Cover image of Digital Pathology Podcast

Digital Pathology Podcast

Aleksandra Zuraw from Digital Pathology Place discusses digital pathology from the basic concepts to the newest developments, including image analysis and artificial intelligence. She reviews scientific literature and together with her guests discusses the current industry and research digital pathology trends.

Weekly hand curated podcast episodes for learning

Popular episodes

All episodes

Warning: This podcast data isn't working.

This means that the episode rankings aren't working properly. Please revisit us at a later time to get the best episodes of this podcast!

Podcast cover

What to consider when choosing an image analysis solution for phenotyping? (part 3) w/ Regan Baird, Visiopharm

This episode is brought to you by Visiopharm.In this third and last episode of the multiplex mini-series with Regan Baird from Visiopharm we look at the considerations when choosing an image analysis software for phenotyping.The two main points to consider when choosing phenotyping image analysis software are segmentation assistance and data visualization. Segmentation assistance:Before different markers are attributed to different cells in the tissue and cell phenotypes are determined, cell boundaries need to be delineated. The automatic delineation of these boundaries by image analysis software is called cell segmentation. Cells in tissue slides can have different shapes and sizes, which depend on the plane of sectioning, heterogeneity of the investigated tissue, and the disease stage. This makes the task of segmentation challenging. Unlike in single-cell confocal microscopy images, where the cell borders are very well-demarcated, in tissue they often need to be estimated. A separate segmentation (e.g., membrane) marker can help significantly, but a perfect cell segmentation is not attainable.To best estimate the cell boundaries, rule-based classical computer vision approaches or artificial intelligence (AI) – powered approaches can be used. In rule-based approaches, we are working with well-defined features on which the segmentation is based, but we need to make concessions. The AI-powered models are only as good as the examples we train the models on. To combine the advantages of both, Visiopharm offers an AI-based nuclear segmentation as the starting point and a rule-based and marker-based second step to obtain the most reliable cell segmentation for phenotyping. Data visualization:The adequate visualization and handling of the obtained data depend on the software used. To understand and interpret the multidimensional multiplex and phenotyping data we need to interpret graphs, plots, two-dimensional reduction plots, and other data visualizations for all the images in multiplex studies. In order to evaluate how well the phenotyping has performed and to export meaningful results, the correct visualization tools need to be used.If you need assistance or have questions about multiplexing and phenotyping visit the Visiopharm’s website and contact the Visiopharm team. This episode’s resources:Multiplexing mini-series Part 1: Introduction to multiplex for tissue image analysis (part 1) w/ Regan Baird, VisiopharmMultiplexing mini-series Part 2: How to make sense of multiplex data with phenotyping? (part 2) w/ Regan Baird


11 May 2021

Rank #1

Podcast cover

How to make sense of multiplex data with phenotyping? (part 2) w/ Regan Baird, Visiopharm

This episode is brought to you by Visiopharm.Multiplex tissue staining can generate large amounts of data to help identify distinct information about particular cells in tissue. Immuno-oncology is a field where it is common practice to use multiplexing, in particular for cell phenotyping in tissue.Phenotyping is the ability to classify every individual cell in the tissue based on the biomarker panel used. The panels are designed to identify cells of different lineages as well as cell activation states within each lineage, which is of utmost importance for the personalized therapeutic approaches in oncology. Although multiplex data can be visualized manually, e.g., by switching on and off different fluorescence channels, its interpretation requires computational assistance. If the multiplex assay only contains a few markers, the rules for detecting potential phenotypes can be designed manually, but as the number of markers increases the number of potential phenotypes increases exponentially. In order to sort through the cellular phenotypes in higher-plexes, machine learning-based auto clustering has been implemented. This method is based on the way cells are characterized in flow cytometry and has been adapted to automatically identify phenotypes of cells in tissue images. The adequate visualization and handling of the generated data depend on the software used. In the next episode, we will be talking about the considerations when choosing an image analysis software program for phenotyping. To learn more visit Visiopharm’s websiteThis episode’s resources:Multiplexing mini-series Part 1: Introduction to multiplex for tissue image analysis (part 1) w/ Regan Baird, Visiopharm


4 May 2021

Rank #2

Similar Podcasts

Podcast cover

OracleBio's path to GCP compliant image analysis clinical trail support w/ Lorcan Sherry, OracleBio

Founded in 2011 by this episode’s guest, Lorcan Sherry, along with co-founder John Waller, OracleBio entered the digital pathology market with a unique value proposition – to be a contract research organization for tissue image analysis and help other organizations in their biomarker discovery work. Fast forward 10 years and they evolved from a small service provider into a Good Clinical Practice (GCP) compliant organization supporting clinical trials, but still within the same paradigm of providing tissue image analysis services.Originally using one software package (Definiens), they diversified into others such as HALO and Visiopharm to be able to utilize the best tool for the job as well as match what their clients might be using for their internal image analysis projects. The transition to supporting clinical trials was a bit bumpy as Definiens abruptly discontinued the software license that OracleBio’s business was based on, but the organization quickly pivoted and secured a diverse image analysis toolbox which helped them stay in business. Currently, as image analysis methods are advancing and deep learning plays an important role in solving computer vision problems applied to pathology, OracleBio keeps expanding the toolbox incorporating not only ready-to-use software packages but also programming capabilities. This variety of tools allows them to address a wide range of projects in the most efficient way.For a CRO specializing in tissue image analysis, it is critically important to provide adequate quality control of the image analysis results. This process has been incorporated into the operations from the very beginning. Each project starts with the evaluation of the image quality – are they good enough for image analysis? Is there enough tissue? What about the tissue processing artifacts and the quality of staining? Only images that passed the QC criteria are used for algorithm development. In the next step, pathologists annotate the regions relevant for analysis (e.g., the tumor mass vs. the non-neoplastic tissue present on the slide) and later they provide region and cell annotations as ground truth for comparison with algorithm markups and correlation calculations.Apart from the annotations, pathologists provide educational sessions for the image analysis scientists to increase their knowledge about the problems which are being addressed with image analysis with the various projects. Although OracleBio is supporting projects along the whole drug development pipeline, their recent focus has been on supporting immune-oncology clinical trials, which lead the company on the path to GCP compliance. This was a big effort for the company and went far beyond image analysis quality control and software validation. This change affected the way work is done across the entire company and positioned OracleBio to bring in the image analysis capabilities to support clinical trials. While histologic techniques evolved from simple brightfield chromogenic single marker IHC stains to immunofluorescence-based multiplex, which are difficult to evaluate visually, the image analysis tools for this more complex imaging data lagged in terms of regulatory compliance. OracleBio’s decision to commit to GCP compliance definitely closed this gap. Even though the company’s focus shifted to the clinical part of drug development, OracleBio continues to serve their smaller biotech clients throughout their whole R&D process not only by providing high-quality tissue image analysis but also by partnering with other service providers, such as histology labs, to facilitate their customers’ drug discovery and development journey. To learn more about OracleBio visit https://oraclebio.com/.


2 May 2021

Rank #3

Podcast cover

Introduction to multiplex for tissue image analysis (part 1) w/ Regan Baird, Visiopharm

This episode is brought to you by Visiopharm.After experimenting with multidimensional, multimarker, and multicolor single-cell imaging modalities during his postdoc at Beth Israel Deaconess Medical Center in Boston, looking at 2D images of tissue stained just with hematoxylin and eosin (H&E) seemed to him a bit simplistic…and then he was tasked with doing tissue image analysis (IA). When relying just on H&E, IA can be a very challenging task. So, to both simplify it and extract more information from the tissue, multiplex staining can be implemented.  In this three-part episode miniseries Regan Baird, Ph.D., scientific sales manager at Visiopharm introduces us to the concepts of multiplexing and cell phenotyping as well as to image analysis approaches relevant for multiplex data analysis.Multiplexing in the context of life sciences is referred to as taking multiple measurements at the same time on the same specimen. With tissue slides the easiest method of multiplexing is immunohistochemistry (IHC) based virtual multiplexing where consecutive sections of tissue are stained with a single IHC marker and later each slide is imaged and co-registered to simulate the presence of several IHC markers in the tissue of interest. More complicated, but more precise methods allowing for visualizing cellular colocalization of biomarkers include multicolor bright field IHC (visualizing up to five biomarkers per tissue but colocalizing reliably a maximum of only two biomarkers per cell), immunofluorescence (IF) potentially with spectral unmixing, to increase the number of biomarkers per tissue section as well as per cell to nine, and imaging mass cytometry where instead of chromogens or fluorophores heavy metals are used, which increases the number of biomarkers up to 60 in a single section of tissue.All these multiplex modalities have their advantages and disadvantages, and the choice of the appropriate method should be guided by the design of the experiment as well as scientific and/ or diagnostic questions we want to address.For example, currently a widely used application of IF multiplexing is phenotyping cells in the tissue. This not only allows for the characterization of single cells but also lets us interrogate and investigate spatial relationships between different cell populations giving us information about the interactome of different cells and the environment in which they live. To learn more about phenotyping join us for the next episode of the Multiplexing Miniseries next week.This episode’s resources:Multiplexing mini-series Part 2: How to make sense of multiplex data with phenotyping? (part 2) w/ Regan BairdVisiopharmTop 20 Pathology podcasts you must follow in 2021 


26 Apr 2021

Rank #4

Most Popular Podcasts

Podcast cover

Accelerating custom design of AI image analysis algorithms for drug development using transfer learning w/ Sylvain Berlemont, Keen Eye

As a computer scientist, he knew that to make a real impact with image analysis there were only two areas: military and life sciences. Sylvain Berlemont, the founder of Keen Eye chose life sciences and never looked back. He started consulting for the industry during his biomedical image analysis research work in academia and he quickly saw that regardless of the applications, the questions asked and the problems to be solved were very similar. Patterns started to emerge and the next logical step was to form a service company and offer solutions to those problems.The service company later turned into a product company and today Keen Eye is a software as a service (SaaS) company leveraging artificial intelligence to design customized deep learning image analysis and computational pathology solutions to support the drug development process. KeenEye's SaaS allows the customers to access powerful computing resources and a user-friendly platform from their own PC. The platform hosts the algorithms and enables their deployment in an easy and scalable way.As KeenEye does not believe in designing "off the shelf" products for the complex image analysis problems of life sciences, the design of the algorithms happens in a customized way and a very close collaboration of pathologists and computer scientists is a key component of the process. Through such collaboration as well as the development of efficient processes and use of transfer learning, the time required to develop high-quality deep learning models was reduced from several months to a few weeks. To learn more, visit Keen Eye's website.


4 Apr 2021

Rank #5

Podcast cover

A new generation of whole slide scanners: faster, smarter, and more flexible w/ Don Van Dyke, Bionovation

Smartphone, smartwatch, smart TV...internet of things (IoT) and artificial intelligence of things (AIoT) is ubiquitous. But did it already make it into any of the digital pathology devices? Oh yes! There is a smart whole slide scanner out there. In this episode, I am talking with Don Van Dyke, the chief business officer of Bionovation Biotech. Bionovation holds a patent to a potentially revolutionary scanning technology powered by AI. Due to the ability of the scanner to predict the 3D architecture of the scanned tissue in the Z-axis the device is able to dynamically adjust camera focus exactly to the surface of the specimen and scan it ca. 70x faster than the classical whole slide scanners. Not only does it make the scanning faster, but eliminates the necessity of Z-stacking when scanning smears and cytology specimen. What is more, it is now possible to obtain high magnification images (80x and 100x) fast too, which practically removes most of the digital pathology hurdles for hematopathology and cytopathology. To learn more about Bionovation offer visit: http://www.bionovationimc.com/


14 Mar 2021

Rank #6

Podcast cover

Reimbursement for digital pathology in the clinic – how does that work? w/ Esther Abels, Visiopharm

Medical tests and procedures can get reimbursed. The basis of the reimbursement are the Current Procedural Terminology (CPT) codes developed by the American Medical Association (AMA). But how can such a code be obtained for digital pathology which is so much more complex than a group of tests or procedures that could be reimbursed on a fee-for-service basis?According to Esther Abels, Visiopharm’s Chief Clinical and Regulatory Officer, to align the digital pathology reimbursement with its value the fee-for-service paradigm needs to shift to a value-based reimbursement strategy. To determine the real value of digital pathology for patient care we need to -          Articulate the services provided and define their added value and uniqueness in patient care (e.g. risk assessment, improvements in responses to therapy, delay in disease progression),-          Gather data relevant to support the claimed added value (e.g. cost-effectiveness data),-          Ensure that the reimbursed fee is based on a combination of technology use and physician involvement,-          And identify the key values relevant for the decision-making stakeholders.Limited work has been done in this area so far, but if we look into the existing care decision-making and treatment patterns and analyze the claims for existing codes in the payers’ databases, we will be able to identify key datasets where digital pathology could make a difference and use this information to start applying for new CPT codes more aligned with digital pathology value. To analyze what steps would need to be taken to prove to the payers that a digital pathology test deserves reimbursement, let us take a tangible example of the Visiopharm’s AI-assisted metastasis detection in Lymph nodes application. This application has a technical, artificial intelligence-based screening component and a pathologist’s reviewing component. Currently in order to assess the presence or absence of cancer metastasis in lymph nodes several (even up to 60) lymph node sections need to be visually evaluated by a pathologist. Finding a metastasis in one of those slides is sufficient to make the diagnosis, but regardless all the other slides need to be reviewed as well. One of the benefits of the AI application would be to save the pathologist’s time, but reducing cost is not the only added value of such an application. The value proposition lies in adding value to patient care. In this case, using a computer algorithm would increase consistency and precision increasing the overall quality of the slide review. AI-aided slide review for metastasis would result in faster turn around not only for the cases where it was used but also for other cases, as the time for visual review could now be used for evaluation of other cases or spending more time on more complex cases again increasing the quality of patient care. Faster diagnosis means faster access to treatment, which often means shorter treatment times.Every time we are able to point out and overcome limitations in the current standard of care with digital pathology applications we have both a legitimate reason to get reimbursed for its use and an incentive to fight for it if we want to make the patients’ lives better. This episode's resources:“Aligning reimbursement for digital pathology with its value” 


8 Mar 2021

Rank #7

Podcast cover

Why and how should pathologists keep up with AI? w/ David Harrison, University of St. Andrews

This episode is brought to you by VisiopharmArtificial Intelligence is starting to cross from pathology research into pathology clinical practice. With several AI-based algorithms approved for clinical use in Europe and many more in the making, it is clear that rather sooner than later it will be an integral part of practicing pathology.  Does everyone practicing pathology have to keep up with this new trend? Those who wish not to and are close to retirement probably not, but everyone else probably yes. AI will become one of the pathologist's everyday tools and the use of this tool should be taught throughout the entire process of medical formation from medical student through to practicing pathologists through continuous professional education. AI is not scary, but it is a new technology we need to adopt, similar to how immunohistochemistry (IHC) was adopted. IHC entered the pathology practice only in the 1980s and today most practicing pathologists are using this method as an integral part of their diagnostic workflow. It was brand new not so long ago and the pathologist community had to figure out this new method and leverage it to better serve patients. An analogous situation is happening now with AI. Unlike some may fear, the primary benefit of AI is not necessarily to make the diagnosis and replace pathologists. The first thing that AI does is help pathologists manage workflow. It may sound unambitious but triaging cases that are safe/ normal, and allowing the pathologist to focus on cases that are more urgent or more high risk already benefits the profession tremendously and improves patient care. Pathologists should keep up with AI both to leverage its power to help pathology as well as leverage the collective pathology knowledge in different aspects of the discipline for the development of reliable AI tools designed to help pathologists. This episode's resources:Prof. Dr. David Harrison researcher profileiCAIRDVisiopharm


22 Feb 2021

Rank #8

Podcast cover

Virtual Immunohistochemistry. How Owkin uses artificial intelligence to generate IHC stains without antibodies w/ Victor Dillard

If you are working with immunohistochemistry (IHC) you know how challenging it can sometimes be to optimize all the steps in the process to obtain a high-quality stain. It often takes testing different antibodies, antibody concentrations, antigen retrieval methods, and incubation times. What if there was a way to produce an IHC stain virtually, without antibodies or even the need to step into the lab?Today's episode's guest is Victor Dillard, the commercial operation director of Owkin. Owkin is a company leveraging artificial intelligence and machine learning for medical image analysis and its offering includes virtual immunohistochemistry staining.  We talk about how it was developed, how it works, and how it can be deployed at interested institutions. To learn more about Owkin visit https://owkin.com/This episodes resources: Deep learning-based classification of mesothelioma improves prediction of patient outcome


31 Jan 2021

Rank #9

Podcast cover

Why and how is AI taking over the tissue image analysis field? w/ Jeppe Thagaard, Visiopharm

Machine learning is not a new technology, but it started to revolutionize pathology relatively recently. The ideal combination of untapped, abundant pathology data necessary to leverage machine learning and the relevance of pathology applications has drawn scientists to this field and caused an artificial intelligence (AI) explosion.  Within just two years from 2018 to 2020 AI-based tissue image analysis went from “cutting edge technology” to “mainstream”. The deep learning explosion started with the Camelyon challenge which served as a proof of concept for the technology. The algorithms performing best in breast cancer metastasis detection in lymph nodes were all deep learning-based. This success combined with greater accessibility of whole slide scanning and recently accessibility of open-source deep learning frameworks led us to where we are today. In computer vision, the task of the computer is to analyze images in a way that mimics how humans see.  This can be achieved in three main ways:·       Through rule-based systems by understanding the visual problem and writing rules such as intensity threshold definition, to solve it. ·       By machine learning, where we still determine the features of interest and manipulate the images to enhance the signal we are looking for, but the rules for detecting our features of interest are learned by the computer. We use approaches such as:Random forest,Bayesian classifier And other classical ML approaches·       and through deep learning, where both the features of interest and the rules to extract those features are learned from the data. This characteristic is at the core of AI power in tissue image analysis. Deep learning enables us to solve problems we could not solve before.It was not possible to solve many of the pathology tasks with rule-based systems because it was not possible to define rules complex enough to achieve a good output. Now that there is no need for rules this barrier has been removed, and we can just give examples of what we are looking for instead. Now instead of writing code, our task is to collect and curate data and generate examples of the structures we are looking for. Deep learning delivers image analysis to a much larger user base and empowers users who were not previously trained in image analysis to take advantage of this technology. This is a major breakthrough in this field. Shifting the main task in designing image analysis from writing code to curating data contributed to the greater involvement of pathologists who are uniquely trained in interpreting tissue and crucial to the process of assuring the quality of the data. However, they are not the only ones who can do this, which broadens the user base of this technology even more. Even though AI is so powerful and accessible, there is still tremendous value in the classical image analysis approaches and even more so in combing the classical rule-based and machine learning approaches with deep learning. Visiopharm’s platform enables this combined approach by having an ecosystem of classical and AI-based approaches that can play together to best solve the problem. In this way, the problem picks the method and not the other way around, which is how it should be. In the long-term, AI will help us get more insights into the pathobiology of diseases by helping in the interpretation of complex diagnostic modalities such as various multiplex assays.AI will be the push to go digital for everyone who wants to stay at the forefront of pathology. The development of this field in the next decade will be extremely exciting. 


17 Jan 2021

Rank #10