Digital pathology · Delivery validation

From glass to platform, end to end, and reproducible.

An eight-phase delivery path for digital pathology, from a departmental scanner to an enterprise platform. Every phase separates field practice from our delivery steps, names what to measure, and ends at an exit gate you can validate.

Walk the phasesRun the numbers
Delivery playbook · Rev A · October 2026
ACQUISITION · SCANNERS, ANY VENDOR MANAGEMENT · IMS VIEWING · PATHOLOGISTS, CONSULTS STORAGE · ENTERPRISE ARCHIVE LIS · CASE OF RECORD
Fig. 0 · The platform modelFour layers, each replaceable

How to read it

Practice, steps, measures and a gate in every phase.

Each phase separates what works in the field from the delivery steps we add, and every number that has to come from your lab is marked as a measurement, never assumed. A phase is done only when its exit gate is true.

In the drawings, solid green is what is proven in practice, hatched is our synthesis, and dashed is what is planned or still has to be measured.

Where you start

The maturity path.

Few labs start with an enterprise platform; getting there can take a decade. Each stage is useful, but only the last one makes digital pathology something clinicians depend on, and at that point it has to be run as a service to the whole institution: scale, viewer performance, retention, disaster recovery and a known security contact.

Locate your lab on this path first. The phases below start from wherever you are; a lab at stage 3 has already done part of Phase 03.

1 SCANNER + DEPARTMENTAL IMS 2 RESEARCH, INTERESTING CASES 3 LIS-INTEGRATED · PARTIAL 4 FULL SCALE · DEPENDED ON
Fig. A · Maturity pathGreen: clinicians depend on it

The delivery · eight phases

The optimal path, start to finish.

Each phase lists what works in the field, the steps we add, what to measure and the exit gate that must be true before the next phase. The phases map to how we work: Phase 00 is the audit, Phases 01 to 05 are implementation, and Phases 06 and 07 are the retainer.

00Before any vendor

Readiness and baseline

Measure the lab before you size the platform.

Digital pathology changes a physical flow into a data flow. Before choosing anything, count what the histology lab produces per hour, where slides go today, how often priors and consults move glass, and how much data the scanners will create. These numbers size scanners, staff shifts, network and storage, and they become the baseline every later phase is judged against.

What works in the field

  • The value in radiology came from throughput and workflow flexibility, not from film savings; glass is not going away in pathology either
  • Large dataset management, archiving, retention and network optimization are non-trivial
  • Slide volume grows quickly once clinicians depend on digital; size for growth, not for today

Delivery steps

  • Hourly histology output for at least two representative weeks
  • Slide volume by case type, stain type and site
  • Measured file size per slide, by scanner and magnification
  • Use cases ranked: primary sign-out, priors, consults, conferences, research, AI
  • Glass-retention policy decided up front, even if the answer is "keep it for now"

Measure

Output per hour, by hour of daySlides per year, and growth rateGB per slide, by scanner
Exit gateHourly output, slide volume and GB per slide are measured, not estimated.
OUTPUT > CAPACITY · BACKLOG GREEN: SLIDES SCANNED PER HOUR
Fig. 1 · Readiness and baselineIllustrative · use the calculator
01Team before vendor sign-off

Assemble the village

Borrow the hospital's existing IT teams, and hire the one team you lack.

Most of what digital pathology needs already exists in a hospital: infrastructure, EMR and LIS analysts, security, storage, network, data and AI teams. What is missing is a pathology informatics team that sits at the pathologists' elbow, owns the IMS relationship and triages technical issues. The team works best hired for the role, placed in the enterprise imaging organization and brought in while vendors are being finalized, so the knowledge built during implementation stays in the building.

What works in the field

  • Leverage your existing village: infrastructure, EMR, security, storage, network, data and AI teams
  • A pathology informatics team sits with the pathologists, owns the IMS relationship and triages issues
  • Create the team while vendors are being finalized, hire for it, and place it in enterprise imaging
  • A multidisciplinary project team is temporary; the support model must survive go-live

Delivery steps

  • RACI across lab, pathologists, informatics, LIS analysts, storage, network, security
  • Informatics team in place before the contract, so custom work is understood by its owners
  • Named escalation path for security vulnerabilities in the imaging stack

Measure

Open tickets and time to resolutionShare of custom integrations with a named owner
Exit gateEvery system in the chain has a named owner who will still own it after go-live.
PATHOLOGY INFORMATICS TEAM LAB + PATHOLOGISTS EXISTING IT TEAMS
Fig. 2 · Assemble the villageHatched: lab and pathologists
02Not a scanner-centric silo

Platform architecture

Separate acquisition, management, viewing and storage.

A scanner tightly coupled to its own IMS works until you buy a second scanner, accept outside images or run out of storage. The platform model keeps the four layers separate, so any one can be replaced without a whole-system upgrade. Storage follows the hospital's existing enterprise model rather than a pathology-only design.

What works in the field

  • Move from a scanner-and-IMS pair to a platform that supports scanners from several vendors and outside images
  • Separate acquisition, management, viewing and storage so components can evolve and be replaced
  • Reuse the enterprise storage model: what runs on SAN, what on NAS, what goes to the cloud
  • The IMS is the main pathology-specific choice; most other components already exist in the enterprise

Delivery steps

  • Vendor-neutral requirements for the IMS: multiple scanners, outside images, open export
  • Storage tiers, retention, backup and disaster recovery sized from Phase 00 numbers
  • Network paths sized for whole slide images between scanners, IMS and storage
  • DICOM support checked now, even if it is used later

Measure

Storage needed per year (calculator below)Peak transfer rate from scanners
Exit gateEach layer can be replaced without touching the others, on paper and in the contract.
SCANNERS · SEVERAL VENDORS IMS VIEWERS ENTERPRISE STORAGE · VNA
Fig. 3 · Platform architectureEach layer replaceable
03The LIS stays the case of record

LIS and IMS integration

Make the pathologist's worklist tell the truth about every slide.

Pathologists live in the LIS. The integration has to show, without guesswork, whether a case's slides are available, pending a stain, or not there yet, and open the right images for the right patient. Identity is the hardest part: a single character difference between LIS and IMS can create duplicate patients.

What works in the field

  • A static "scanned" icon is not enough once stains are ordered; use green, yellow and red
  • Show the digital view link only on slides that are actually available
  • A one-letter mismatch between IMS and LIS can create duplicate patients that then have to be merged
  • Outside cases get an internal accession tied to the outside case number; the outside slide number stays on the label and in the IMS
  • Keep gross and other lab images in the IMS, where pathologists work
  • Use a sharing mode that hides patient information at conferences

Delivery steps

  • Identity rules: which fields must match, which system wins, how mismatches are caught before a duplicate is created
  • Status model for every case and slide: available, stain pending, none
  • Outside-case mapping for slides and for blocks
  • Test script of real case types, rerun after every LIS or IMS update

Measure

Duplicate patients or cases created per 1,000 casesCases where status and reality disagree
Exit gateZero duplicate identities in the test set, and the status matches reality on every test case.
LIS · CASE AND PATIENT OF RECORD IMS · IMAGES GREEN · AVAILABLE YELLOW · STAIN PENDING RED · NONE
Fig. 4 · LIS and IMS integrationSolid · hatched · dashed = the 3 colors
04QC that scales

Scan quality and validation

Catch missed tissue at the scanner; catch blur from the pathologist; audit both.

Checking 100% of scans by hand does not last. The sustainable split: scan technicians use the scanner's rack preview to confirm the whole tissue area is captured, pathologists report blurred scans from inside the IMS, and those slides go back for rescanning. A periodic 100% snapshot measures how much blur goes unreported. Some findings, such as H. pylori on H&E, are known to be harder to see digitally, which is one reason glass stays on site.

What works in the field

  • 100% manual QC does not scale
  • Scan technicians confirm full tissue coverage with the scanner's rack preview
  • Pathologists report blurred scans in the IMS; slides are retrieved or dropped in a scan bin for rescanning
  • A periodic 100% snapshot shows how much blur goes unreported
  • H. pylori is not reliably seen on digital H&E; IHC is not a problem; glass stays available
  • The informatics team owns the regulatory documentation

Delivery steps

  • Validation study design and acceptance criteria agreed with the medical director
  • Rescan workflow with an owner and a time target
  • Quarterly 100% snapshot to estimate the true blur rate
  • List of findings that default to glass or a stain

Measure

Reported rescan rateTrue blur rate from the snapshotUnder-reporting ratio
Exit gateRescan rate and under-reporting are measured, with a named owner for each rescan.
RACK PREVIEW · FULL TISSUE SCANNER BLUR REPORTED · RESCAN BIN
Fig. 5 · Scan quality and validationDashed slide: reported blurred
05Comfort, not a cutover date

Go-live and adoption

Let pathologists choose digital, and measure how fast they do.

Go live without a date to take glass away. Adoption grows month by month, including among reluctant users, when their annoyances are found and fixed. Consults from outside should arrive as a regular LIS case, so pathologists never log into a separate system.

What works in the field

  • Go live with no set date to withdraw glass
  • Adoption rises month by month as users get comfortable
  • Reluctant users convert when their common annoyance is found and fixed
  • Digital consults arrive in the LIS like any other case; no separate login

Delivery steps

  • Go-live support at the elbow for the first weeks
  • Adoption dashboard per pathologist and per case type
  • Glass-retirement criteria written down, even if not yet triggered

Measure

Digital slide views ÷ slides signed out, per monthUsers active digitally, per month
Exit gateAdoption is reported monthly, and every reluctant user's main annoyance is on the backlog.
DIGITAL USE BY MONTH GLASS · STILL ON SITE
Fig. 6 · Go-live and adoptionIllustrative · measure monthly
06The support model that survives go-live

Operate and improve

A quarterly loop, and a standing check that scanning keeps up with histology.

After go-live the system keeps changing: IMS upgrades, interface changes, tickets. The informatics team reviews feedback by theme every quarter, passes the key items to the IMS and LIS vendors, and tells users what changed. Separately, hourly scanning throughput is compared with hourly histology output; any hour where output exceeds capacity becomes backlog, and pathologists wait. The fixes: more histology shifts, more scanning shifts, more scanners.

What works in the field

  • Feedback reviewed thematically every quarter, passed to vendors, tied to quarterly updates, and communicated back
  • Slides scanned per hour are compared with slides produced per hour; a deficit means a backlog
  • Responses: add histology shifts, scanning shifts or scanners
  • Every slide off the stainer should go straight onto a scanner

Delivery steps

  • Quarterly feedback review with the medical director
  • Weekly capacity report by hour of day
  • Regression test script after every upgrade

Measure

Backlog at end of each hour (calculator below)Stainer-to-available time, median and 90th percentileFeedback themes closed per quarter
Exit gateBacklog returns to zero within each shift, and every quarter ends with a user update.
1 FEEDBACK 2 THEMES · QUARTERLY 3 VENDOR UPDATES 4 BACK TO USERS
Fig. 7 · Operate and improveOne turn per quarter
07Standards for what comes next

Extend

DICOM connects pathology to the imaging infrastructure the hospital already has.

Radiology already routes DICOM studies to AI, to other hospitals and into one archive with one retention policy. Pathology can join that infrastructure instead of building its own: DICOM routers can send studies to AI, consults can travel with their demographics attached, and one archive can hold everything. Algorithms, AI checks for blur and missed tissue, and a home for non-patient QC work are the next steps and open problems.

What works in the field

  • DICOM routers already route studies to AI infrastructure; multimodal AI could follow
  • Hospital-to-hospital exchange in radiology is DICOM-based, with demographics attached
  • One storage, retention and backup strategy in the archive is an easier conversation
  • AI checks for blur and missed tissue are still emerging
  • Non-patient work such as lot validations and antibody workups usually has no home in the LIS

Delivery steps

  • DICOM conformance and archive integration plan
  • AI intake process: validation, monitoring and who reviews results
  • A repository decision for non-patient QC slides

Measure

Share of slides stored in the enterprise archiveAI results reviewed and overridden
Exit gateEach extension has a validation plan and an owner before it touches a live case.
AI INFRASTRUCTURE OTHER HOSPITALS ENTERPRISE ARCHIVE MULTIMODAL AI · LATER
Fig. 8 · ExtendCenter: DICOM router · dashed: planned

Run the numbers

The metrics, written so anyone can recompute them.

Each metric is defined by its formula and its inputs. Use the same definitions every time, and two people looking at the same data get the same answer.

Capacity balance

Δ(h) = produced(h) − scanned(h) backlog(h) = max(0, backlog(h−1) + produced(h) − capacity(h))

Per hour of the day. A positive backlog at the end of a shift means pathologists wait.

Time to availability

t_avail = time slide is viewable − time off the stainer report median and 90th percentile, per day

The pathologist-facing measure of the same balance.

Rescan and under-reporting

reported rate = rescans reported ÷ slides scanned true rate = blurred in a 100% snapshot ÷ snapshot size under-report = true rate ÷ reported rate

Run it on a fixed sample size each quarter.

Adoption

adoption(month) = slides viewed digitally ÷ slides signed out active users(month) = pathologists with ≥ 1 digital sign-out

Per pathologist and per case type, monthly; watch the trend.

Identity integrity

duplicates per 1,000 = duplicate patients or cases created ÷ cases interfaced × 1,000

A one-letter mismatch is enough to cause one. The target in validation is zero.

Storage forecast

TB per year = slides per year × GB per slide × copies (primary + disaster recovery) ÷ 1,000 TB held = TB per year × years retained

GB per slide must be measured on your scanners.

Capacity calculator

Slides the histology lab produces each hour of a shift, against what your scanners can scan per hour.

–Enable JavaScript to calculate; the formula is above.

Storage calculator

Use your own volume and your own measured GB per slide. The default volume is an example.

–Enter GB per slide to calculate.

Still open

What is still unsolved in the field.

An honest plan carries these as known risks rather than hiding them.

  • When and how to withdraw glass as the backup
  • Where non-patient QC work (lot validations, antibody workups) should live
  • AI-based detection of blur and missed tissue
  • Digital reliability for H. pylori on H&E
  • Storage strategy as volume grows
Phase 00 · Audit

Baseline

Hourly output, slide volume, GB per slide, maturity stage and the gap list.

Chain of Custody program →
Phases 01–05 · Implementation

Platform

Village, architecture, LIS and IMS integration, QC and go-live, each with its exit gate.

Epic Beaker →
Phases 06–07 · Retainer

Operate

Quarterly feedback loop, capacity reports, regression tests and the move to standards.

Slide routing →

Drawings marked illustrative show shape, not data. Epic and Beaker are trademarks of Epic Systems Corporation; DICOM is the registered trademark of NEMA. CarolinaEMR is independent and not affiliated with any vendor named.

Begin at Phase 00

Bring us one week of hourly histology output.

We will run the capacity balance, measure GB per slide on your scanners, place your lab on the maturity path and hand back the gap list with exit gates. The numbers become the plan.

Start the baselineOne case, four workflows