Digital Pathology Workflow Software: Build Custom, Buy Proscia or Sectra, or Keep the Viewer and Build the Routing Layer

When is off the shelf genuinely the right call here?

Proscia Concentriq, Sectra Digital Pathology and the image management platforms that ship with Leica Aperio and Philips scanners are real products, and some of them are excellent. Whole slide imaging as a diagnostic modality is established, and none of these vendors is failing at the imaging. Buying is frequently the correct answer.

Buy, and stop reading here, if this describes you:

• One site running one scanner brand, so a single ingest path and a single identity domain.
• Under roughly 200,000 slides a year.
• Research, education and tumour board use rather than primary diagnosis.
• An existing enterprise imaging estate from Sectra, where putting pathology inside the same archive is straightforward.
• Case assignment that a saved search over subspecialty and date genuinely handles.

At that scope a packaged product is cheaper, faster and better integrated than anything custom, and the operational irritation you feel is not yet worth a decade of ownership.

Two further cases where buying stays right at larger scale. If your laboratory information system integration has no political sponsor, do not start, because ingest without reconciliation produces an image library rather than a workflow. And if you have not completed the validation work required before whole slide images can carry primary diagnosis, do that first. Building against an undecided clinical scope is the most reliable way to double a budget.

When does a custom build actually pay off?

The category has an unusual property: the products are built around the image, and a pathology department runs on the case. An accession with blocks, stains, levels, recuts, a responsible pathologist, a turnaround clock and a report. Every practical failure sits at the join between those two models, and the join is where a build earns its money.

Build when two or more of these are true:

• You run scanners from more than one manufacturer, and clearing the unmatched image queue is somebody's daily manual task.
• Case assignment needs rules your platform cannot express, such as routing a prior specimen back to the pathologist who reported it, or capping daily complexity rather than case count.
• Pathologists sign out remotely and latency is quietly driving them back to glass for anything urgent.
• You route consultations between institutions regularly, and today that means mailing glass or issuing external accounts your security team dislikes.
• Algorithm results are entering the workflow and nobody has decided how they are recorded, who saw them and which model version produced them.

The mixed fleet signal is the most common and the least anticipated. Laboratories buy scanners at different times, so mixed fleets are the norm, and every manufacturer assumes its own end to end path. What arrives is an ingest reconciliation problem owned by nobody.

Our position on algorithms is deliberately conservative and it affects the build decision. Algorithm output belongs in the record as an advisory observation carrying its model version, the region it applied to and the time it ran, never presented as a finding. If you are heading that way, you want governance you control.

How do they compare on the things that matter in this industry?

Reconciliation, not ingest. A scanner produces a file. That file has to become slide 3 of block B of a specific accession, at a specific level, with a specific stain. Barcodes are read imperfectly, printed by different label printers and formatted differently depending on which module produced the slide. Ask any vendor or developer how they handle a barcode that decodes badly. If the answer is filename parsing, they have worked with one scanner and will not survive your second manufacturer.

Worklist versus rules engine. Packaged platforms offer configurable worklists, which is a queue with a saved search. What they generally cannot express is a departmental rule set: hold cases until a stain results, cap complexity units per person, escalate anything approaching the turnaround target, and know that one of the three people on the gastrointestinal list is covering frozen sections today.

Latency as a design requirement. A pathologist's tolerance is roughly the time it takes to move a slide on a stage, and a workflow slower than glass gets resisted, correctly. Vendor platforms leave this to infrastructure, which means your experience of digital pathology is largely decided by a storage team who never had a pathologist in the room. Ask what first view target a supplier designs to and how they hit it. You want to hear pre-fetch triggered by assignment and pyramid level caching at the edge.

Consultation identity. This is digital pathology's clearest clinical benefit and the workflow least well served by the products. The requirement is a scoped, time limited grant of access to one case for a named external pathologist, with their opinion returned as a structured record attached to the case rather than as an email.

Storage policy visibility. Ask for stored volume by case class and age. If that report does not exist, retention policy is emergent from whatever infrastructure could afford that year, and the reflex response to the annual surprise is to reduce scanning, which defeats the programme.

Annotation portability. An annotation set built over several years is a research asset, so confirm it exports as structured data rather than burned pixels.

What does total cost of ownership look like at your scale?

From Digital Heroes delivery experience, a focused first release runs $100,000 to $200,000 over 14 to 20 weeks, covering vendor neutral ingest with accession, block, stain and level reconciliation, slide quality checks, the subspecialty worklist engine and pre-fetch backed viewing. A full platform runs $280,000 to $700,000 phased over 9 to 15 months, adding consultation routing between institutions, annotation as structured regions, algorithm result governance, storage lifecycle tiering and laboratory information system write back.

Component by component: ingest and reconciliation $30,000 to $60,000, worklist and case assignment $25,000 to $50,000, viewing with pre-fetch $25,000 to $55,000, consultation routing $40,000 to $90,000, annotation and algorithm governance $50,000 to $120,000, storage lifecycle $35,000 to $80,000, and legacy archive migration $40,000 to $150,000. A second scanner manufacturer adds roughly $25,000 to $50,000 and about a month of calendar, most of it testing against real stained slides from your own histology laboratory rather than the sample files the vendor ships.

A three hospital academic system with two scanner makes, roughly 450,000 slides a year, primary diagnosis in four subspecialties and inbound regional consults priced out at $340,000, or $381,000 with a 12 percent contingency for the metadata problems that always surface in month three, delivered across about eleven months.

Annually, plan on support at 18 to 25 percent of build cost, scanner and format drift at $10,000 to $20,000, laboratory information system upgrades at $8,000 to $25,000 per major upgrade because every interface gets retested whether you want it or not, accreditation and revalidation at $10,000 to $20,000 per inspection cycle, and training at $4,000 to $10,000 for rotating residents and new pathologists. Then storage, which at high volume routinely exceeds the software maintenance line by itself, since a 40x whole slide image commonly lands between one and three gigabytes and retention horizons in pathology are long.

On the buy side, the comparison is harder than it should be because vendor quotes bundle scanners, service contracts, storage and software into one figure. Ask for the software and storage lines separately before you compare anything.

What does the hybrid look like, and when is it the honest answer?

For most departments this is the answer, and it is the one we give even though it shrinks the engagement. Keep the viewer. Build the layer above it.

Building a whole slide image viewer that matches an existing one for pan, zoom, pyramid handling, annotation and colour fidelity is roughly a year of specialised work, and it adds nothing your pathologists will notice except delay. A developer proposing to build one is either inexperienced or selling you time. Keeping the scanner vendor viewer commonly saves $30,000 to $50,000 outright and far more in schedule.

What you build instead is three pieces:

• Vendor neutral ingest and reconciliation, $30,000 to $60,000. Normalised scanner output, barcode decoding with fallback strategies, matching against the laboratory information system with a confidence assessment, and a short exceptions queue with the image visible so a human resolves it in seconds rather than opening three systems.
• The assignment rules engine, $25,000 to $50,000. Subspecialty, prior case history, workload measured in complexity, service coverage from the schedule and turnaround risk, with reassignment carrying a reason code.
• Pre-fetch backed viewing, $25,000 to $55,000. Assignment is the prediction signal: when a case is assigned its images move to warm storage and low magnification levels cache to the pathologist's edge.

Two further honest reductions. If your hospital infrastructure team already runs an object store paid for centrally, the storage lifecycle work shrinks to policy rather than procurement. And defer algorithms, because governance is the expensive half of computational pathology and there is no reason to build it before a model is actually in clinical use.

Which should you choose, by operator size and stage?

Find your row and act on it.

• One site, one scanner brand, research and education only. Buy the vendor image management product and stop. Put the money into scanners and histology staff.
• One site, one brand, moving to primary diagnosis under 200,000 slides. Still buy. Spend on validation, label printing consistency and barcode quality, which cost little and remove most of what people blame on software.
• 200,000 slides and up, or a second scanner manufacturer arriving. This is the decision point. Keep the viewer and build ingest reconciliation plus the assignment engine, roughly $55,000 to $110,000, and see how much of the daily irritation disappears before funding anything larger.
• Multi site with remote sign out and latency complaints. Add pre-fetch backed viewing to the above, roughly $80,000 to $165,000 in total, and set a first view target you will hold the supplier to. If the architecture cannot hit it, nothing else in the project matters.
• Academic system routing consults between institutions, with algorithms arriving. Build the full platform, phased: reconciliation and worklist first, then deeper laboratory information system integration, then consultation routing when a partner institution will test with you, then storage tiering once you have a year of growth data.

Two conditions apply. Sample a few thousand cases from your archive before fixing a migration price, because metadata repair is priced by how bad the metadata is. And gather validation evidence continuously as each subspecialty goes live.

When the shortlist is down to two and you need a tiebreaker, Digital Heroes contracts through India LLP, US LLC and UK LTD entities, so the agreement and the intellectual property assignment sit under law your own advisers already read. Nothing about that commits you to the build.