Spotlight: Connecting Pathologists and Computer Scientists
Blog Posted: Jan. 12, 2021

The pathologist—no longer the mysterious doctor behind the microscope. Pathologists are increasingly becoming more visible through patient-pathologist consultations, high profile research collaborations and social media like Twitter and YouTube. We wanted to further highlight pathology by spotlighting pathologists from a variety of specialties to not only connect pathologists with one another but to show the critical connection pathology plays in every aspect of medicine. We welcome you to our fourth Spotlight featuring Dr. Aleksandra Zuraw.

Zuraw 1 - Copy (3).jpg

Dr. Aleksandra Zuraw (@olkazuraw) is a toxicologic pathologist at Charles River Laboratories and an online publisher at the Digital Pathology Place. She studied veterinary medicine in Wroclaw, Poland, where she worked as a practitioner before pursuing a veterinary pathology specialty in Berlin, Germany. After finishing her residency and PhD she worked for a digital pathology company, which started a great passion for this discipline and resulted in the creation of the Digital Pathology Place website – a blog and podcast with digital pathology resources, including a “crash course” video. The next steps in her career led her into the field of toxicologic pathology through Canada into the US, the sixth country where she has lived and worked!

When were you first introduced to digital pathology and what did you think about this new technology?

My first encounter with digital pathology was during my residency. We had an Aperio whole slide scanner and it was mostly used for illustrative and educational purposes. One of my fellow residents was starting to figure out the tissue image analysis possibilities, but my first impression was that it was an elite and restricted technology requiring a lot of expertise and not really accessible to ordinary pathologists. Then my first job as a pathologist happened to be with a digital pathology company, Definiens, and I found myself quickly and fully on board with digital pathology. Even though I did not have prior experience, everything felt intuitive. I was able to navigate the digital pathology environment and easily give guidance for tissue image analysis. This is when I developed a true passion for digital pathology and felt the need to spread it to all the people who felt the way I felt during my first encounter with digital pathology.

Your mission to bridge the gap between pathologists and computer scientists is critical in advancing digital pathology. How does one even begin?

I was not even aware of this gap until I started working as a pathologist supporting tissue image analysis and software development efforts. I was the first pathologist doing this job at my former company and quickly realized that pathologists and computer scientists not only look at the challenges connected to these tasks from different perspectives, but they also have knowledge gaps and lack understanding of how the other side works. On my side, for image analysis I had to learn how the image analysis tools worked and what were their strengths and limitations. On the other hand it was important to teach the computer scientists about the relevance of different tissue regions for our analysis as well as about the pathologist's workflow for optimal digital pathology software design and integration. This was over three years ago. Since then a lot has changed. More and more pathologists and computer scientists work together and understand each other, but there are always people just starting this journey, and if they are equipped with the "gap bridging" knowledge up front, their journey will be a lot more successful.

Aside from monetary resources, what are some top concerns clients have about going digital?

One concern, especially in the drug development world, is the regulatory status of digital pathology solutions. The perception is that validation of such solutions is an enormous effort and it is really not clear how to conduct it. In fact, the effort is similar to validation of any other system and scientific method and there is industry guidance available. Where the guidance is not straightforward or not specific enough then it is possible and advisable to engage with the regulators before starting the process. There will be digital pathology-specific challenges and hurdles, such as the size and management of whole slide images which are very data heavy, but remembering that validation is a team effort and engaging members of all involved teams, including IT, at the beginning of the validation project is crucial for success.

Another concern is often if the particular digital pathology system or software will work well for a specific pathology workflow and environment. Here my answer would be to go ahead and try it out and see. Digital pathology companies are often open to letting users try the systems before committing and the current trend in the industry is to support interoperability. If you identify areas preventing you from going digital then you are already a step ahead. You can both address the areas on your end as well as give the digital pathology solution providers feedback on what to improve. This is often a great incentive for digital pathology companies to improve their product. This is how it works in this dynamic industry - a constant feedback loop between the user and the provider. We may not be aware that we can contribute to the product improvement by giving feedback. This may not be the case in other industries, but it definitely is in digital pathology. This is still a relatively young field, so I very much encourage giving structured, constructive feedback to make the tools we will be using in the future intuitive and pathologist friendly.

How can pathologists contribute their expertise to image analysis software?

Now in the era of supervised deep learning it may seem like the main pathologist's contribution to image analysis is providing annotations, when in fact this can be taught to other image analysis team members. Pathologists can further help in a wide range of areas including image analysis study design, translation of semiquantitative pathology scores into image analysis outputs, and very importantly quality control strategy of the results.

What can computer scientists and pathologists learn from one another?

What we can learn from each other can be summarized in one sentence: pathologists learn from computer scientists how digital pathology and image analysis tools function, and computer scientists learn from pathologists where those tools can be applied. This enables both sides to recognize the potential and limitations of the tools early on in the project and find the best solutions to address the project's goals. To give a tangible example, in order to help with the annotations, pathologists need to know that in computer vision the inputs are pixel based and translating them into objects comes later. This will allow the pathologist or the pathologist's trainee to annotate with the adequate level of detail and consistency. On the other hand, the computer scientists working on the neural networks to address the image analysis problem need to understand the dependencies within the tissue in order to set the hierarchy of the different networks correctly (e.g. detect the larger tissue compartment, like gastric mucosa first and later look for changes specific for that compartment, such as ulceration only in this compartment). We also need to understand each other's language and be familiar with each other's terminology. The field is changing so rapidly that next year we might be learning totally different things from each other. The important thing is to keep an open mind and embrace the collaborative and multidisciplinary nature of digital pathology projects and provide expertise where it is needed.

A huge thank you goes out to Dr. Zuraw for her time in answering our questions and being a part of this series. We’d love to feature you in our next Spotlight article. Just email Kristin at kmitchener@instapathbio.com. Stay tuned because we are spotlighting more pathologists in the weeks to come! Be sure to read previous Spotlight posts which we’ve listed below:

Dr. Elizabeth Neyens, Toxicologic Pathologist

Dr. Nicola Parry, Veterinary Pathologist

Dr. David Suster, Bone and Soft Tissue, Pulmonary and Molecular Genetic Pathologist

Built on the vision of better patient outcomes, Instapath was founded in 2017 by engineers and scientists to enable patients to immediately know their cancer diagnosis. Our team made it our mission to develop fast and easy digital pathology technology so diagnosis can be made in minutes instead of days. To learn more about Instapath and our technology, visit https://instapathbio.com or contact us at info@instapathbio.com.

Mei Wang Mei Wang, Ph.D. | CEO at Instapath
Tags:  digital-pathology image-analysis