Editor’s note: Dirk Soenksen, founder and chief executive officer of the digital imaging company Aperio, spoke about “The Road to Digital Pathology” at the company’s 2006 Pathology Visions show. This is an edited transcript of his remarks.
Digital pathology is not a term that has been around for many years. Last year everyone was using “virtual microscopy.” Our space is evolving. We’re now using “digital pathology” because it more appropriately defines the space we’re in. Digital pathology goes well beyond slide scanning. This field started with a box—a slide scanner—that created virtual slides. But there are differences between the technical solution called virtual microscopy and this bigger space called digital pathology.
We think of digital pathology as a digital environment for managing and interpreting pathology information, specifically information that is enabled by the digital slide, which is a whole-slide image. Our definition is different than “Digital pathology is slide scanning.” In our view, it’s what you do with the information that matters.
Digital pathology can be thought of as an adjunct to traditional microscopy. In traditional microscopy you require a microscope to view the glass slide. You can view one slide, one field of view, one magnification at a time. If you had four eyes, you could view two fields of view, one slide, one magnification at a time. So there are limitations to traditional microscopy.
If you want to do any sort of analysis with a microscope, you have to remember the information from each field of view. For example, if you want to compare two slides, you have to look at one slide, store the information in your brain bank, look at the other slide, and try to remember the one you looked at before to make a comparison. This is what we’re used to in microscopy. There’s nothing new here.
On the digital pathology side, you have the benefit of doing things differently. For example, you can view multiple digital slides on a computer monitor. You can pair them side by side—something that’s so simple, yet it can’t be done easily on a microscope. If you want to count cells, if you want to quantify protein expression, these are easily done by a computer program that can be run on an image file called a digital slide.
With traditional microscopy, if you want to share the information with somebody else, you have to get the glass slide, the microscope, and the other person all in the same room. If you want to have somebody at a remote location view the microscope slide, you have to put the glass slide in the mail or you have to put the pathologist in a car or on an airplane. None of that’s true with digital pathology. You can view the information instantly; you can share the information with anyone in the world immediately.
It is also relatively easy to integrate a digital pathology system into a laboratory information system. In contrast, a microscope and glass slides are somewhat standalone from the LIS and represent only a small portion of all the other bits and pieces of information required to make an interpretation. Digital pathology can support the viewing and consolidation of disparate sources of information required for pathologists to do their work more efficiently and more productively.
A generic digital pathology system supports data acquisition from three types of devices: a slide scanner, a robotic microscope, and a digital camera. Any device that generates digital imagery information in pathology is part of a digital pathology system. Whether it’s a digital slide, whether it’s a live image in the form of video, or whether it’s just a still image captured from a camera on a microscope, it’s all part of digital pathology. That’s how the space has evolved. What’s important to remember is that digital pathology is much more than whole-slide imaging.
One could even make the argument that digital pathology encompasses images created by digitizing pictures obtained with film cameras attached to a microscope. In our view, digitized film is part of digital pathology. Our view of the world is that whatever the source of the image, it’s critically important to have a way to manage that information, whether it’s a still image, a video stream, a whole-slide image, or, at some time in the future, a three-dimensional whole-slide image. Ideally you want to manage these images with information management software.
In a digital pathology system you need to worry about multiple things in addition to viewing.
A couple years ago, you had a scanner that created a digital slide, and you had software to view the digital slide. This is what people called virtual microscopy. Digital pathology is much bigger than that. With digital pathology you have to worry about information security and creating an environment for managing data that comprises multi-gigabyte images. Also, you have to think about integrating the digital pathology information management system with autostainers, coverslippers, and the laboratory information management system. All this has to be worked out for digital pathology to become mainstream.
Broad adoption isn’t going to happen unless we can overcome some of these integration and information management challenges. For example, suppose you want to generate a report from a digital pathology system; if the report doesn’t already exist in the LIS, it has to be created by the digital pathology solution.
A digital pathology system must also be able to interface with other elements of the laboratory, for example, devices for image storage, ranging from arrays of discs to a PACS (picture archiving and communication system). PACS are popular in radiology. Companies like GE, Siemens, McKesson, Agfa, and Fuji are selling PAC systems. Believe it or not, it’s a $1 billion-plus market per year just on the software side.
Integrating a digital pathology system with a laboratory information system is important for adoption. You don’t want to have to switch between a digital pathology system and the LIS to get your work done. There has to be some convergence of these two systems. Then there’s integration with other types of lab equipment. For example, the concept of having a slide scanner next to a coverslipper that sits next to an autostainer is not new. This would be required to bring automation to the laboratory, something everybody believes will happen at some point.
Our view is that the pathologist’s role in digital pathology will be to operate the pathologist’s cockpit. This is an array of monitors loaded with information, so that, in principle, a pathologist sitting down to read a case doesn’t have to get out of his chair to get the work done. Everything is at his or her fingertips. It’s assimilating information from multiple sources. It’s using image analysis and computer data diagnosis tools to make more objective interpretations.
Consolidating information in the pathologist’s cockpit would allow pathologists to provide more comprehensive diagnoses to their customers: oncologists, surgeons, physicians. At the highest level, digital pathology has the potential to reinforce the central role of the pathologist in patient management.
How will this happen? I don’t think full adoption is going to happen tomorrow. I don’t think any laboratory is going to take all of its glass slides, digitize them, and operate in a 100 percent digital environment. When I sit in meetings and pathologists say, “My lab processes 1,000 slides a day, so your scanner would have to scan a slide in 10 seconds for me to consider adopting your technology,” I say, “You’re right. We’re not going to do that tomorrow.”
This may seem obvious, but this is not a trivial insight. You’re not going to go from an all-analog world to an all-digital one overnight. But when people talk about why a one-minute scanner is important, or why a 30-second scanner is important, it’s because everybody’s first reaction is, “I need almost infinitely fast scan speed to achieve full adoption.” Even if you had that 30-second scanner, you wouldn’t be able to achieve the goal of full adoption. So you might ask, “Why is Aperio having a conference where they’re telling us they don’t believe the market is ready to fully adopt digital pathology?” The answer is, adoption is going to happen in niches. You’re going to start using it for specific applications, and in those niches there is immediate value.
We’ve divided the use of digital pathology into three areas. First, you could digitize glass slides and use the digital slide after you’ve looked at the glass slide. That’s the simplest way. Let’s say you’re a professor. You take a glass slide and you digitize it and say, “I’m going to use that for education.” Or, “I have a slide, and I need to have somebody else look at it,” so you use the digital slide for a secondary consultation. Or you might say, “This is just an interesting case that I want to keep in my collection.”
Those are all uses that are valuable today. It’s the lowest-risk way to use digital pathology. You do everything you do now, and you have the added step of digitizing the slide. Then you manage the digital slide in some way, shape, or form. We think applications where the digital slide is used after the glass slide are the earliest applications, or niches, for adoption.
You can also think about using the digital slide before the glass slide. In this case, you’re going to do something with a digital slide before you look at the glass slide, but you’re always going to go back as a safety valve and look at the glass slide afterward. There’s a virtual IHC product on the market where pathologists send glass slides to a reference lab, the reference lab scans the slides and puts the resulting digital slides on a central server, and the pathologists review the digital slides before they look at the glass slide.
One benefit of looking at the digital slide before the glass slide is to evaluate the stain quality. They can say, “Based on this, I want to order additional tests.” If they want to, they can make a diagnostic decision. Or they say, “You know what? I’m going to make an interim decision that I’m going to confirm once I get the glass slide.” So there’s a series of applications where you would look at the digital slide before you look at the glass slide, but you would always look at the glass slide.
Then there are cases where you would use the digital slide instead of the glass slide. Nobody’s doing that today, and until somebody actually gets FDA approval, you’d be doing this at your own risk. These are cases where you would make a digital slide read; you would look at a digital slide and say, “Based on this, I can make an interpretation.” And you might never go back to the glass slide.
We think the adoption of digital slide reads is also going to happen in niches. You’re going to start with certain organ systems. Let’s suppose you’re in a hospital system, and you don’t have a dermatopathologist on site. So you might say, “I’m going to send all my derm path cases to somebody at a remote location for a digital read.” Another example is if somebody goes on vacation, and you might say, “For two weeks, we’re going to read all of that pathologist’s slides remotely, without looking at the glass. When the pathologist returns from vacation we’re going to revert back to reading the glass slides.”
Evolving from where we are today to full adoption—that is, looking at the digital slide only, never at the glass slide—is a path that doesn’t start in the upper right-hand corner. An evolution has to happen. You have to get comfortable. Think about what we all went through when we adopted e-mail from fax. Initially we said, “I’m going to use it for the less important documents, but for the most important documents I’m still going to use the fax machine.” That’s certainly how I was, until I had enough confidence in the technology to say, “E-mail is the primary way I’m going to communicate.”
We have to go through that sort of adoption with digital pathology as well. So when you hear somebody talking about full adoption and the need for a one-minute scanner, you should challenge them and say, “Why do you think you’re going to do this tomorrow?”
It’s interesting to think about digital pathology applications in a 3D space because there are three elements that are important. In fact, every digital pathology application is a linear combination of these three core capabilities. Along one axis, you have analytic techniques, which is essentially image analysis. On another axis, you have remote viewing. The third axis is data management.
For example, IHC image analysis doesn’t require any remote viewing. If you add remote viewing to image analysis, you get to the virtual IHC product, where you’re doing remote image analysis. The archive and retrievel application falls along the data management axis only, because if you’re doing archive and retrieval in your local environment, you don’t care about image analysis or remote viewing.
We think of some digital pathology applications as comprising three elements: how much remote viewing, how much analytic technique, and how much data management?
Here are some of the reasons why full-scale adoption is in the future: We’re not integrated with LISs yet; we don’t have FDA clearance; we lack published success stories; and we’re still challenged by IT integration. Anytime we talk to a prospect, we’re asked, “How do we get digital pathology integrated into our IT environment?” Another challenge is workflow integration, which is only now beginning. Also, there are no standards. We’re all working on a DICOM standard to store images in PACS, but today there is no digital slide standard. Finally, there haven’t been any studies to measure the return on investment for deploying this technology.
So rather than aim for full adoption today, we can work in niches like education secondary consults, archive and retrieval, virtual IHC, and IHC image analysis. There is significant value to using digital pathology for these niches. We have customers who have adopted digital pathology for many of the niches I talked about. There are, however, challenges that we have to overcome as a community to start using this technology for digital slide reads. I think it’s some years away. It’s just not reasonable to think you’re going to go from zero to 100 miles per hour overnight.
With this framework for adoption our strategy is to take some of these niches and make them bigger. Let’s add more niches, and let’s get everybody more comfortable with the technology. Let’s focus on publishing more clinical studies. Some companies will take the path of moving toward FDA adoption. I think next year we will have LIS vendors working in the solutions gallery. Some of those efforts have already started, and we’re well on our way.
What’s the market today? It’s emerging and generally recognized as being a large opportunity. Our estimate is that the install base is growing at 50 percent annually, based on Aperio’s numbers. There is growing awareness that software is what delivers value. That doesn’t mean scanners are less important. It’s great if you have a cool scanner—but what are you going to do with that information?
There is low awareness of digital pathology among pathologists. The average pathologist in a community hospital does not know anything about digital pathology. Or, if he does, he has the wrong idea about it. Important critical path applications have emerged. One is in reference labs, which are using this technology to gain competitive advantage by providing better service, better turnaround time.
In the marketplace, there’s more competition. More companies are entering the market, and we’re seeing more consolidation. Olympus has purchased Bacus Laboratories. Clarient has purchased Trestle Holdings. Some of the early players are being acquired. Those are all good signs for the industry.
The focus today is still primarily on the hardware: How cool is my scanner? Today on the scanner side we have the ability to do 20x, 40x and live (robotic), primarily in histology. And we now have a two-minute scanner.
Over time, you’re going to see these scanners used for other specimen types, whether it’s hematology, which requires high-resolution oil scanning; cytology, where you have to do three-dimensional scanning, faster than two minutes; and fluorescence and multispectral. Those are the areas where you’ll see the hardware capabilities evolving.
With data management capabilities we can view, conference, and analyze digital slides. In the future, you’ll see integration with laboratory information systems and the use of data management tools to do intelligent retrieval. The goal is to query databases to bring out information that’s more than ”Hey, show me the slides I scanned yesterday.”
Digital pathology technology can help automate certain tedious tasks and can be integrated into workflow. You’re also going to see integration with PAC systems.
In terms of image analysis, today it’s focused mostly on immunohistochemistry. Today, you can take a digital slide, and you can apply algorithms against that digital slide, and you can interface to third-party software packages. In the future, you’re going to see a proliferation of additional algorithms for applications other than IHC. You’re going to see some of those algorithms evolving in computer-aided detection, or CAD, very similar to what’s happened in radiology, where you might have the ability to take a set of slides and to put them into two piles, normal versus abnormal, using a machine.
In terms of compliance, there are still some digital pathology vendors with proprietary file formats. You’re going to see more clinical studies. You’ll see the acceptance of a DICOM medical imaging standard. And I think you’ll see FDA clearance, beginning with IHC image analysis, very soon. In fact, we plan to have a submission by Aperio in early 2007. That will set the stage for obtaining clearance for digital slide reads. There will be the beginning of more formal studies aimed at doing digital slide reads as well.
There are lots of opportunities for automation in the laboratory. And image analysis will continue to grow. It will become an important part of what we do as we move toward the everyday use of this technology.