Digital Pathology Part 2: Ditching the Scopes
Pathologists don’t like change. If you’ve ever met one of our kind, you know we are creatures of habit who’d rather be left alone in the dark corner of the lab clutching our microscopes. In this way, we’re like many other professions populated by introverts and nerds, only more so. Despite our inclinations toward constancy, technological advancements in the past decade have forced us to to radically change how we work. We’ve ditched our microscopes and replaced them with dual monitors and keyboards. I realize most of you probably don’t care whether we are looking at microscopes or monitors, and that this feels like a niche irrelevant topic. As we continue this series, I want to convince you that the conversion to digital pathology is about more than just enhanced efficiency and less eye strain for a select few. It’s about the discovery of new and valuable diagnostic information being illuminated by AI. It’s about a source of untapped data found in traditional and seemingly archaic images. It’s a part of the bigger -omics revolution changing the way we think about diagnostic testing and will be a valuable part of future multimodal machine learning models. But first, we need to set the stage to understand where we are with the technology and how we got here. Here’s some of the story from someone who was in the trenches during this transformative time.
The microscope at the center
Since Rudolf Virchow and the dawn of pathology over 150 years ago, pathologists have been peering into microscopes. Like Virchow’s pupils, I trained on a microscope. Virtually all human and veterinary medical students today still train on microscopes. During my pathology residency, much of my world was centered around this tool, which was core to our trade. For our annual migration to the national board exam in Ames, we opted to drive across the country instead of flying, ensuring our microscopes’ safety by carefully buckling seatbelts around them in the backseat, treating them like our cherished mechanical offspring. We built our offices and desks around the microscope. Kept them clean and maintained them. I’m sure our retinas still carry residual effects of the twin beams of light traversing our globes for hours on end.
The Death of the Microscope. All Eyes on Radiology
My profession has changed much since those days. I haven’t used a microscope for diagnostic purposes more than a handful of times in the past decade. If you look at the office of a commercial veterinary diagnostic pathologist, you won’t find a microscope. Or, if you do, it’s a decoration alongside raccoon baculums and potted succulents. The digital pathology revolution has transformed how we work and is doing the same in human pathology.
The transition to digital pathology began with the emergence of digital cameras. In the 80s and 90s, pathologists started mounting digital cameras to the top of their scopes, allowing one to take static images of things seen microscopically. With the relatively new communication platform known as “email”, you could then send those images to colleagues around the world for consultation. This was helpful and a big step, but at the same time everyone was watching the rapid transformation happening in radiology as the traditional film processing was being replaced with a rapid digital workflow. While not intuitive at the time, technology to digitize radiographs and microscope slides makes radiology and pathology remarkably similar, both workflows composed of:
- A digital image uploaded for diagnostic evaluation.
- A professional diagnostician reviewing the image and writing a descriptive report with findings and recommendations.
- A clinician proceeding clinically based on findings in the diagnostic report.
Radiology is a harbinger for what is to come in pathology, and it’s been that way since radiology went digital. Right now, we watch as radiology is in the midst of the AI revolution and grappling with the effects, just as pathology is starting down the same path. We’ll revisit this topic later, as it is worth a deep dive.
While digital radiology was exploding, a key component to the digital pathology workflow was introduced: the digital whole slide scanner. Digital whole slide scanners came online in the late 1990s and early 2000s, allowing for the digitization of glass slides. But guess what was a big and early barrier? Anybody remember internet speeds in 1999? Those 2 GB images weren’t going anywhere fast. Navigating them was reminiscent of playing solitaire while waiting for your next Google Earth frame to refresh on dial-up speeds of 14.4 kbps. Existing technological hurdles of the day meant that digital pathology was really more of a research tool with limited practical applications. As technology advanced and these barriers were removed, exploration of the commercial opportunities began in earnest, starting with the large commercial laboratories in the US.
Commercial adoption of Digital Pathology
As often happens in veterinary medicine, the lack of regulatory oversight facilitated quick adoption as compared to human medicine, with Antech and Idexx converting their pathologists to digital in 2014 and 2015 over the course of a few months. I joined Antech in 2012 just as the massive project for the conversion to digital pathology was being launched. We had a million questions.
- Would the images be good enough for diagnostic purposes?
- What kinds of lesions, if any, would still need glass slide review?
- Can we scan slides quickly enough and distribute them in time to make it worth replacing the existing model?
- Will our pathologists rebel? Will they hate life? Will they make our lives miserable?
Dr. Jeff Edwards is a long-time friend and fellow pathologist who I’ve dubbed the “Father of Veterinary Digital Pathology”. He served as Antech’s Chief of Anatomic Pathology at that time and led the project to overhaul the pathology workflow from analog to digital. It was no small feat convincing the staff of 30+ anatomic pathologists to ditch their scopes and embrace dual monitors and a mouse for daily work. To validate the technology, we recruited a team of pathologists from Antech including: Dr. Phil Labelle, Dr. Cindy Bacmeister, Dr. Rich Moreland, Dr. Fabiano Oliveira, Dr. Jeff Edwards, and myself (I’m likely forgetting some others, I’m sorry!).
Dr. Jeff Edwards
What did our validation study find? We discovered that some things were easier with digital pathology. Low power diagnoses were faster due to the ability to zoom out and visualize the entire image on the screen. Measuring margins with digital rulers was a snap and much more accurate than the old method of holding the microscope slide up to a light and squinting at a ruler held against it. High power diagnoses were a little slower, and some structures like cytoplasmic granules and small bacteria weren’t as crisp as observed by light microscopy. But to our pleasant surprise, the images were more than adequate for confident diagnoses and results were comparable to light microscopy.
Once validated, we broke the news to the IT team, dumped a giant project in their lap and forever changed their world. We installed 30 Leica scanners in 2 lab locations. Every day, thousands of slides were scanned and distributed to pathologists across the globe. Turnaround times improved, and the oft-experienced weather-related shipping delays became non-existent. As one might imagine, this required a large overhaul of Antech’s IT infrastructure. Omar Ondoy’s IT team has long been a jewel in the Antech crown and did incredible work. Superstars like Patricia McDevitt and Daniel Garcia (again, many others I’m forgetting!) built the system to be pathologist-proof! Much to our surprise, the pathology staff took to the digital images quickly. We realized the advantages that digital brought, not to mention the ergonomic benefits of not having to hunch over a scope all day. Pathologists had the opportunity to get glass slides to compare to a tough digital image, but after 6 months nobody was asking for them anymore.
One of the most appreciated benefits of digital pathology that continues today is the ability to rapidly consult with pathologists around the world, opening up an amazing opportunity to have experts chime in on tough lesions. You see, even though we pathologists are trained on all organ systems, the reality is many of us have subspecialities or favorite organ systems. For example, I’ve been embedded in dermpath and oncopath for many years and feel that I excel at tough skin and cancer cases, but I’m just okay at liver pathology. In the light microscopy days, if I had a tough liver case I’d have to take my best stab at it and hope I did it justice. Or if the client was able to wait a couple of weeks I might pack the slides up and mail them to a hepatopathologist across the world somewhere. But with digital pathology I was suddenly able to click a button and have an expert hepatopathologist look at my slide and give feedback within an hour. Not only did this collaborative ability increase the accuracy on individual cases, but over time it grew the knowledgebase of the entire pathology team, breaking down the geographic barriers that once kept us in silos and made us prone to the type of bias that comes with working in isolation. A 30+ pathologist team scattered across the globe was instantly connected in a way that allowed for better collaboration than if we’d been in the same building. In fact, over time we found ourselves sending digital consultations to other pathologists even when they were working in the office next door, saving everyone the hassle of having to get out of a chair and walk 10 steps.
While we were patting ourselves on the back for a job well done, we hadn’t actually told the world anything about our amazing feat yet. Before our marketing efforts to do so got off the ground, Idexx smartly put out a press release about their own upcoming conversion to digital pathology that hadn’t actually happened yet. Despite our hard work, Idexx had stolen our thunder and received the attention we were hoping for. It may sound petty now, but we were proud of being the first veterinary commercial lab to move to digital pathology!
In subsequent years, a handful of smaller veterinary labs and academic centers followed suit. Today, however, many labs still use traditional light microscopy. The early argument by these labs was that the cost of conversion just wasn’t worth it. Their pathologists were all on site, and digitization just added a new step. While that argument may have been valid in the early days, I fear that these small labs and institutions will be left behind when the upcoming wave of image analysis AI models hits our profession full force. To be ready for what’s coming, digital pathology training is essential and should be embraced as quickly as possible.
A room full of digital slide scanners
Cytology is just more of the same, right?
At this point, we do need to make a distinction in what we mean by “Digital Pathology”. Most of our discussion thus far has been around digital histology slides. The term “Digital Pathology”, however, is an umbrella term that refers to digitizing both histology and cytology slides. Digital cytology presents special challenges, including larger file sizes, z axis scanning needs, and increased scanning times that made early adoption slower than histology. Scanning technology wasn’t quite there yet and bandwidth issues were a barrier to ocvercome. Despite these challenges, an innovative startup, Lacuna Diagnostics, emerged to rise to the task. Led by a bright young team of entrepreneurs, Lacuna launched in 2016 with point-of-care cytology scanners in hospitals. Although the story is forgotten by many today, Lacuna prepped the world for digital cytology. In doing so, they paved the way for the large labs to later convert to digital cytology in 2020. The digital cytology story alone is fascinating and worthy of its own dedicated letter. We’ll try and tackle that later and hopefully snag one or more of my friends from the Lacuna team to weigh in. For now, it’s worth at least highlighting the fact that digital cytology slide scanning from point-of-care units in hospitals is the closest thing we have to replicating digital radiology in diagnostics, with potential for synergy and cross-collaboration that nobody is paying attention to.
Digital microscopy in human diagnostics
So, what about digital pathology in human diagnostics? This story is better for someone with more human medical regulatory expertise to tell, but I’ll summarize. The regulatory hurdles required for FDA approval and clinical use resulted in a slower rollout than we’ve seen in veterinary medicine. It was actually quite comical in the early days. Dr. Edwards and I attended a human digital pathology conference thinking that we’d learn something that would improve our system. To our surprise and humor, the speakers were talking about their success stories of digitizing small labs with handfuls of pathologists and “large numbers” of slides being scanned that were in actuality only a few hundred per day, meanwhile we were scanning 5-10k slides per day. We quickly realized that our conversion was much larger than any attempted in human pathology to date and that veterinary medicine was much farther along in the digital pathology conversion process. The good news is that today digital pathology in human medicine is gaining traction. It’s becoming more commonplace, with several FDA-approved systems for primary diagnosis.
A few notable events in regulatory approval for human diagnostic use:
- Approval of Philips’ Intellisite pathology solutions in 2017
- Leica Aperio in 2019
- Hamamatsu Nanozoomer S360MD in 2022
- Paige Prostate AI in 2021
- Proscia’s Concentriq AP-DX in Feb. 2024
The Future
Why does any of this matter? Who cares how the pathologist views the slide as long as they write a good report? Who cares whether the lab saves money on shipping costs, or is slightly more efficient? As we will demonstrate in subsequent newsletters, the digital pathology revolution may have started out as a way to enhance efficiency, but has set the stage for artificial intelligence and new categories of diagnostic development. It’s paved the way for unrealized and unexpected data and solutions to come out of seemingly archaic pathology images. The AI pathology revolution is already here; its potential is just not realized. Let’s keep exploring this topic together. In the next segment in this series, we’ll dive into AI in pathology and where all this is going.
