Chaim Linhart, PhD is the CTO and Co-Founder of Ibex Medical Analytics. He has more than 25 years of experience in algorithm development, AI and machine learning from academia as well as serving in an elite unit in the Israeli military and at several tech companies. Chaim has a PhD in Computer Science from Tel Aviv University and has won multiple Kaggle machine learning competitions.
Since 2016, Ibex has led the way in AI-powered diagnostics for pathology. The company set out to transform pathology by ensuring that every patient can receive an accurate, timely, and personalized cancer diagnosis. Today, Ibex is the most widely deployed artificial intelligence platform in pathology. Developed by pathologists for pathologists, their solutions serve the world’s leading physicians, healthcare organizations, and diagnostic providers. Every day, Ibex has the privilege of impacting the lives of patients worldwide. The platform raises physician confidence, streamlines diagnostic workflows, helps clinicians provide more personalized diagnoses, and, most importantly, enables better clinical outcomes.
Can you share the journey and vision behind Ibex’s founding and its mission to transform cancer diagnostics with AI?
In 2016, my co-founder, Joseph Mossel, and I learned about the direct impact a digital revolution in pathology could have on improving cancer diagnostics. Radiology had gone through a similar transformation 20 years earlier, which had a prominent impact on how the specialty was practiced. With pathology becoming digitized, we recognized it provided an opportunity to develop new advanced tools that utilize artificial intelligence (AI) to perform sophisticated image analysis. We have focused on developing AI-powered tools that help physicians in reaching more accurate, objective, reproducible diagnoses, and thereby helping each patient receive the right diagnosis, in a timely way, which leads to the best possible treatment.
How has the landscape of cancer diagnostics changed since Ibex’s inception in 2016?
Labs have been adopting digitization at an increasing rate, even further accelerated by Covid-19. The digital revolution has enabled the labs to broaden their capabilities beyond the microscope in an impactful and meaningful way, leveraging AI that helps pathologists analyze and understand results efficiently.
The cancer diagnostics AI field has grown exponentially, as we’ve been seeing startups and other companies working on various aspects of AI for pathology in the cancer diagnosis realm. Precision medicine, for example, is data-driven patient stratification enabled by an accurate diagnosis and various informatics approaches that lead to optimal, personalized treatment. An increase in precision medicine comes with an enhanced need for more complex diagnostics to support the new targeted treatments.
We’ve also seen an increase in academic publications and industry associations focusing on the field. When Joseph and I attended our first conference on digital and computational pathology in 2016, AI was a small sliver of the conversation surrounding cancer diagnosis, as it wasn’t as mainstream. Now, when attending a large pathology conference, AI is the main event.
What differentiates Ibex from other companies in the field of AI-powered pathology?
When we talk about AI-powered pathology, there are several subdomains. There are companies that prioritize research applications, like tools that analyze tissue images to help understand disease processes at the morphological and cellular level, for example. Secondly, there are companies that focus mainly on clinical applications, i.e., products that are used in labs to support routine diagnosis.
Ibex is focused on clinical applications, and we have the largest and most widespread installation base with pathologists around the world using our tools daily for cancer diagnosis. We are also partnering with Pharma to develop AI-powered clinical applications that support pathologists in quantifying biomarkers that enable targeted therapies.
Additionally, while some companies focus on specific, limited indications per tumor type, like cancer detection, our approach is to train the AI to analyze everything a pathologist would see in these tissues. It’s not only about cancer detection, but also the type and subtype of cancer, the grade, its size, as well as cancer-related morphologies and other clinical features. We know pathology is more than just determining if the patient has cancer or not. We want to help pathologists realize the vast benefits that AI brings to the table.
Can you explain the core technology behind Ibex’s solutions and how it assists pathologists in cancer detection and grading?
Our approach is that pathologists essentially train the machine. We have a large team of pathologists around the world annotating slides. This means, they mark specific areas within those slides and label them. They may mark a low-grade tumor, a blood vessel, a nerve, inflammation, and so on. We then take that data and use it to train the AI models. This ensures that the AI is very accurate, even for rare and difficult cases, which is vitally important. Our AI is taught by pathologists and is trained to identify many different types of structures and morphologies of the tissue, which is very helpful to pathologists and inevitably increases its accuracy. By having access to a breadth of data and knowledge, we’re able to improve our AI and implement learnings with the feedback obtained directly in the field.
How does Ibex ensure clinical-grade accuracy across different cancer types such as breast, prostate, and gastric cancers?
This takes a lot of hard work. We collect data from many partners around the world. We ensure the data is very diverse, with representation from different labs and various tissue preparation techniques, scanners, and clinical findings. We enrich the training data with rare types of cancer. This ensures the AI is trained with a wide variety of features. During the training process, we measure what the AI does well, and we also determine where improvements need to be made. The team, with vast experience in machine learning, tests the AI on thousands of slides that we collected from different labs. We run studies and clinical trials and compare two fundamental aspects of the system. First, we review its standalone performance compared to the ground truth. Second, we determine how accurately the pathologist works with and without AI. In doing so, we ensure the AI is accurate, robust, unbiased, and safe. We measure its impact on the pathologists using the AI. Across our applications, we see that the pathologist, with the assistance of AI, reaches better results (meaning more accurate, higher agreement with the ground truth) than in standard of care (i.e., when they are not supported by the AI). We also measure the efficiency of their work and other important benefits of the AI platform, such as optimizing the workflow in the lab and decreasing the turnaround time (how quickly the patient receives the results).
What are some unique features of Ibex’s solutions that enhance diagnostic workflows and improve patient outcomes?
Our integrated system includes a slide viewer, the AI results, and built-in reporting tools. This holistic system was designed to enhance accuracy and productivity. It walks pathologists through the diagnostic process, showing them the main findings in every case and slide. Instead of searching for features, which can be small and hard to detect, the AI highlights everything very clearly. From there, the pathologist can confirm or modify. The AI shows measurements and quantifications; it also scores everything. With built-in reports, the pathologist doesn’t have to look at the slide, make the diagnosis in their mind, and then go to another system and report everything; instead, reporting is done while the AI is driving the integrated workflow. Even the number of mouse clicks was optimized. Everything was built with pathologists in mind to enhance diagnostic accuracy and efficiency, thereby creating a better work environment for these physicians with better outcomes for their patients.
How does Ibex’s solutions integrate with existing digital pathology software solutions and laboratory information systems?
We work with several vendors in the field that sell image management solutions or offer lab information systems. For each partner, there are different types of integration opportunities. In some cases, we embed our AI into their tools so the pathologist can use their platform with our AI inside it. In other cases, we integrate with these tools in a way that allows pathologists to launch Ibex from the other system. Regardless of the integration, we always want to make sure the users have the most optimal way of using the AI. Additionally, we have developed an open application programming interface (API) that allows third parties, including other companies or customers’ IT departments, to retrieve information from our AI and integrate it into their environment.
What challenges did Ibex face in achieving widespread adoption of its AI-powered solutions in pathology?
Upon reflection, I’d say the main challenge Ibex faced was around the sheer complexity and the amount of work, effort, and time required to bring diagnostics products to market. This includes multidisciplinary approaches: collecting data, working with pathologists, training the AI and testing it rigorously, running clinical trials, and, in some geographies, gaining regulatory clearance – and doing all of this under strict quality assurance measures. In the medical field, it is also extremely important to generate scientific evidence and publish results with multiple labs to demonstrate the performance and benefits of the AI platform.
Another notable challenge is integration. We need to make sure that pathologists can use the AI in a way that is efficient and natural. There are multiple systems in the lab: digital pathology scanners, the lab information system and workflow, and reporting tools. Put simply, we make sure everything comes together in the most efficient way possible, despite the challenges.
Can you share some success stories or case studies from healthcare organizations that have implemented Ibex’s solutions?
We’re very proud of our partnerships and global reach. For example, we have the first nationwide deployment of AI in Wales – all of the Health Boards in Wales are using Ibex’s AI solution. Another example is CorePlus Laboratories in Puerto Rico – they have been using Ibex for several years and published a paper, which shows the impact the platform has had on their clinical practice. As an example, using the AI algorithm, the pathologists were able to identify 160 men that otherwise would have been misdiagnosed. Those patients were given the right treatment thanks to the AI’s support. That’s really the impact that we’re making. It’s something we can’t forget – we’re here to impact people’s lives.
What role do you see AI playing in the future of pathology and cancer diagnostics over the next decade?
Throughout the next decade, we’ll continue to see pathologists use AI to support them in their primary diagnostic efforts. I envision pathologists will use AI on most of their workloads to make sure that the quality is high, and everything is objective, reproducible, and timely. Additionally, AI will help physicians do things they don’t currently do. It can help them decide which additional tests need to be performed on a specific case, as well as provide a more accurate prognosis and streamlined treatment selection.
AI will be integral throughout the entire patient journey, not just the cancer diagnostic part in the pathology lab, but also, for example, the oncologist who decides on the course of treatment. Also, I think AI will help combine disciplines. With time, the different modalities (pathology, radiology, genomics, clinical records) will be fed to various AI modules to support new and improved precision medicine. From a health equity perspective, patients that don’t have access to the best doctors in the world will experience a huge leap in the quality of their diagnosis and their treatment. AI will bring everyone to the level of near expert. Everyone deserves access to quality care, and AI will help bring us in the right direction to democratized health access.
Thank you for the great interview, readers who wish to learn more should visit Ibex Medical Analytics.
