Warning! You won't be able to use the quotation basket until you enable cookies in your Web browser.
Warning! Your Web browser is no longer supported. Please upgrade to a modern browser.

Going Digital: The"Why's" and "How's" of Adopting Digital Pathology

Overview

In recent years, digital pathology has gone from an intriguing idea to an integral part of how academic and commercial labs operate. Join Leica Biosystems and Procia Digital Pathology as they discuss why institutions are going digital today and how they approach best practices in digital implementation. Hear from an adopter of digital pathology about the promise they saw in this technology, how easy it was to implement, and how their work has changed because of adopting it. Learn the exciting potential that digital and computational medicine hold for pathology and discover why now is the time to bring digital pathology into your organization.

Learning Objectives:

  1. Learn from industry and lab representatives about the value of adopting digital pathology
  2. Discover how easy implementation can be by viewing example applications, best practices and important considerations for transitioning to a digital workflow.
  3. Understand the recent and future changes that digital pathology will bring to the laboratory space.

Presenter:

Dr. Alexander Baras, Associate Director of Pathology Informatics, Johns Hopkins Medicine

Dr. Olga Colgan, Director of Marketing, Pathology Imaging EMEA, Leica Biosystems Pathology Imaging Business

Nathan Buchbinder, VP of Operations, Proscia, Inc.

Webinar Transcription

Evolution of Digital Pathology - (Olga Cogan, Ph.D)

It was around the 1600s that we saw the emergence of microscopy and over the next 300 years we saw huge developments in the imaging but staying predominantly with the same modality of a sample, a light source, and magnification where your microscope, your sample, and your pathologist all have to be in the same place at the same time.

Around the late 1900s we started seeing the emergence of photomicrograph use giving some level of sharing to images, and in the 1980s seeing the emergence of telepathology; being able to remotely review slides in a live view-type scenario. It wasn’t until the turn of the century that we started seeing the emergence of whole slide imaging and the birth of digital pathology. In the last 18 years we’ve seen a technological maturity around digital pathology in the digital pathology-specific solutions with the hardware and the software and also in the infrastructure and the technologies that underpin a digital pathology deployment; things like the processor speeds and servers to be used, the storage costs associated with housing whole slide images and the performance of broadband and increased connectivity and availability. This leads us to a situation where digital pathology today incorporates all these elements of scanning, management, sharing, and interpretation of pathology information in a digital environment.

Market Trends in Pathology

When we look at some of the trends we’re seeing in pathology, we see a decline in practicing pathologists which is predicted to worsen in coming years as many pathologists are reaching retirement age and the positions are not being backfilled at the rate needed globally to cover the pathologist requirement. It is forecasted that there will be a net deficit of almost 6,000 pathologists worldwide by 2030. If we add the increasing and ongoing demands on pathologists and pathology departments with the emergence of new subspecialties and the emergence of new biomarkers and new detection methods, we’re faced with a situation where we’re being asked to do a lot more with a lot less and we need to look at innovative solutions to support that. For many pathology and anatomical pathology departments still using glass slides and paper files it can be a huge challenge when faced with shipping slides between different experts. The innate fragile nature of glass slides doesn’t lend itself to packaging and shipping both from the safety of the sample perspective, but also the time that it will take to ship those slides around.

Leica Biosystems Overarching Strategy

Leica Biosystems looks at the whole pathology workflow taking a holistic systemwide view of pathology from the excision of a biopsy piece of tissue right through to reading out that slide and looks at a fully integrated approach on how to enable control of the whole lab through integrated IT solutions and managing the different steps throughout the process to ensure a standardization and quality, ultimately delivering a perfect glass slide which can then be digitized and reviewed.

There are a lot of pressing economic and social factors at play for people when they’re choosing to go digital; it is about the use case and the problem you’re looking to solve. Digital pathology can support this by being able to implement a fully digital solution where you can open up your laboratory information system and review slides, review your H&E, review your IHC slides from your LIS interface, and eliminate the need for glass handling at that point. It will also leverage rapid peer-to-peer collaboration to receive input from colleagues when reviewing slides without having to physically shift the glass around between sites and people. Digital pathology can remove the necessity for the slide, microscope, and pathologist to be in one place at the same time. By leveraging digital pathology, you have anytime, anywhere access to the slides by the right person, at the right time, and the ability to externally expand your field of experts making this as easy to use as possible.

Steps to Going Digital in Pathology

There are a number of factors or steps when going digital in pathology and it can be quite transformational and impact multiple departments within an organization. The first big step is finding your champion for digital pathology. Who is that person who really knows what they want to achieve and can help drive that message, get buy-in throughout the department, define the needs and the goals that you’re looking to cover, and set those out as quantifiable steps? It’s not a one size fits all approach and you need to specify the type of samples you’re looking at, the needs, the infrastructure required, and things like integrating with a LIS, what things are required upfront and what things can be added at a later date.

Building the workflow includes looking at the physical side within the laboratory and the different steps in the workflow to determine where some steps may be eliminated by things like automated case sorting utilizing bar codes that automatically aggregate all slides associated with one case. This can eliminate the physical glass slide sorting that can be hugely time consuming.

Rolling out your configuration which is now unique and tailored to your organization and getting all your users up and running allows you then start analyzing the benefits of that use case and going back to the defined needs and goals and whether this solution supports your initial premise, then you can look at how to grow the solution to leverage it most effectively and expand on the applications.

Comprehensive Suite of DP Solutions

Three main areas to consider when looking at your solution are scan, manage, and analyze.

Not all scanners are the same and not all needs of departments are the same, so it is a case of considering what type of samples you're looking to scan and to look at your current throughput requirements and the volume that you’re looking to put through, as well as future proofing for what your needs might look like within the next one, three, or five years.

Once you’ve defined what you need from a scanning perspective, the key consideration is how you manage these images and the associated metadata, and integrating with the laboratory information management system so that you have automation of the transfer of information between the two systems eliminating the need for double data entry. This system should be secure and robust, but also flexible enough to manage your different images, your different levels of users, as well as your different deployment options tailoring the solution that’s specific to the needs of your department.

Analyze is the third pillar to consider. Are you going to interpret these manually where you can leverage things like the improved accuracy in doing measurements onscreen, or take it one step further and start moving towards the automated image analysis side of things where you start to bring in a level for biomarker interpretation or a level of reproducibility and quantitative analysis that give you objective quantifiable data? These are all things to be considered when looking at a solution, but also take into consideration that you can start small and scale these things up to provide the optimal solution for your department.

Digital Pathology at Johns Hopkins - (Alexander Baras, M.D., Ph.D)

There are two aspects when we talk about digital pathology. The first is whole slide imaging in which we take a slide and construct a high-resolution image of the entire slide and that is a static file that can be looked at over and over again. An alternative version of digital pathology is where we do live microscopy in which a live feed is coming off of a microscope and it’s presented on different monitors in the same office, in the same building, the same state, or around the world. The two have different strengths and weaknesses. Consideration has to be given to the difference in the time it takes to generate the digital version of the slide, with the live imaging from the microscope being nearly instantaneous, and whole slide imaging requiring an amount of time to prepare the digital version of the slide, distribute it into some framework, and then be able to analyze it. The biggest use cases in the context of this presentation has been in education and research along with the ability to generate high-quality publication-level images from our repository along with various use cases for telepathology and internal conferences at Johns Hopkins.

Digital Pathology Overview

An overview of the hardware and software stack that one should consider when implementing digital pathology starts with some type of hardware that’s going to take the slide or alternatively directly scan the tissue. Some of the criteria you'll want to look at for conventional whole slide imaging modalities include how it handles the slide, what kind of optics it uses to visualize those slides, and what kind of detectors it has. That set of the technology stack basically is a hardware component but there still is going to be some software that sits atop of that within the scanner to stitch together the different high-resolution patches that are taken of the image and eventually compress the data into a manageable format. From that point you want to consider what types of viewers and analytics you want to layer on top of that. Is your repository going to be a network-attached storage device within your institution’s infrastructure? Will you be looking for cloud storage? How will you visualize the slides? Will it be a local application on someone’s computer, will it be a web-based viewer, and what implications does it have for how well you can share and analyze your data? It’s important to think about these whole slide imaging formats being generally stored in a manner that mimics the way we think about visualizing slides on a microscope with different objectives representing different magnifications of the whole image.

Our overarching goal to further research education at Hopkins is to provide a simple web-based resource for viewing and obtaining whole slides. We want to foster collaboration both within Hopkins along with our collaborators around the world. We want to lay the framework down for a common infrastructure that will enable analytics on these slides because that’s the promise the digital pathology allows; to enable analytics image analysis routines to generate more data from these slides than we previously could. We are proponents of opensource algorithms in the context of these analogies. One of the biggest bioinformatics pushes came around 2000 when algorithms that process bioinformatic data were opened up to the academic community to be refined and perfected. That fueled a whole library of tools like Bioconductor in the R language, and similar things are occurring now in digital pathology that will propel the field forward. Our research and education web-based platform is available at the website https://digital.pathology.johnshopkins.edu/. You can log in as a guest or you can register an account within our system so there are more capabilities you will be able to do, and our internal Hopkins staff also has a dedicated login where they can come in and upload slides.

Research & Education

The ability to annotate cases starts with a concept of repositories. In order to enable our researchers to do this, each repository within this framework has its own set of attributes that each project might be relevant to. One repository may have certain attributes that relate to the pancreas cases that are relevant, however, other repositories might be combinations of TMAs and we may need to have different sets of attributes that we’ll want to be able to annotate these slides with. We have a variety of different cases that are represented in these TMA formats and that’s a natural place where a lot of these analytics are being applied because it’s a quick way to have a lot of different cases represented on the same slide. These are active areas of research that various investigators at Hopkins are using the platform to answer very interesting questions.

We also have for education the well-known Johns Hopkins Surgical Pathology Unknowns Case Conference Series. This is an example of a public repository you can access directly without having to log in. It represents a collection of nearly 2,000 cases. Each week at Hopkins we collect maybe five to ten interesting cases from the cases encountered that week. We annotate them with what the diagnosis is and then the residents prepare their thoughts on them to discuss with an attending each week. This has been a public conference for many years but the ability to share it with the world the last few years has been very exciting for us. While it does serve as an educational piece to be able to show these materials, in some sense it might also be an interesting case collection from a research and analytic perspective because these may be looked at for a potential to develop algorithms. We’re very interested in having a framework for analytics and this platform does enable that. We’re working closely with our technology partners on this to enable an API to access annotated slides for folks who want to dive a little bit deeper into these images and the data behind them.

For various degrees of difficulty in the analytic space, the most interesting and probably most applicable is to refine our ability to do quantitation of immunohistochemistry, in particular some of the more routine biomarkers. Most of the time people think about genomics but proteins are still molecules so this is still molecular diagnostics and is a critical piece of interpreting these tumors. Some of the common quantitative biomarkers that we look at today include ER, PR, Ki, HER2 and PD-L1, which has been a big important thing in the immuno-oncology space. These are considered low hanging fruits because it’s something we should be able to do a good job with. There are a lot of algorithms that are out there with promising results and we look to try and do largescale validations to get these off the ground.

The riper fruits are various deep learning approaches in imaging. This example is a project that my team worked on for the BACH Breast Cancer Challenge using deep convolutional neural networks to distinguish benign from malignant breast cancer. I think these techniques hold a lot of promise, in particular towards the notion of developing screening technologies, a concept that is definitely not foreign in pathology. There are a variety of things already in place in pathology that screen materials, both from a workflow perspective, humans screening slides for pathologists, or a machine. I think this holds a lot of promise to make pathologists very efficient.

More exotic fruits coming up in the next few years are various multiplex, multichannel approaches looking at the immunofluorescence along with conventional light microscopy approaches to multiplex or have multichannels of biomarkers on the same slide, which has a corollary to our ability to save precious biopsy tissues. If we only have to use one 5-micron section for the three or four biomarkers that we need at the immunohistochemical level, it saves material for other analyses that have be performed.

Some new fruits not yet ready are different techniques that have been published on recently such as this example on histographic microscopy in which we move away from conventional Brightfield intensity-base imaging of these slides and consider other modalities that allow us to render digital image, but with potentially more information than can be generated from Brightfield microscopy.

Computational Pathology

The regime of computational pathology is where we want to focus on numerous quantitative needs or presence in the molecular era, even if you're just talking about immunohistochemistry. There’s definitely a lot of novelty in the single cell parametric space so the notion of being able to do something like flowcytometry in situ means you’re now going to have the ability to analyze both the parameters of the cell along with its location. That’s becoming an interesting concept in immuno-oncology as well in the sense of what type of immune cells are present and where are they relative to the tumor cells and the novelty in the way with which we digitally capture these slides and how can we go beyond the conventional light microscope.

When considering your infrastructure needs try to be SAVI; think about Scanners, Analytics, Viewers, and Informatics needs. For research applications we focused on our ability to help our residents with didactics along with attendings throughout the world be able to generate a framework that people can layer annotations on top of visual pathology and an API for analytics with a focus on opensource for the actual algorithm development piece.

PROSCIA- (Nathan Buchbinder)

Proscia is a digital pathology, artificial intelligence, and software company. We build products that enable pathologists to effortlessly transition from their traditional microscope workflow to the digital world through a digital pathology software platform with modules for collaboration, image analysis, and artificial intelligence. By bringing together cloud and on-premises workflow software and pioneering work in artificial intelligence, Proscia provides a simple, flexible modular and futureproof software platform for pathology labs to go digital.

Data Management

The first broad set of functionality relates to our data management and workflow management platforms. We have two broad operating systems. The first is repository-oriented and it enables you to store, manage, organize, and otherwise work with large collections of whole slide images and data. The second is a workflow-oriented operating system which allows our users and our adopters to start with a case or with many cases, and then move them through a customer-defined digital set of processes.

The second set of use cases that our customers leverage our platform for relates to collaboration, whether that’s through case assignments or real-time telepathology with large groups or in a smaller group setting for research, for interinstitutional reviews, all of that happens on that same base platform on cloud or on premises. The third set of functionality and features pertains to digital IHC. Proscia offers a suite of reserchee’s-only stain-specific algorithms that aid in producing consistent assessment of IHC and H&E stained slides.

Proscia is a software only company and we believe that while digitization of these whole slide images in and of itself is necessary and useful, the real value comes in what software enables you to do with those whole slide images.

Considerations for Going Digital

Two big changes have happened in the past couple of years. The first is that computing technology including cloud technologies, have advanced tremendously and made their way into organizations. This radical shift has enabled new advances in the technology, new advances in the use cases that are provided, and in the user experience. At the same time, these advances have driven down the cost of going digital. The second major shift is not a shift so much as increased pressure to drive revenue to do more of those cases and at the same time to reduce costs. One thing that’s remained constant throughout the entire digital pathology history is that the customer, consumer, and user of digital pathology are in the driving seat. One thing you should never concede on is that digital pathology systems must conform to your imagined and ideal workflow and not the other way around.

Meeting Your Expectations

Your pathology group should be thinking about the wide range of requirements coming from your IT staff, your legal department, purchasing, pathologists and histotechnologists, being summed up into three basic questions. Does this work with what I’m already doing? Will this product serve me well today and tomorrow? Is this able to do everything that I want and need it to? Proscia views those questions as requirements concerning simplicity, flexibility, scalability, modularity, and building a futureproof product.

Adopting digital pathology can sound quite intimidating and it doesn’t have to be. At Proscia we focus on a few things to make going digital simpler, easier, and less intimidating. The first is ensuring we provide a flexible and scalable cloud and on-premise platform. This has to be secure, compliant, and well-managed with the right user permissions and user settings controls. The flexibility ensures it addresses the needs of your organization.

The second is provide use cases that can be added modularly and plug in directly to your existing platform as needed. Rarely is it a good idea to go from zero to completely digital in one swoop. That tends to cause a lot of frustration for everyone working at your institution and for the vendors themselves. A modular adoption allows you to start small, start with use cases that you know are very real, very concrete, and could benefit from digital and then scale from there.

The third is we make sure that we take the complexity of implementation, training, and actually using the digital pathology system out of the hands of the pathologists. If digital pathology is going to be adopted it has to be usable by the pathologist and not just by technologists or technology-oriented individuals.

Digital and Computational Pathology

Computational pathology and AI concepts mean taking the powerful advances in technology and applying them in a way that enables pathologists and labs. An example of this was recently announced by Proscia’s R&D team concerning dermatopathology, which is an AI-powered algorithm that was trained to be able to identify tumor regions on a whole slide image. While this remains a researchees-only technology that’s still is in development today, AI-powered technology like this, when plugged into a digital pathology operating system, like Proscia’s or others, and when applied appropriately to a workflow has the potential to elevate the economics and science of laboratory medicine and ultimately put powerful technology in the hands of pathologists.

Questions and Answers

Q: How do I make sure that the system I initially adopt is able to accommodate my lab as it expands its digital presence?

A: MR. BUCHBINDER

There are a couple of things you can do to ensure that you’re adopting a technology that lasts you not just for the use cases that you have today, but also for other ones moving forward, is to be as transparent and curious as possible upfront. If you know what you’re potentially looking to do in the future, certainly ask any of the vendors in the digital pathology space, and even if you’re not certain, it’s always helpful to have a comprehensive and complete view of what it is these companies, including Proscia and Leica, are looking to do. In addition, you should really be looking at systems that talk openly about compatibility and interoperability. While there’s a chance that one company may be able to address all of your needs in perpetuity and that the system that you adopt today is going to be able to expand indefinitely, realistically, that’s not the case. You need to find a way of making sure that regardless of what system you adopt today the technology itself is able to expand and integrate with other solutions.

Q: Can you tell us anything more specifically about the difference between digital and computational pathology?

A: MR. BUCHBINDER:

I think computational pathology is a newer phrase or term and everybody has their own nuances in the definition as to what it is. The way that I perceive it is that digital pathology, to put it simply, is what makes pathology data available and accessible while computational pathology is about turning that data into insights and putting those insights into the hands of pathologists and the laboratories. The key thing is very hard to have one without the other. You don’t have computational pathology without digital pathology, and for many individuals there's a real value out of having computational pathology on top of digital pathology. Computational is often the reason that people are going digital. It’s the promise and the potential that it holds. We see a synergistic relationship between digital pathology and computational and you really need to have one to have the other.

A: DR. BARAS

Digital pathology is a question specifically of digitizing the histologic images into something that can be viewed on a computer screen and diagnoses could be rendered directly off that via humans, much akin to radiology. Computational pathology speaks to taking that as one data source, potentially others such as genomics or potentially even immunological data, whatever that source is, deriving some additional data features from that raw data via imaging, genomics, or whatever, to drive insight particularly towards clinical application. I think digital pathology speaks specifically to digitizing these histologic slides and then computational pathology speaks more to deriving more numerical data from those images.

Q: What would you describe as the most important factors in evaluating digital pathology vendors?

A: DR. COLGAN

I think there are a number of things to look at when evaluating a vendor or different vendors. The first is starting out with your specific use case in mind and knowing what it is that you’re looking to achieve and keeping that at the forefront in keeping the end goal in mind, and making sure it satisfies all the different stakeholders that are going to touch this. There’s the laboratory element to this, usually from the scanning perspective, so your lab staff will need to be comfortable with the system that’s chosen. Interoperability with your existing systems you may have in place, and then also satisfying the IT and personnel side of this, as well that the system you choose can satisfy and plug into your existing infrastructure and your network, as well as the people who are going to be reviewing the slides, the pathologists, and the tactile element to this and the attraction of the usability of the system are going to be key. It is set out with the end goal in mind and then make sure that you bring all your different stakeholders along with you in the process.

Q: Why would anyone ever go from a 1.1-centimeter max diameter field of view on a microscope to a 20+” screen?

A: DR. COLGAN

I think if you were just to evaluate the use of digital pathology on your field of view, and yes it may not look like a fair comparison; however, I think it’s all the additional benefits digital pathology can bring to the table but just aren’t possible on microscope. Even just from a simplistic viewer perspective, the ability to look at two slides side by side, two or more slides. If you’re looking at serial sections of your H&E and then your associated IHC stains, to be able to view all of those within one single visualization can be key as well as other things like performing measurements to bring that level of accuracy and usability under some of the basics. Add to that then all of the things that we spoke about today around computational pathology and the potential benefits of that, I think there are an awful lot of additional benefits, not just the field of view, to take into consideration.

Q: What clinical field do you think will benefit most from computational pathology?

A: MR. BUCHBINDER

In my presentation I spoke specifically about results that relate to dermatopathology, but I’ll give you a little bit of insight into our philosophy as to why we think dermatopathology was a good start, but why it’s certainly not the only specialty that’s going to benefit from digital and computational pathology.

When you look at what some of these computational tools are doing, artificial intelligence included, they’re augmenting what a pathologist is able to do. You’re providing additional insight, additional metrics, additional quantification that aids the pathologist in doing their day-to-day work, or it automates or simplifies other parts of the workflow that are mundane but still important as part of the process that goes in a standard lab. The fields of pathology that are likely to benefit most from going digital or from adopting computational pathology tools are those where there’s a need for improved standardization, quantification, or where you could really benefit most from having these additional AI-driven insights, or those in which there’s the highest volume, the highest percentage of your workday that’s spent doing tasks that are either mundane, tedious, or easily taken over by a system that enables you to then focus the remainder of your time on the parts of pathology that really require you to practic

Leica Biosystems webinars, training presentations and related materials provide general information regarding particular subjects and are not intended to be, and should not be construed as medical, regulatory or legal advice. The views and opinions expressed are the personal views and opinions of the speaker(s)/author(s) and do not necessarily represent or reflect the views or opinions of Leica Biosystems, its employees or agents.

For the use of any product, the product information guides, inserts and operation manuals of the various products and devices should be consulted. Leica Biosystems and the editors disclaim any liability arising directly or indirectly from the use of devices, techniques or procedures described in these materials.

Copyright © 2018 by Leica Biosystems Richmond Inc. All rights reserved.LEICA and the Leica Logo are registered trademarks of Leica Microsystems IR GmbH.