What Radiology Figured Out… And Pathology is Still Learning

There's a transformation that already happened in radiology, playing out now in slow motion in pathology. If you work in pathology and you're watching digital imaging platforms proliferate around you, you’re probably wondering how the reporting workflow fits in, whether your current documentation tools will keep up, and what the right architecture looks like. The answer is already visible. You just have to look one department over.

How Radiology Changed

Twenty years ago, radiology reporting looked a lot like pathology reporting looks today. Radiologists dictated. Transcriptionists typed. Reports moved through a workflow that had been essentially unchanged for decades.

Then PACS (picture archiving and communication systems) arrived. PACS digitized image storage and retrieval, and the reporting workflow changed… not because anyone planned it that way, but because of a simple gravity: when the images moved into PACS, the natural place to author the report was inside that same environment. Report authorship migrated into the diagnostic workflow because that's where the work was happening.

PowerScribe became the dominant radiology reporting platform not because it had the best speech recognition engine. It won because it was built to be driven by the PACS workflow. It was embedded deeply enough that reporting happened naturally within the same environment where radiologists were already working, rather than as a separate step outside of it. The lesson radiology learned, eventually and at scale, was that the diagnostic workflow and the documentation workflow have to be the same workflow — or the documentation workflow loses. 

Pathology: On the Same Path

Pathology is following the same arc. Whole slide imaging platforms like Proscia, Philips IntelliSite, Hamamatsu NanoZoomer, Leica Aperio, and more, are moving from research and subspecialty use into primary diagnosis at major medical centers and health systems. AI analysis tools are layering on top, generating quantitative findings, flagging cases for review, and surfacing data that wasn't available in the analog workflow.

The question this creates for pathology departments isn't whether to adopt digital pathology. It's how the reporting workflow connects to it.

Right now, for most departments, the answer is: it doesn't, or not well. A pathologist reviewing slides on a digital platform still has to context-switch to their AP system or a separate reporting tool to document findings. The image is in one place. The report is in another. The workflow is fractured, and every fracture costs time and creates opportunities for error.

This is the same problem radiology solved. Pathology hasn't solved it yet.

The Orchestration Problem

The reason pathology is slower to converge than radiology isn't a technology problem. It's an architecture problem.

In radiology, PACS was a single platform managing both imaging and workflow, and PowerScribe built directly into it. In pathology, the landscape is more complex: the AP system (such as Epic Beaker, CoPath, or PowerPath) is the system of record; the digital pathology platform is a separate environment; AI tools are separate still; and the EHR sits above all of it. There is no single platform that owns everything, and there probably shouldn't be.

What's needed instead is an orchestration layer: a platform that brokers data and workflow across these systems, presenting pathologists with everything they need at the moment of diagnosis, and writing structured results back to the appropriate systems of record. It's not a replacement for Epic Beaker or Proscia or PathAI… but a connective layer that makes them all work together seamlessly.

What That Looks Like in Practice

Voicebrook's pathology reporting solutions are designed to be that orchestration layer. The architecture has three components working together:

The AP system remains the system of record. Orders flow in. Finalized reports flow back. Nothing changes about how the AP system manages case identity, accession series, or report storage.

The digital pathology platform handles image management, AI analysis, and slide-level workflow. Voicebrook can integrate with any digital pathology vendor, so that data and workflow signals can cross between environments.

Voicebrook's VoiceOver PRO sits in the middle: pulling case context from the AP system, surfacing relevant data at the moment of diagnosis, capturing structured findings, and writing finalized reports back to the LIS as the record of what was found. And it does this without requiring a custom backend integration for each vendor — meaning departments aren't trading one complexity for another.

With Voicebrook, the pathologist doesn't have to navigate between systems to complete a case. The workflow comes to them.

The AI Layer Coming Next

The radiology parallel extends one step further. After PACS-integrated reporting matured, AI analysis tools began integrating directly into the radiology workflow: flagging findings, pre-populating structured fields, triaging worklists. That integration happened smoothly in radiology because the reporting workflow was already centralized.

In pathology, the same AI integration is coming. It will land much more cleanly in departments where the reporting workflow is already serving as an orchestration layer — where there's a single environment that can receive AI findings, present them to the pathologist at the right moment, and capture the pathologist's response in a structured, auditable way.

Departments that are still running speech engines as standalone tools, disconnected from both their AP system and their digital pathology environment, will have to rebuild their workflow from scratch to accommodate the AI layer. Departments that have already invested in an orchestrated reporting workflow will extend it.

Radiology figured this out over a decade of sometimes challenging transition. Pathology has the advantage of doing it with foreknowledge of where it ends up.

Is your lab ready for what comes next? 

Let us show you the Voicebrook difference...