How Automation Can Help Fight Disease
Blog Apr 18, 2018 | by Ruairi J Mackenzie, Science Writer for Technology Networks
Incorporating automation into pathology workflows could potentially make for faster and more accurate disease diagnosis. We caught up with Coleman Stavish, Chief Technology Officer at Proscia Inc., providers of digital pathology workflows, to discuss how these benefits can change the field and what obstacles still lie ahead.
Ruairi Mackenzie (RM): How can the battle against major diseases be helped by automation?
Coleman Stavish (CS): The application of automation in medicine promises two things: helping solve the supply-and-demand problem in patient care and helping to improve the accuracy and consistency of diagnosis. China has 1.4 billion people and some 20,000 licensed pathologists. Compared to the United States, which has a similar number of pathologists for a population less than a quarter of China’s, the difference is staggering.
Specifically, in anatomic pathology, which plays an integral role in cancer diagnosis, there are a number of areas where the thoughtful application of automation can help pathologists and patients alike. For example, immunohistochemistry (IHC) is a big piece of the personalized medicine “puzzle,” yet pathologists’ interpretations of IHC slides are often manual, laborious, and non-standardized. Image analysis applications that can automate cell counting and the detection of rare events, such as mitoses, allows the pathologist to spend more time on higher-level diagnostic decision-making, while improving diagnostic and prognostic accuracy.
Elsewhere in laboratory medicine, we saw how almost two decades ago, automation was making its way through cytopathology. Innovations in both specimen preparation (e.g. ThinPrep) and automated or semi-automated computer screening systems yielded tremendous efficiency gains while improving the quality of care.
RM: This battle is one often fought at scale, by global teams across many different labs. If one lab in such a collaboration adopts a smart workflow, will that disrupt collaboration with labs not possessing such innovation?
CS: If anything, one lab’s adoption of digital pathology usually serves as a catalyst for other labs to adopt digital pathology. The main reason is the world can see that the digital lab is seeing the benefits promised by vendors, as well as some additional benefits they may not have anticipated. The second reason is digital adoption requires careful planning by lab management. Seeing another lab go through the process first allows subsequent labs to learn from any mistakes made and feel more confident as they transition themselves.
In pathology, collaboration between labs usually takes the form of a consultation, in which a referring organization sends slides to a consulting organization with greater sub-specialist expertise. The standard operating procedure for sending consults is FedEx--you can’t get any more low-tech than that! Mailing specimens will always be an option (for domestic consults, at least), but as labs adopt digital imaging capabilities, we’ll see more and more consults be sent electronically.
Now, as pathology labs increase their digital footprints, interfacing between different systems will become more important than ever. Everyone knows how to deal with the interface that is shipping and receiving a FedEx package. But will one lab’s digital pathology system be able to communicate with another lab’s system? It’s critical that purchasers hold vendors accountable on the matter of interoperability.
RM: How easy is it for a lab to adopt a smart workflow? How has Proscia streamlined this process to make installations seamless?
CS: Incorporating digital pathology into lab operations is more complicated than one might expect. Lab workflows are the products of years of gradual optimization; the prospect of changing them makes many people nervous. The prospect of large, upfront IT costs (a small lab might produce 15TB of data per month) doesn’t help either. Proscia’s solution addresses both of these concerns. First, our software can be configured for multiple different use-cases and workflows, enabling labs to pick just one or two areas to pilot digital. This enables lab personnel to get comfortable in the revised workflow through gradual change and productivity gains. Second, Proscia leverages Amazon Web Services, Microsoft Azure, and Alibaba Cloud to provide labs with scalable, pay-as-you-go storage infrastructure. When your storage backbone is a worldwide leader in secure, compliant storage, your data is probably safer than it would be housing it on-premises. That said, many institutional IT policies are cloud-averse, and Proscia also supports deployments to customer-owned infrastructure. Additional lab concerns revolve around interoperability. For example, “I have two scanners from two different vendors, can one software system read both file types?” or “I am heavily invested in my anatomic pathology laboratory information system (APLIS)-- can my digital pathology system interface with it?” Proscia software is compatible with all the major file formats (vendor-agnostic) and supports integration with existing LIS and EMR systems.
A smart workflow might look something like this: a breast pathology specimen comes into a lab. The tissue is fixed, embedded, and the initial H&E slides are prepared, as usual. Instead of those slides going directly to a pathologist, they are scanned and an algorithm predicts if any re-cuts or IHC stains are needed. If yes, those additional slides can be prepared right away and then the pathologist can have the full complement of slides needed for diagnostic and prognostic assessment. This can equate to much faster turnaround times for labs--without the algorithmic intervention, the standard workflow would have the pathologist review the first H&E slides, then manually order additional stains. This could take a day or more to arrive, delaying the final pathology report.
Proscia makes implementing workflows like this seamless, by automatically loading images off of the whole-slide scanners (any vendor) and by interfacing with the anatomic pathology laboratory information system (APLIS), which typically manages ordering, reporting, and other day-to-day laboratory tasks.
RM: Tell us about digital pathology. Are there examples of digital techniques advancing pathology research?
CS: It is a very exciting time to be working in biomedical informatics, particularly on projects that make use of whole-slide images. For over a century, pathologists have looked at tissue prepared on glass slides through light microscopes. The visual patterns of tissue on those slides are incredibly meaningful--they tell the pathologist whether a patient is suffering from benign inflammation or a deadly cancer. However, the amount we could ever learn from those patterns was limited to human specialists, a major bottleneck for research. With whole-slide images, a glass slide containing tissue is scanned at high resolution, creating an image that is just as clear as one would get looking under a microscope. As tissue samples are captured as digital images, they can be analyzed in exhaustive detail by specialized computer algorithms, finding completely novel correlations and insights that escaped the notice of human eyes for decades.