Memorial LIS Extracts
De-identified cohort extracts pulled from the pathology LIS — the structured clinical backbone of nearly every study in the group, with documented reliability caveats.
What it is
Data pulls from the laboratory information system, cleaned and de-identified before they reach any repository. The underlying data model is case / specimen / block / slide with accession numbering.
Available content, per the LIS documentation: structured case and specimen fields; narrative report text (extracted via Report Text Extraction); several distinct timestamps — received, grossed, signed-out, amended; and quality signals including turnaround, amendment rates, and missing-field rates. ICD codes and report contents are retrievable from the Virapis system, and the ecosystem notes record a regex-based labelling file used to tag reports in production.
Three documented caveats define how this data must be used:
- Field reliability varies by subspecialty practice. The presence of a field does not mean it is populated consistently or trustworthy. The LIS repo documents which fields are reliable for research.
- Timestamps have specific documented meanings. Received, grossed, signed-out, and amended each mean something particular; guessing produces wrong turnaround figures.
- Accession numbers are identifying even without names, because their patterns are. Research exports use hashed accession IDs exclusively.
Why it matters for my work
This is the join key for the whole research estate: it links image cohorts on the Memorial WSI Share to clinical outcomes, and it supplies the raw material for the operational and text-mining work. Anything that goes wrong here propagates everywhere, which is why the reliability caveats deserve more weight than their brevity suggests.
How it connects
Patoloji Bilgi Yönetim Sistemi (LIS) — the project page documenting the system, its interfaces, and the request process.
Report Text Extraction — converts the narrative portion into structured fields.
Turnaround Time — computed from these timestamps, and only as valid as their definitions.
De-identification — the mandatory stage between the LIS and any repository.
Memorial WSI Share — the image counterpart, joined by hashed accession ID.
Konsultasyon Log — a specialised extract for consultation cases.
Open questions
- Which fields are documented as reliable, and where is that list? It is referenced but not reproduced in the portal, so it is effectively invisible to anyone not already in the repo.
- Is there a standard cohort-request form and turnaround for research pulls? A dataset-request process is mentioned as a shared resource but with placeholder links.
- Are extracts versioned and archived, so a published analysis can be regenerated from the same snapshot?