Onboarding

Start here.

Five things to have, three steps, and then your first question.

What you needFive things, once.

  1. 01Wayne State AccessIDYour university sign-in, the one you use for Academica.
  2. 02Grid accountApply once, with the grid account request form.
  3. 03GlobalProtect VPNConnect it before you open the grid's portal.
  4. 04Claude Pro or MaxYour own plan, which includes Claude Code.
  5. 05BeethovenThe Chrome extension. Ask for access. Private listing, by invitation: after approval you receive an invitation at the Google account you gave.

Three stepsConnect once. Then ask.

  1. Open OnDemand and sign in.

    With GlobalProtect connected, open ondemand.grid.wayne.edu and sign in with your AccessID (how to connect).

  2. Click the Beethoven icon, once.

    It does the rest: it points your account at the lab's copy of Beethoven on the grid, starts your Claude Code worker there, and asks you to sign in to Claude one time.

  3. Open claude.ai or the Claude app, and ask.

    In Code, pick the environment named after your site, and type your question.

The worker is a job on the grid, not a program on your laptop: close the browser and it keeps going. You need the portal again only to stop it, or to start a new one.

Your first questionAsk it the way you would ask a colleague.

Is the female pelvis relatively wider?

  1. Opens the labels of CTSpinoPelvic1K; dataset_labels.json says which identifier is which bone.
  2. Finds each femoral head centre (identifiers 32 and 33) and the width of the L4 body (23).
  3. Divides the distance between the heads by the L4 width, for every record.
  4. Compares females and males: a Welch test and Cohen's d.
  5. Draws the three panels below.
  6. Writes the report from the CSV it just made. The heavy pass runs as a SLURM job.
Three box plots by sex: bi-iliac width, femoral head distance, and head distance relative to L4 body width
Pelvis width (a), hip joint separation (b) and separation relative to L4 body width (c), by sex, from the released labels of 791 records.
  • 164 vs 162femoral head distance, female vs male, mm (mean)
  • 50.2 vs 55.4L4 body width, female vs male, mm (mean)
  • 3.29 vs 2.94head distance / body width, female vs male
  • d = 1.09Cohen's d; Welch p = 2 × 10-43

The answer is yes. Male pelves are wider at the iliac crests in absolute terms (a), and the femoral heads sit about 163 mm apart in both sexes (b); on a smaller skeleton that is a relatively wider pelvis: 3.29 against 2.94 (391 females, 338 males) (c). The published ratio is 3.95 against 3.48; ours is lower because a body width taken from labels runs a few millimetres wider than a caliper endplate width, and the female to male ratio, 1.12 against 1.14, agrees. The far outliers are records where a femoral head was not where the label said; the medians do not move.

Every number above is in results/report.txt, which the script writes from results/pelvic_width.csv. The demo study's code is one 120-line script.

Your dataWhat goes where.

  • You sign in to Claude with your own account, on your own plan.
  • No protected health information, FERPA-protected records, data under an IRB protocol that has not approved this use, or export-controlled data in the working directory.
  • Your prompts, the files the worker reads and every command's output go to Anthropic, and the session transcript is stored on Anthropic's servers.
  • On a personal plan Anthropic keeps it five years if model training is on, and 30 days otherwise. No BAA covers a personal plan.
  • The details: Beethoven's privacy statement and Anthropic's privacy policy.

Other ways to get involvedJoin the consortium.

Whether you are a radiologist, an ML researcher, a data engineer, or a clinician with imaging to contribute — there is a place for you here. The consortium grows through public pull requests and open conversation.

  • Contribute code or data

    Open an issue or a pull request on GitHub. Dataset contributions are reviewed against our origin and licensing checklist.

    GitHub organization →

  • Collaborate on research

    Interested in co-developing a benchmark or a study? Reach out to the project lead to start a conversation.

    gregory.schwing@med.wayne.edu

  • Students: start here

    Templates, the demo study above, and a list of project ideas, in one public repository.

    Student_Projects →

Go deeper with the six workshop tutorials.