HER2 Gastric Cohort
Gastric and gastroesophageal cases with HER2 IHC and reflex ISH, plus scanned slides and QuPath annotations — the curated basis of the HER2 heterogeneity and agreement work.
What it is
A curated cohort with three linked layers:
- Cases — gastric and GEJ resections and biopsies with HER2 IHC and reflex ISH. Cohort identification is complete and the IHC/ISH database is curated.
- Images — de-identified scanned slides on the Memorial WSI Share.
- Annotations — QuPath annotations exported as GeoJSON per the group SOP, following the
{organ}__{class}__{reviewer}.geojsonconvention.
One structural feature of this cohort is a genuine analysis constraint rather than a housekeeping note: it contains both biopsies and resections, and HER2 scoring rules differ between them — the biopsy rule accepts focal positivity, the resection rule does not. The specimen-type column is therefore mandatory in any aggregation, and pooling without it produces a meaningless combined score.
A second: slide-level tumour percentage is required for downstream heterogeneity metrics, so tumour-area annotation cannot be skipped even for cases where only the HER2 score is of interest. That is a data-collection decision which is expensive to retrofit.
Cohort size, date range, and scanner are not recorded. [unverified]
Why it matters for my work
It supports all four objectives of HER2 Intratumoral Heterogeneity simultaneously — interobserver agreement, heterogeneity mapping, IHC/ISH concordance, and biopsy-versus-resection divergence — which makes it unusually efficient: one curation effort, four questions. The pilot scoring round is complete and heterogeneity annotation is active.
How it connects
HER2 Intratumoral Heterogeneity — the project page with objectives and status.
Intratumoral Heterogeneity — the concept this cohort was assembled to measure.
Interobserver Agreement — the pilot scoring round’s analysis framework.
QuPath Annotation Workflow — how the annotation layer is produced and exported.
Memorial WSI Share — where the images live.
Open questions
- How many cases, over what period, and on which scanner? Not recorded. [unverified]
- What is the biopsy-to-resection ratio? It determines whether the divergence sub-question is adequately powered.
- Do the same cases drive both IHC/ISH discordance and interobserver disagreement? If so that is a strong finding, and the cohort is already assembled to answer it.