Minimum Viable Cohort
What's the minimum needed to join?
We keep the bar deliberately low so any HGSOC center can contribute. A founding-quality cohort is a small, well-characterized set of cases that the model can actually learn from.
Cohort & slides
- 01 At least 50 HGSOC cases. More is better, but 50 is the floor for a meaningful contribution to the foundation model.
- 02 H&E whole-slide images, stained within the last 5 years. Stain quality has the biggest single effect on model performance, so recent slides keep batch variability in check.
- 03 One slide from the primary tumor, and one from a metastatic site where available. Two-site cases let the model learn intra-patient heterogeneity.
- 04 From primary debulking surgery (PDS), with no neoadjuvant chemotherapy. Treatment-naive tissue only, so the model learns underlying biology rather than chemo-induced morphology.
- 05 Scanned at 40× magnification (~0.25 µm/pixel). 20× scans work as a fallback if 40× isn't available, but 40× is the target.
Pulling slides
Practical guidance from the pathology side on which slides to pull from each case.
Priority
Non-NACT (PDS) cases
Use pre-treatment resection slides only, never biopsies. If there are no metastases, take the single slide with the highest tumor percentage. If metastases are present, take at least one adnexal slide plus at least one metastasis slide. Prioritize omentum mets, but include multiple metastatic sites when they appear morphologically different to the eye.
NACT cases
Take the pre-treatment biopsy and the post-treatment resection.
Clinical fields
Five fields per case. The first four are required; HRD score is optional but materially improves model performance where available.
| case_id | slide_id | age at debulking | relapse* | time to relapse (mo) | HRD score |
|---|---|---|---|---|---|
| OV-001 | OV-001-P | 62 | 1 | 14 | 72 |
| OV-001 | OV-001-M | 62 | 1 | 14 | 72 |
| OV-002 | OV-002-P | 58 | 0 | NA | 41 |
| OV-003 | OV-003-P | 71 | 1 | 8 | 65 |
| OV-003 | OV-003-M | 71 | 1 | 8 | 65 |
| OV-004 | OV-004-P | 49 | 1 | 22 | 88 |
| OV-005 | OV-005-P | 66 | 0 | NA | 18 |
| OV-006 | OV-006-P | 74 | 1 | 5 | — |
| OV-006 | OV-006-M | 74 | 1 | 5 | — |
*Relapse follows the standard GCIG (Gynecologic Cancer Intergroup) definition: the earliest of (a) radiologic progression by RECIST 1.1, (b) CA-125 elevation meeting Rustin criteria — CA-125 ≥ 2× the upper limit of normal confirmed on a second sample at least one week later, or doubling from nadir in patients whose CA-125 normalized, or (c) clinical / symptomatic progression. Time to relapse is measured in months from the date of completion of first-line platinum-based chemotherapy.
Ready?
If your cohort meets the floor — or you'd like to discuss how close yours is — leave us your email and we'll set up a call.
Get more information →