Scanner Logs
Per-slide scanner and transfer event logs from the Leica scanners and the Sectra transfer path — the raw material for the group’s operational workflow research.
What it is
Event-level records of scanning and transfer, analysed in the scan-time-in-real-life repo (R + Quarto). The repo ships a sample CSV used for the onboarding reproduction task, so the schema is discoverable there.
Two documented data properties determine how the logs must be handled, and both are cleaning traps rather than analysis choices:
- Timestamps are local time with DST. The repo already normalises to UTC. Applying a second correction produces an hour-shifted dataset that looks entirely plausible and invalidates every hour-of-day figure.
- Some logged “failed transfers” are benign retries that later succeeded. The analysis de-duplicates by
(scan_id, final_outcome), and that logic must be preserved or the error rate is inflated.
Derived measures include hour-of-day and day-of-week load patterns, case-level timing metrics, and queueing indicators. Coverage is incomplete by design: the onboarding task asks contributors to flag transfer steps that are not instrumented, which implies some steps are currently invisible to the logs. [unverified]
Total record count and date range are not documented. [unverified]
Why it matters for my work
This is the evidence base for Scanning Time in Real Life, whose deliverables are a journal figure set, an internal optimisation memo, and an operations dashboard. It is also the clearest case in the repo where the scientific validity rests on data cleaning rather than statistics — the two traps above are the study.
How it connects
Scanner Log Analysis — the method page describing the cleaning and analysis pipeline in full.
Turnaround Time — scanning and transfer are components of the digital workflow’s overall turnaround.
Whole Slide Imaging — file size is why transfer is a bottleneck worth instrumenting.
WSI Quality Control — quality failures cause rescans, which appear here as additional load.
Open questions
- What is the date range and record count? Not documented. [unverified]
- Which transfer steps are un-instrumented, and can they be? That gap bounds what the analysis can conclude.
- Is scanner identity retained per record? It is both an operational finding and a potential confounder for pooled timing analyses.