Ask most people what it takes to “go digital” in pathology and they'll picture a scanner: glass slides in one end, high-resolution images out the other. It's a reasonable guess. It's also the part that turns out to matter least.
On episode six of Flight Path, Dr. Liron Pantanowitz, Chair of Pathology at Pitt and UPMC and one of the most influential voices in digital and computational pathology, described his own institution's path in a way that should sound familiar to anyone who has modernized imaging: “Early on we were all about the hardware, which scanners did we need. Then we realized it's more about the software, and now we realize it's really about the applications.” The scanner was table stakes. The hard part was everything around it: the standards, the workflow, and getting an enterprise to talk to itself. If that sounds like radiology's story from twenty years ago, that's the point.
Here's the plain version. Pathology has historically lived on glass. A tissue sample becomes a slide, and a pathologist reads it under a microscope. Digital pathology scans that slide into a whole-slide image, a very large digital file, so it can be viewed, shared, measured, and analyzed on a screen instead of at a bench.
The reason a CIO or imaging leader should care isn't the screen. It's what the screen makes possible: a biopsy taken at a small hospital can be read by a subspecialist a hundred miles away; a patient's slides can sit alongside their CT and their cardiology images in one record; and, eventually, algorithms can help triage and support the read. UPMC is now, in Pantanowitz's words, a 40-plus-hospital integrated healthcare system, and his goal for it is disarmingly simple to say and hard to engineer: a patient who walks into any hospital in the network should get the same level of care. That only works if the images can move.
Digital pathology breaks the tidy math that justifies a new CT or MRI. With a scanner, you can project studies and the revenue you'll bill against them. Pathology doesn't reimburse that way. The billing codes for digitizing work, and for AI, still lag. So the return shows up indirectly: efficiency gains, workload you can distribute instead of duplicating, lower courier and slide-retrieval costs, and reduced medical-legal risk. “It's difficult because the return on investment is hard to prove right now,” Pantanowitz said, not because it isn't there, but because the field is still short on long-term outcomes studies that quantify it.
The lesson for leaders isn't “wait for better codes.” It's to build the business case on the honest levers: access, efficiency, risk, and the data foundation you're creating, and to say so plainly to the people funding it.
This is where a pathology decision becomes an enterprise decision. Should pathology contribute its images to a shared, vendor-neutral record, the same place radiology and cardiology already live? Pantanowitz makes the case both ways, and the tension is worth sitting with.
The upside is substantial. When the enterprise funds a common imaging foundation, pathology gets implemented as part of the whole rather than fighting for a departmental silo's budget. Patients benefit because a surgeon, a dermatologist, or a nephrologist can actually see the images with them. And when multimodal AI arrives in earnest, pathology images already sit in the same data foundation as everyone else's.
The tensions are just as real, and he names them: a non-pathologist viewing a slide may misread what a trained eye would catch; proprietary image formats mean there may be no universal viewer that just opens a slide on a clinician's desktop; and some pathologists are, understandably, protective of images they worry will be used without them. None of these is a reason to stay siloed. They're the design requirements for doing it right, and every one of them is an interoperability problem, not a scanning problem.
For any organization dipping a toe in, Pantanowitz's advice is refreshingly unglamorous. Skip the big bang. Some people want to go fully digital in three weeks and take away everyone's microscopes. “I haven't found that successful,” he said. Go stepwise: start with education, then image sharing (telepathology), then primary diagnosis on digital, then quantitative image analysis, then the more advanced algorithms. And go bottom-up: when pathologists are the champions, adoption sticks; when it's mandated top-down onto people who were never trained for it, you pay for it in retention and recruitment.
Then the reality check: this is a journey with no finish line. “I don't think there's a day you sit back and say ‘we did it, we're fully digital,’ because it still requires upgrades, new applications, troubleshooting, and IT has its downtimes,” he said. He expects a hybrid world where some cases still go back to the microscope. He borrows a line often attributed to Bill Gates, that computerizing a process exposes its inefficiencies, and says digital pathology is a lot like that.
The most useful thing Pantanowitz does with AI is refuse to treat it as one thing. He sorts it into functions a leader can actually plan around:
The caution he keeps returning to is the foundation underneath all of it: data. Commercial algorithms are usually trained on populations other than your own, so generalizability is a genuine question. His own team used an algorithm that had not been trained on the very advanced, high-grade cancers they see. Commercializing archived images is one option, and he favors labs partnering with industry, but only with privacy, ethics, consent, and disclosure handled first. “That train has left the station,” he said of digital and AI adoption. What he asks of the field next is evidence, shared standards, and far better AI literacy.
If you're weighing where pathology fits in your enterprise imaging strategy, a quick self-check:
If the answer to any of those is “not yet,” the gap isn't a pathology gap. It's the plumbing, and it's the same plumbing that decides whether your AI roadmap pays off across every specialty, not just this one.
This is the terrain Mach7 Technologies has worked in for years: helping healthcare organizations take control of their imaging data and make it accessible, portable, and ready for what's next. The Mach7 family spans enterprise imaging, the vendor-neutral archive (VNA), the communication workflow engine, the eUnity™ enterprise viewer, and AI enablement, with Flamingo™ as the next offering.
That is why the enterprise question in this episode matters to us: the goal is a foundation where a pathology image can live alongside the rest of a patient's record, without locking anyone into a single vendor. We didn't build it to win a pathology argument. We built it so that when the enterprise decides pathology belongs at the table, the table is already there. From Pixel to Point of Care.
Hear the full conversation with Dr. Liron Pantanowitz on Flight Path, available on YouTube, Spotify, and Apple Podcasts.