Author: Visiopharm

Ken Pritzker co-founded the company Proteocyte AI in 2011. The company just brought to market their first prognostic LDT test for oral pre-cancer lesions, Straticyte.

Visiopharm: Can you tell us more about your company Proteocyte and how it all began?

Visiopharm: What led you to choose Visiopharm and how has it been working with the company?

Dr. Ken Pritzker: We already knew the company and had worked with the software for several years back then at the Mount Sinai Hospital in Toronto. Already at this time, Visiopharm was one of the market leaders of morphometry applied to digital images of tissue slides. We already had a good and long relationship with them and the team, who provided advice. Back then it was just a few people, and we had a great relationship. They have been excellent, and the quality of support has been really superb, so we’ve been very happy with it. Visiopharm software is used to quantitate the immunohistochemical marker and it’s also used to measure certain morphological aspects of the cells. With the support of the Visiopharm team, we have created several APPs in Visiopharm, covering the whole workflow from region detection to score calculation.

Visiopharm: So how does Straticyte work and how can physicians use it?

Dr. Ken Pritzker: Our product Straticyte is an LDT test and it combines the evaluation of this specific IHC cell marker with an evaluation of cell morphometry. There are very few, if any, other biomarkers products out there which have a combination of two very different modalities. So this was an innovation. The other is that we saw ourselves from the beginning as an information company. The idea would be that the image of the oral epithelium would be transmitted to the company for analysis, and that we would have a serial database from which we could continue to develop the product and that has been a successful approach. Straticyte can calculate the 5-year probability that dysplastic lesions will progress to cancer. It has a sensitivity of 96% and a NPV of 95%, and those are significantly higher than the values for histological dysplasia grading, which are 75% and 59% respectively.⁴ The test is so good that if the Straticyte value is low, the patient can go back to live a normal life. But for cases with an elevated Straticyte value, those patients need to be watched closely and followed up against a possible recurrence or cancer. So compared to the current practice, Straticyte provides greater objectivity, sensitivity, and predictive power.

The way it works is that the physicians anywhere take a biopsy and that is assessed by the pathologist for whether there is pre-cancer present. If so, the pathology lab forwards unstained slides to one of our reference laboratories, which are contracted with Proteocyte and this lab performs the stains. The scanned image is then sent to us and we analyze the image and issue a report.

Figure 2: From: Hwang JTK et al, 2017, doi: 10.1016/j.oooo.2016.11.004

Visiopharm: How did you develop and validate your product?

Dr. Ken Pritzker: We started with samples from 250 patients with oral pre-cancer. From there we expanded to other sites in Canada, a site in Ireland and several sites now in the United States. By now, we have a database of more than 600 patients with clinical follow up. That’s probably the largest database of oral pre-cancer that exists, because typically, these patients present to local hospitals and if they have pre-cancer, they may not be followed up very well. So we know exactly the relationship between the test and the prognosis out beyond five years. It’s now more than ten sites and very different populations. And we noticed that lesions from different places have different characteristics. Those might reflect the state of health care in different places or the state of nutrition and this cannot be picked up if you do a single-site database.  

At this point, the algorithm is very mature and our laboratory has received ISO15189 and CLIA accreditation. In the background we work on improving the tests, but it requires a considerable validation operation to move it from the version which we’re using, which is quite good to the next version, which might be a bit better.

Learn more about Straticyte.

——————————————————————————————————————–

Main references related to the interview

1. Tripathi et al, 2010, DOI: 10.1371/journal.pone.0011939 ↩︎
2. Dost F, Le Cao K, Ford PJ, Ades C, Farah CS. Malignant transformation of oral epithelial dysplasia: a real-world evaluation of histopathologic grading. Oral Surg Oral Med Oral Pathol Oral Radiol. 2014;117:343-352. ↩︎
3. Fleskens S, Slootweg P. Grading systems in head and neck dysplasia: their prognostic value, weaknesses and utility. Head Neck Oncol. 2009;1:11. ↩︎
4. Hwang et al, 2017, DOI: 10.1016/j.oooo.2016.11.004 ↩︎

This blog post is inspired by the scientific poster presented at the Pathology Visions conference in 2021 by Daniel E. Winkowski, Ph.D., Jenifer Caldara, Brit Boehmer, Ph.D., and T. Regan Baird, Ph.D.

Today, let’s explore three different approaches to segmenting cells in samples stained with various multiplexed fluorescent assays.

Starting with the challenge: Precision in Complexity

In tissue pathology, multiplex tissue images have become crucial, allowing us to observe each cell’s location and understand its complex features. Yet accurately and automatically defining each cell’s outer boundary for analysis proves challenging. The reason for that is that tissues contain a heterogeneous mix of cells, each of their own shape and size. Commonly used methods for cell segmentation only focus on nuclei, but those often fall short in precisely defining cell edges.

Let’s compare the methods: Traditional vs. A.I.-Powered 

In this study, Daniel Winkowski and team used Visiopharm’s image analysis software to explore three different approaches to segment cells in samples stained with different multiplexed fluorescent assays:

1. Traditional Image Analysis (Trad): This method uses built-in filtering to enhance objects of interest, relying more on morphometry than intensity. While effective in identifying nuclei based on shape, it’s time-consuming and can struggle with diverse nuclear morphologies.

2. Ready-To-Use Deep Learning (AI): This approach uses a deep learning algorithm for nuclear signals identification and segmentation. It estimates cell boundaries based on dilation – a no-hassle method but still an estimation. This approach can underperform in identifying cells with clear biomarker staining but lacking a clear nucleus in the plane of section (i.e., CD68 macrophages).

3. Adapted Deep Learning (AI+): The real star of the show! This method augments the deep learning algorithm to include signals from biomarker channels, assisting in accurately delineating cell boundaries. This blend of nuclei and biomarkers, coupled with an edge-finding filter, offers a more precise approximation of the cell border and assists in finding complex cells (i.e,. macrophages, etc.) where nuclear identification methods fall short.

Visiopharm offers a complete toolbox for design and customizing your own algorithm but also offers a general nuclear detection algorithm which uses Deep Learning AI to segment nuclei of most morphologies (Ready-to-use AI approach).

To create the AI+ approach, the team adapted the ready-to-use algorithm to combine its nuclear AI classifier with signals from biomarkers in the multiplex panels to expand the segmentation beyond the nuclei, to the outer cell periphery.

The performance of these approaches was quantified by comparing them with manual annotations as a ground truth.

Ground truth annotations were prepared by three human observers on images from two different instrument manufacturers: Vectra Polaris 8-plex lung cancer (Akoya Biosciences) and Hyperion 13-plex spleen (Standard BioTools).

Manual annotations were prepared by three unique human observers on images from two different instrument manufacturers. Human observers utilized both nuclear and biomarker signals to aid during annotation placement. Note the variability across observers.

The AI’s performance was assessed against the ground truth annotations using Precision (a ratio of true and false positives), Sensitivity (a ratio of true positives and false negatives), and DICE Score (a ratio of all true/false, positive/negatives).

Results: AI+ Takes the Lead 

After rigorous testing against manual annotations (the ground truth), the results were clear. While all algorithms performed admirably, the AI+ method showed the highest concordance with ground truth annotations. This means a more complete and accurate segmentation of cells across varying multiplex images.

• DICE Scores for AI+ method show improved performance vs Ground Truth
• Precision and Sensitivity (measures of object detection) are similar across methods

• DICE Scores for all methods show similar performance vs Ground Truth
• Precision and Sensitivity (measures of object detection) are roughly equal across methods

The Visiopharm has developed a flexible, AI-based strategy enabling comprehensive cell segmentation across various multiplex images.

• All algorithms performed at a high-level using industry standards including DICE Scores, Precision and Sensitivity. 
• The Adapted Deep Learning algorithm showed highest concordance with Ground Truth annotations.
• Performance differences across instruments could be attributed to resolution differences of the instruments, the number of biomarkers in the panel, and differences in images tissue and disease state.
• Despite annotation differences among human observers, the algorithms perform consistently across the image set.

Would you like to see how Visiopharm’s software could support your cell segmentation challenges in multiplexed images? Learn more about Phenoplex, a complete workflow for all your multiplexed image analysis needs.

Welcome to our latest blog post, where we delve into the evolving landscape of histology and image analysis through an exclusive interview with industry experts from NovAliX – a fully integrated drug discovery CRO based in Strasbourg, France.

In this insightful conversation, Didier Merciris and Florence Anquetil-Besnard from NovAliX share their experiences in incorporating Visiopharm’s image analysis software into NovAliX’s workflows, and the profound impact it has had on their projects and clients.

Visiopharm: Welcome, Didier and Florence. Can you tell us about NovAliX, the services you offer, and how you differentiate yourself from other CROs? 

Florence Anquetil-Besnard: NovAliX is a fully integrated drug discovery CRO. We support biopharma companies with a range of programs, from initial target identification to the delivery of preclinical candidates. We have seasoned teams in medicinal chemistry and pharmacology, covering a variety of fields – oncology, inflammation, fibrosis, infectious diseases, kidney diseases, and more. We work with a diverse selection of screenings and characterization techniques, including a unique DNA encoded library platform, along with a vast array of biophysical methods, including Cryo-Electron Microscopy (Cryo-EM). Such advanced technologies are rare finds in the market and enhance our research capabilities significantly.

Didier Merciris: In 2023, our former company, Galapagos, transferred all its discovery resources located in France to NovAliX. This strategic move included integrating a team of highly trained experts, each with over 20 years of experience in fields such as chemistry, protein production, in vivo pharmacology, DMPK (Drug Metabolism and Pharmacokinetics), translational science, and histology. We now offer comprehensive histology services tailored to any requirement within the histology workflow. This ranges from basic tissue processing to advanced staining techniques and in-depth image analyses.

Visiopharm: What problems were you trying to solve when you started working with us compared with previous analysis software and how did Visiopharm help you overcome those challenges? 

Didier Merciris: For years, we had been working with an old-fashion image analysis software with limited machine learning capabilities. Developing an image analysis project with this tool required extensive annotations and post-processing steps, which were not only time-consuming, but also yielded unsatisfactory results.

We were literally stuck due to the software’s limitations, which significantly affected our ability to accurately measure data on slides. This was a major source of frustration in my daily work.

Florence Anquetil-Besnard: When I joined NovAliX, one of the first things I did was to share my previous experience with Visiopharm. Indeed, I thought the software could significantly improve our capabilities due to its deep learning features; so, we quickly organized a proof of concept with David Mason (Technical Sales Specialist at Visiopharm). The aim was to present him with our most common and challenging issues. We aimed to test the system thoroughly. As expected, Visiopharm, through David’s efforts, was able to provide solutions to all our challenges.

Didier Merciris: During David’s demonstration, I was really amazed by how Visiopharm easily resolved all our imaging analysis problems. Visiopharm is a complex software to handle, however, thanks to the exhaustive training that we had when we bought the software, it was manageable to learn all its nuances easily.

Florence Anquetil-Besnard: I would say that we are now better equipped to address the complexity of organs. Taking kidney as an example, it is an organ divided into four main compartments: glomeruli, tubules, blood vessels, and interstitium. These parts are independent from each other, but can also be all connected together, posing unique challenges for analysis. Now, with Visiopharm, we can conduct detailed analyses in each area, a task that was significantly more challenging before. For instance, we can independently assess fibrosis in the glomerulus and in the interstitium. This improvement is largely due to the deep learning capabilities of the software.

Even more impressive, we are not limited to broad compartment analysis; for example, in mice, a single section can contain up to 200 glomeruli. Currently, we can isolate and derive specific data from each glomerulus independently.

This represents a substantial advancement for our research. Consequently, we can achieve more accurate and representative results.

Typically, working with a pathologist yields a score for the overall tissue. Now, we can provide a more detailed picture of the tissue section, not just an overall score but insights into specific changes occurring within the tissue.

Visiopharm: What were you specifically looking for in a solution, how did you go about your search, and what ultimately led you to choose Visiopharm?

Florence Anquetil-Besnard: For advanced users like me, the support team is also there to bring you even further than what you thought. For example, once I had a big challenge with a complex macrophage staining project involving multiplex techniques. The heterogeneity of the staining was problematic, with intensity varying significantly across samples. In the image analysis community, heterogeneity of fluorescent staining is a nightmare. So, I called the support team, and they guided me through an unconventional use of the software, not something that is taught at first sight. With their help and additional processing steps, we managed to overcome this variability, enabling us to deliver a reliable and robust analysis to our client.

This experience highlights the importance of not only overcoming technical challenges but also of being able to deliver solid, scientific solutions to our clients.

Florence Anquetil-Besnard: In the imaging community, it’s crucial to stay updated and leverage the most advanced tools to remain competitive. Our requirement was for a flexible solution capable of meeting the market’s existing standards. While the Oncotopix® Discovery software works for up to 4 markers, the transition to real multiplex analysis requires a more sophisticated approach. The introduction of Phenoplex™ has significantly simplified the analysis of complex multiplex immunostains.

Didier Merciris: Another major advantage that influenced our decision was the exceptional support provided by Visiopharm. As a “newbie” to the software, I really appreciate the constant availability of the support team whenever I need them. They respond quickly and efficiently to any queries or issues, which has been incredibly reassuring.

With Visiopharm, you’re never alone with your challenges; there’s a supportive community, the blog, and a dedicated support team ready to assist.

Visiopharm: What specific features of our software do you find most helpful?

Florence Anquetil-Besnard: With Visiopharm, the process is intuitive: start with an image, add annotations, and if opting for deep learning, proceed to train and validate, among other steps. But what is also really nice is the APP Center – it’s like entering a magical world where you can pick an existing app created by someone else, make minor adjustments to fit your needs, and then all of a sudden you have your own customized app, that you didn’t spend time to create, yet still reliable and efficient.

This flexibility is really helpful to adjust our workflows, offering both time savings and inspiration by exposing us to ideas from other APPs, created by other users in multiple different ways.

Moreover, the software’s versatility extends beyond just brightfield images, but multiplex fluorescent and special stains, enhancing our research scope. We are empowered to combine different histological techniques, like integrating IHC and ISH, allowing simultaneous investigation of proteins and RNA, or merging special stains like H&E or PAS with targeted IHC staining.

We can extract as much information as we want from the minimal amount of tissue, which is very helpful when you don’t have a lot of samples to work with, so we can maximize the analysis with just one image.

Didier Merciris: I used to work with another solution, and to be honest, Visiopharm is faster than the other software. It’s significantly quicker because it can process multiple images concurrently. While the speed can vary based on the user’s IT infrastructure, Visiopharm consistently delivers faster results.

Compared to our former software, we’ve seen a substantial reduction in processing time, enabling us to deliver data to clients much more rapidly. Another advantage is the user experience, the software was built with an accessible interface that is very easy to use after a little training.

Visiopharm: Could you share some insights into how the software has added value to your business and your customers?

Didier Merciris: As you know, the power of Histology lies in its spatial resolution at the cellular level.

We had a client interested in quantifying mucus production in epithelial cells from a diseased colon, without considering inflammatory cell infiltration.

The AI capabilities of Visiopharm significantly aided us in meeting the client’s needs within an extremely short timeframe.

Moreover, thanks to deep learning, we can tailor our algorithm to work with any staining. Now, if there are any color variations on the slide, we can simply add them to the APP.

This allows for much greater consistency in our data, which is a significant advantage for us and our clients.

Florence Anquetil-Besnard: An important aspect is our adaptability as a CRO; it’s not limited to Histology. We can adjust to almost any image. A recent unique project involved working with an image from a camera—not even a microscope, but a camera. We developed a program that automates the counting of tumor nodules in photographs from an ex vivo organ.

Previously, scientists had to manually count all the nodules. We’ve automated this task. By collaborating closely with our colleagues, we’ve enhanced the reproducibility and reliability of organ assessments. This project was internal, but ultimately, our client benefitted the most because we were able to deliver faster results.

Stay Ahead – Realize the potential of AI-driven precision pathology.

In conclusion, NovAliX’s integration of Visiopharm’s image analysis software has significantly transformed their histology and image analysis workflows, offering substantial benefits to their services and clients:

• Enhanced research capabilities: The adoption of deep learning and artificial intelligence within Visiopharm has empowered NovAliX to overcome previous limitations in image analysis, enabling more detailed and accurate data extraction from tissue samples.
• Customization and flexibility: The ability to customize algorithms and adapt to a variety of image types has allowed NovAliX to tailor their services to the specific needs of their clients, enhancing the adaptability and reach of their offerings.
• Overall business value: By leveraging Visiopharm’s technology, NovAliX has not only enhanced their service quality but also solidified their position as a cutting-edge CRO, capable of delivering high-quality, reliable results to their clients in the drug discovery industry.

Curious to hear more about Visiopharm software? Contact us here.

Reach out to NovAliX to hear more about their services here.