard GitHub repo: https://github.com/OSIPI/dashboard
• Current public dashboard: https://osipi.github.io/dashboard/
• Challenge paper / PubMed: https://pubmed.ncbi.nlm.nih.gov/38115695/
• Challenge repository: https://github.com/OSIPI/TF6.2_DCE-DSC-MRI_Challenges
• OSF data page mentioned by the challenge repo: https://osf.io/u7a6f/
• RIDER Neuro MRI source dataset: https://doi.org/10.7937/K9/TCIA.2015.VOSN3HN1
• Earlier alternative considered: QIN Breast DCE-MRI, https://www.cancerimagingarchive.net/collection/qin-breast-dce-mri/

Proposed first dashboard use case

Use the OSIPI DCE Challenge as the initial realistic workflow for dashboard development:

1. Select the OSIPI DCE Challenge use case.
2. Show dataset/paper/source metadata.
3. Let the user point the dashboard to a local copy of the data.
4. Generate or edit the OSIPY YAML/configuration.
5. Validate the configuration.
6. Run or mock-run the OSIPY workflow.
7. Inspect outputs, especially Ktrans maps and optionally scoring results.

The challenge repository is public and contains useful material for development:

• ChallengeDICOMData/
• additionalDROData/
• Scoring/challengeScoring.py
• Scoring/Masks/
• OSIPI_DCE_Challenge_Guidelines.pdf
• expected maps/arrays such as Ktrans, vp, kep, AIF, and R10

The GitHub repo is large, around 2.3 GB. For dashboard development we should probably avoid making the dashboard depend on large binary data stored directly in GitHub.

Recommended data-location pattern:

• keep code, metadata, manifests, small examples, and documentation in GitHub;
• use OSF/TCIA/Zenodo or another citable data repository as the canonical data source for large data;
• have the dashboard work with a locally downloaded copy of the data;
• include a small fixture/demo path for fast UI development.

Dashboard implementation direction

Add a first Use case / dataset selection card for the OSIPI DCE Challenge.

Initial UI fields could include:

• use case name;
• modality: DCE-MRI;
• source links: paper, GitHub, OSF(Open Science Framework)/TCIA(The Cancer Imaging Archive) if applicable;
• expected local data folder;
• expected outputs: Ktrans, vp, kep, scoring outputs;
• status: fixture/mock mode vs full local dataset mode.

First flow:

Use case → Local data path / fixture mode → Config/YAML preview → Validation → Run/mock-run → Output-map/scoring preview

Acceptance criteria

• Decide which subset is the first dashboard path: synthetic data, clinical data, or both.
• Decide canonical data location for dashboard references: GitHub, OSF, TCIA, Zenodo, or other.
• Add OSIPI DCE Challenge as a selectable use case in the dashboard UI.
• Show paper/data/source metadata in the dashboard.
• Define expected local folder structure or manifest for the data.
• Support a lightweight fixture/mock mode so the dashboard can be developed without downloading the full data.
• Generate or display an OSIPY-compatible YAML/config draft for this use case.
• Define first output view: config validation, Ktrans map inspection, scoring, or a step-by-step combination.
• Avoid committing large datasets, patient/participant data, or local paths into the dashboard repo.

[ctwhome]

To discuss:

• Which exact challenge subset should be the first dashboard workflow?
• Should the first implementation use synthetic data only, clinical data only, or both?
• What should be considered the canonical data location: GitHub, OSF, TCIA, Zenodo, or another repository?
• Should scoring be part of the first dashboard slice, or should we first focus on config and output-map inspection?

[ctwhome]

Created a focused follow-up issue for choosing the smallest useful first subset:

• Identify minimal OSIPI DCE Challenge subset for first dashboard example #5

This keeps #4 as the broader OSIPI DCE Challenge use-case issue, and tracks the narrower decision: which synthetic/reference subset should drive the first dashboard example.