Joseph Mossel is the CEO of Ibex Medical Analytics. His career in the tech industry spans more than 20 years, starting off in software development and product management followed with leadership positions in startups, large multinational corporations and non-profits. Joseph has led products from inception all the way to maturity as multi-million-dollar businesses. He holds a MSc in computer science from Tel Aviv University, and a MSc in environmental science from VU Amsterdam.
Developed by pathologists for pathologists, Ibex is a clinical-grade, multi-tissue platform that helps pathologists detect and grade breast, prostate and gastric cancer, along with more than a hundred other clinically relevant features.
Seamlessly integrated with third party digital pathology software solutions, scanning platforms and laboratory information systems, Ibex’s AI-enabled workflows deliver automated high-quality insights that enhance patient safety, increase physician confidence and boost productivity.
What inspired you to co-found Ibex Medical Analytics (Ibex), and what problem were you aiming to solve?
Cancer, unfortunately, touches everyone–whether they are personally affected, have been a caregiver for someone with cancer, or know of someone who has been impacted. I have relatives and friends who have been affected by cancer, and tragically, one of our employees passed away from cancer.
As cancer incidence continues to rise worldwide, there is an increasing demand for cancer diagnostics that is being compounded by a global shortage of pathologists, whose jobs are becoming more complex with advances in therapy and a demand for more complex diagnostics.
Our platform helps overcome these challenges by empowering pathologists with AI tools that enhance accuracy and streamline workflows to ensure that every patient receives an accurate and timely diagnosis, which is instrumental both in guiding treatment decisions and ultimately improving patient outcomes.
We’re proud of the work we do for our customers, many of whom rely on our technology daily to deliver better diagnoses. Their trust in our solutions highlights the real impact we’re making, transforming the field of pathology, and improving patient outcomes.
Can you share a bit about your background and how it led to your work in AI-powered pathology?
If I look back at my career, there have been two driving forces: a search for a sense of purpose and a preference for interdisciplinarity over deep specialization. I am lucky to run a company that gives me a deep sense of purpose and allows me to work with an incredibly talented team from diverse backgrounds and disciplines.
My original academic background was in computer science, specializing in computational neuroscience. I then worked as an algorithms engineer and moved on to product management. After a stint at a large corporation, I decided that it was not for me. I earned a degree in environmental science and ran an environmental non-profit for several years. Sustainability remains a passion of mine and is considered the great challenge of our time.
Around ten years ago, I met my co-founder, Chaim Linhart, who was equally driven to make a meaningful difference and shared my passion for technology. Chaim, unlike me, is a specialist. He has a PhD in computer science and more than 25 years of experience in algorithm development, AI, and machine learning (ML). In the first days of Ibex, Chaim was busy winning Kaggle (ML) competitions.
When we learned that pathology is being (slowly) digitized, we talked about the impact a digital transformation in pathology could have on improving cancer diagnostics. Hundreds of companies were already developing AI in radiology, and we asked ourselves, why not do the same in pathology? It seemed like a natural fit to bring our technological expertise into the field, collaborating closely with pathologists every step of the way.
What were some of the biggest challenges you faced in the early days of Ibex, and how did you overcome them?
The idea -which we were not the first to come up with- of applying AI to pathology slides was the easy part. Execution is hard. The three main challenges we encountered within the early days of Ibex were access to data, access to capital, and access to domain-specific knowledge.
We solved the data challenge through partnering with Maccabi Health Services of Israel. At that point, we were two fledgling entrepreneurs with no medical knowledge who decided to open a medical startup in a very complex domain. Still, Varda Shalev, who headed Maccabi’s innovation arm at the time, believed in our vision, and we signed a partnership and data-sharing agreement with Maccabi. At this point, Dr. Judith Sandbank, the chief pathologist at Ibex came on board as our Chief Medical Officer (CMO), a position she still holds. With a strategic partner and a CMO, we were now well-positioned to raise a seed round, which we raised from Kamet Ventures, a French venture studio that was part of AXA Insurance.
We were now positioned to make history. We hired two engineers and developed our first algorithm for prostate cancer detection. Once we were happy with the performance, we deployed it at the Maccabi pathology lab as a second read, reviewing all of the cases after an initial read by the pathologist. To our surprise, within a few days, the system raised an alert for a case of cancer that was missed by the pathologist. As far as we know, this was the first case ever where the initial diagnosis of cancer was made by an algorithm, back in 2018.
Congratulations on receiving FDA 510(k) clearance for Ibex Prostate Detect! What does this approval mean for Ibex and the broader field of AI-powered diagnostics?
Thank you! This approval marks a significant milestone in Ibex’s journey and exemplifies our dedication to developing clinically validated solutions that help improve patient health outcomes. It affirms our commitment to the safety and efficacy of our solutions and strengthens our ability to provide cutting-edge innovation to pathologists, ultimately benefiting the patients they serve.
We envision that this tremendous milestone will break down barriers and accelerate the adoption of AI and digitization in pathology. We hope this accomplishment will bolster industry-wide confidence that the technology is easy to implement and ready for wide-scale use. Long-term, FDA clearance is an important step towards achieving reimbursement for AI in pathology and fostering widespread adoption.
The FDA validation process highlighted a 13% rate of missed cancers in initial benign diagnoses. What does this tell us about the potential of AI to improve diagnostic accuracy?
In the robust precision and clinical validation studies conducted at multiple United States and European laboratories as part of the FDA clearance, the system identified a 13% rate of missed cancers in a cohort of consecutive patients initially diagnosed as benign. This statistic reinforces the accuracy and impact of Ibex’s products, and it also validates that Ibex’s AI platform can be integrated safely into clinical workflows, enhancing diagnostic precision and ultimately improving patient care. By providing an additional layer of analysis, our technology is helping to reduce errors, enable better clinical decision-making, and promote patient safety.
As for potential, while the clearance serves as a critical validation of our technology, our solution has already been making a meaningful impact in the market. This is a testament to the daily hard work in pathology labs, and we see this as a step forward in improving health outcomes globally. We can’t help but imagine the impact this would have if labs across the United States embraced a digital transformation.
How does Ibex Prostate Detect work, and what makes it unique compared to other AI-driven pathology solutions?
Ibex Prostate Detect is an in vitro diagnostic medical device that harnesses AI to generate heatmaps identifying missed prostatic cancers. Acting as a safety net, Ibex Prostate Detect assists pathologists in ensuring that patients receive an accurate diagnosis. It leverages AI algorithms to enhance the accuracy of a prostate cancer diagnosis.
The device is intended to identify tumors that may have been missed by the pathologist. If suspicious tissue for prostate cancer is identified, the system generates an alert and includes a heatmap, directing the pathologist to areas likely to contain cancer. Ibex Prostate Detect is the only FDA-cleared solution that provides AI-powered heatmaps for all areas with a likelihood of cancer, offering full explainability to the reviewing pathologist.
Can you explain how the heatmap feature assists pathologists in identifying cancerous tissue?
Ibex Prostate Detect is intended to identify cases initially diagnosed as benign for further review by a pathologist. If it detects tissue morphology suspicious for prostate adenocarcinoma (AdC), atypical small acinar proliferation (ASAP), and other rare cancer subtypes, it provides alerts that include a heatmap of tissue areas in the whole slide images that is likely to contain cancer, offering full explainability to the reviewing pathologist.
Generally, the heatmap is accurate and precise and may provide the pathologist with areas of concern that they can focus on and determine the correct diagnosis. In the precision and clinical validation studies conducted as part of the FDA clearance, Ibex Prostate Detect’s heatmaps demonstrated extreme pixel accuracy and determined the following:
- Nearly all cancer areas are covered by the heatmap (sensitivity=98.7%).
- Almost everything highlighted as high probability of cancer in the heatmap is indeed cancer (PPV=99.6%).
- The missed cancer cases (false negatives) identified by the system were subsequently verified by expert pathologists, confirming the product’s clinical utility and benefits compared with the current standard of care.
How does the AI model differentiate between benign and malignant tissue, and how was it trained?
The Deep Learning algorithm is based on multilayered convolutional neural networks, operating on several magnification levels. The AI is exceptionally robust, demonstrating high accuracy across multiple labs and patient demographics. Of note, in line with our mantra of ‘by pathologists, for pathologists,’ the model was trained on over a million slides painstakingly annotated by world-renowned pathologists at leading medical centers. This approach is costly, but we believe that without the insight of pathologists it is very difficult to reach the level of performance we are aiming for. By doing this, we equip all pathologists with expert insights and ensure that every patient, regardless of their location, receives a level of diagnosis on par with the world’s leading specialists.
Beyond prostate cancer, Ibex is also working on solutions for breast and gastric cancers. What’s next for the company in terms of new diagnostic capabilities?
Ibex is already having a huge impact on AI-powered diagnostic solutions for breast and gastric cancers. As the worldwide leader in live clinical rollouts, many labs – including those in the United States – are already using Ibex products to transform their medical practice. Our products are proven to deliver real-world clinical impact, and pathologists both trust the AI and attest to the value it brings. Now, we are working to release a new type of technology into the market, a technology that was developed and validated by Ibex in collaboration with AstraZeneca and Daiichi Sankyo. The specific algorithm that is the first to be released helps quantify HER2 expression, which helps providers determine the course of treatment for the patient.
Looking ahead, we’ll continue to expand our offerings to provide additional insights within the tissue types we already support. We’re also looking to provide offerings within other tissue areas and continue improving our customers’ workflows.
How do you see AI-powered pathology evolving in the next five to ten years?
I envision that AI will have a profound impact on the practice of pathology and the way cancer is diagnosed. I don’t see us replacing pathologists, but as with every new technological development, the practice will be transformed. AI will continue to be instrumental in addressing the growing workforce challenges in healthcare, particularly the global shortage of pathologists and their increasing caseloads driven by rising cancer cases. Implementing responsible AI will help pathologists manage their workloads more effectively, improving diagnostic efficiency and reducing delays. By automating routine tasks, AI can lower error rates, improve the quality of diagnosis, and ultimately boost pathologists’ confidence in their work. I strongly feel that AI, together with a human in the loop, is the best combination for transforming healthcare.
Another area with great promise is expanding beyond the current practice of pathology into the realm of predictive algorithms. Algorithms that potentially combine several modalities to predict outcomes or, crucially, treatment efficacy.
AI can also enhance health equity through democratized health access. Regardless of location, every patient, everywhere deserves a trusted diagnosis. It would be great for AI technology to be deployed as part of standard practice in every pathology lab worldwide. However, this starts with collaboration among physicians, the industry, and agencies to accelerate the deployment of this technology–I feel we owe it to patients.
Thank you for the great interview, readers who wish to learn more should visit Ibex Medical Analytics.