Header: Should digital pathology have a true revolution in the quality of input devices?
###
Recently, we blogged about how digital pathology advances slowly. And rather than looking to radiology for advice, we should look to...the adoption of Da Vinci robotic surgery.
Robotic surgery probably took half the prostatectomy market in ten years - at huge cost in equipment and infrastructure. There were four potential drivers:
1) Health outcomes
2) Economics (faster-cheaper)
3) Surgeon enthusiasm
4) Hospital competition
Per that blog and Da Vinci white paper, there was little early evidence for better outcomes. The costs were high, not self-propelling. However, #3 surgeon enthusiasm was high, and #4 hospital competition (keeping up) was a big fact.
1) Health outcomes
2) Economics (faster-cheaper)
3) Surgeon enthusiasm
4) Hospital competition
For digital pathology, big health outcomes aren't expected, it's more expensive, some pathologists like digital but it's hardly a stampede of excitement, and hospital competition (e.g. billboards or radio ads) is off the table.
One thing we might be able to change is #3, pathologist enthusiasm. I think it's pending two things: AI prepares a full digital report ready for careful validation and sign out (it's not so far off) and second, there may be a missing "quantu leap" in user interactions.
On the latter count, see a new article by Lopes et al, new horizons for input devices. It got a write-up in CAP TODAY as well.
Here, let's try to take it further with a 14-page white paper from Claude.
Whole-slide imaging has spent twenty years solving the
hard problems of optics, file formats, scanner throughput, FDA clearance, and
now artificial intelligence. It has spent almost no effort on the question that
sits between the pathologist and all of that machinery: how, physically, a
human being should steer a gigapixel image with their body. The de facto answer
in 2026 is mostly the same as in 2009, which is a computer mouse, a scroll wheel, and
a toolbar of small icons designed for spreadsheets. The small literature that
exists on alternatives, which runs to perhaps a dozen papers across fifteen
years, is consistent in finding that pathologists would rather use almost
anything else, and consistent in being underpowered, task-contrived, and
politely ignored by the companies that build viewers.
The most recent and most interesting entry, a 2026 Archives study from Western University testing eye and head tracking, is good work that nonetheless illustrates the field’s smallness of ambition, right down to its survey question asking whether pathologists would spend a hundred dollars a year of their own money on a better way to look at cancer.
This paper argues
that the interaction problem is real, that it has consequences for diagnostic
accuracy and not just for wrists, that incrementalism is structurally
guaranteed by the way the field is organized, and that a deliberately radical,
well-funded, year-long design effort by a small mixed team is the appropriate
response. The goal should be a different way of working, not a better mouse.
