Beyond the Scanner: What “True” Digital Pathology Looks Like in Practice

Practical lessons from the South 4 Pathology network on building an end-to-end, LIMS-centred digital pathology service that stands up to NHS operational reality.

In the NHS, ‘digital pathology’ can sound like a procurement line item. In practice, it is a service transformation, one that succeeds only when technology, governance and operating models evolve together. Interfaces, networks, identity and access, storage, barcoding, QC gating, support processes, and governance all matter — as does how the service behaves at 09:00 on a Monday when the laboratory is under pressure.

Kieron White, Anatomic Pathology IT Manager, Oxford University NHS Foundation Trust discusses the Oxford approach to Digital Pathology and argues the case for making the pathways work, and not just making the technology run.

Kieron manages Anatomic Pathology IT at Oxford University Hospitals NHS Foundation Trust, has been central to its digital pathology rollout, and is currently installing the Clinisys^TM WinPath LIMS across the South 4 pathology network (covers Buckinghamshire Healthcare, Milton Keynes University Hospital, Oxford University Hospitals, and Great Western Hospitals NHS trusts). Clinisys^TM WinPath, once live, will be used for block and slide tracking and Digital Pathology reporting.

His role is the one most of us recognise immediately: joining up the IT. From LIMS to specimen processing and tracking, from scanners/IMS to authentication, networks, and the operational glue that keeps it reliable across Trust boundaries.

When people talk about digital pathology, they often mean different things. What’s your take?

For South 4, “true” digital pathology is not a digital slide viewer. It’s the coordinated use of digital technologies to support safer and efficient anatomic pathology pathways. It includes voice recognition, barcoding, instrument interfacing, resilient networks and storage, workflow visibility, and a reporting experience that reduces (not increases) cognitive load.

The easy mistake is treating scanners as “plug and play”. The hard work sits around them: Data Protection Impact Assessments (DPIA), firewall rules, identity/access, bandwidth, endpoint constraints, vendor boundaries, and storage that quickly moves from terabytes to petabytes. If those foundations are weak, the service will be brittle, even if the images look great.

What does the South 4 digital pathology strategy look like, and what drove the approach?

South 4’s strategy grew out of a practical need: share slides for MDTs, second opinions and cross-site working, without losing control of governance. One key design choice was local data ownership. Each Trust manages its own image data, paired with controlled sharing of images and access across the network.

Programme reality matters. Funding can arrive with spend deadlines, creating pressure to procure digital pathology solutions before integration and operating model decisions are ready. But integration is where programmes succeed or fail, shaped by networking and cybersecurity constraints and by the simple fact that whole‑slide images become an enterprise IT issue very quickly.

Scale shifts the discussion from “can we do this?” to “can we keep doing this?”. Oxford has scanned around 2.4 million slides and generated approximately 420,000 images every year; across a network, that becomes petabytes of data to run, secure and retain. At that scale, resilience is non‑negotiable: a single scanner failure can force a reversion to glass unless redundancy, support cover and planned maintenance windows are in place.

Is it really ‘digital’ if specimen management is still manual, and where should the LIMS sit?”

Different options on this subject abound, but Kieron’s view is clear: the LIMS is the “ground truth”. It’s where the case is managed, the reports are authorised, and where specimen state must be unambiguous. As digital pathology makes work more fluid, with recuts, restains, rescans and re‑work becoming routine, the need for a single source of truth increases, not decreases.

Here’s an operational safety example where an incorrect approach can have a major impact. If images are visible to a pathologist before laboratory QC is complete, and a slide is later rejected and recut, you can end up reporting from an image that should never have been used. In a digital pathway, “ready to view” has to be a controlled state, with explicit gating and auditability at a sample/block/slide level.

What happens when tracking and workflow logic sit outside the LIMS?

When workflow logic and tracking are split across a LIMS, an Image Management System (IMS) and multiple bolt-ons, you create duplicated truth. Someone updates status in one place (e.g., “reviewed” or “extra levels requested”), but it doesn’t land everywhere unless you’ve engineered, tested and maintained robust interfaces. Over time this creates reconciliation work, audit risk, and uncertainty about what is actually true.

It also complicates incident management. Faults can become ‘in‑the‑middle’ problems: scanner vendor versus LIMS vendor versus the integration layer. For NHS services already running at capacity, that delay is not just inconvenient; it has direct pathway impact. Support models and contracts need to be designed with this shared reality in mind.

What does a sustainable digital pathology platform look like?

For a sustainable digital platform, it is essential to reduce, and ultimately remove the reliance on glass slides. To do this, the digital solution has to work well for the histopathologists and be built with resilience in mind. With limited budgets, it is understandable to just have enough scanners, but at Oxford we took the decision to incorporate redundancy to ensure we can maintain the digital pathology pathways.

The quality of the digital image is also key. To avoid the tendency to revert back to the glass slide, histopathologists need to be provided with trusted high quality images.  

Standardisation, usability and clear ownership matter. Kieron stresses a “one label, one time” approach: consistent barcode identity flowing through staining, IHC, scanning and reporting is the difference between a controlled pathway and a workaround factory.

Looking beyond a single Trust or network, the value grows when images can be shared safely where expertise is scarce — renal pathology workforce scarcity being one example.

The NHS does not need more isolated islands of digitisation; it needs governed interoperability that supports cross-cover, second opinions and training without shipping glass.

AI is a major driver; how do we adopt it safely and meaningfully?

AI is often presented as a solution to structural pressures in the NHS, particularly in strategic narratives such as the NHS 10 Year Health Plan.

The risks are well‑understood — false positives and false negatives — but the near‑term value is practical rather than transformational: triage, highlighting areas of interest, prioritising slides (for example, prostate biopsies) and triggering downstream steps earlier, such as initiating IHC sooner in selected pathways.

AI is valuable today primarily as decision support. It is not the answer to everything — yet. We are not ready for AI‑generated reporting, and pathologists must always retain the final word in the definitive diagnosis or exclusion of cancer. Safe and accepted AI adoption depends as much on workflow design and governance as on model performance.

If you were advising another NHS Trust at the very start of their digital pathology journey, what would you encourage them to think about before investing?

It starts with talking to people, particularly those who have already implemented digital pathology or who have struggled through early decisions. There is now a broad base of lived experience across the NHS.

The same technology behaves very differently depending on local deployment and usage, staffing models, estates constraints, and whether digitisation is aimed at internal throughput, network sharing or primarily referrals.

Before procurement, make pathway decisions explicit: where digital pathology sits, what different flags/status changes mean, how exceptions are handled (recuts and restains), and how barcoding flow can be used to provide  an end to end pathway. Clarity here makes supplier conversations easier and materially safer.

A practical checklist for NHS Pathology IT & Anatomic Pathology Service Leads

• Define system of record: what system owns specimen state, and what is allowed to be “advisory” only?
• Provide a safe working environment: Make sure the users are fully trained and provided with the right environment e.g. up to date documentation, the right equipment, good quality images, LIMS with contextual launch between critical digital systems
• Identity & barcoding: apply “one label, one time” wherever possible; eliminate parallel identifiers across platforms.
• Integration & support: map every interface and agree who supports which failure modes, including cross-vendor faults.
• Capacity planning: plan for storage growth, retention policy, bandwidth, and workstation/display requirements for pathologists.
• Resilience: design for scanner redundancy, maintenance windows, and a clinically safe fall-back mode.
• Information governance: complete DPIA early, apply role-based access, and define clear rules for cross-site sharing.

Three takeaways for NHS teams implementing digital pathology

• Digital pathology is an operating model, not a device. The scanner is the start of the conversation, not the end.
• Make the LIMS the system of record for specimen state. Safety and governance depend on one place that represents “what’s true” right now.
• Design for scale and resilience from day one. Petabytes, vendor boundaries, maintenance, versatility, and staffing are not ‘later problems’, they are the programme.

Digital pathology pays off when it is treated as a governed service with clear ownership of specimen management, robust integration, and operational resilience. Get those right and scanners, IMS and AI become accelerators, not sources of new risk.

The views and opinions expressed in this article are those of the author, reflect personal experience, and are not necessarily representative of all users.