This month we sat down with Body Vision’s Software Development and Algorithms Team lead, Eran Harpaz. After completing his B.Sc. in Computer Engineering at Technion and graduating Cum Laude, Eran went on to work for both Apple and Amazon. After five years in the industry, he was looking for a way to make more of an impact so he left corporate tech and joined Body Vision. It was a perfect fit—a medical startup where he could have ownership and encounter difficult and highly interesting challenges while helping to improve the lives of lung cancer patients.
As a key leader on the team developing our AI Tomography, in this blog, Eran shares his expertise on Artificial Intelligence in the industry and how Body Vision's AI tomography is changing the landscape. Hear directly from Eran about the complex process and continuous evolution that was required to create this proprietary advanced imaging technology.
Body Vision: What is the role of Artificial Intelligence today in the medical industry?
Eran Harpaz: AI can make magic happen. Sometimes you have a problem that people have tried to solve for years but couldn’t and now, with the power of AI, you can. In essence, AI unlocks a whole new range of solutions for medical procedures and applications. This is what we are doing with our AI tomography - enabling diagnostic and therapeutic options that did not exist before.
From a technical point of view, what does Artificial Intelligence mean for you?
While the basic definition of Artificial Intelligence is that a machine or computer can simulate human intelligence, it’s far more sophisticated than that. Rather, it’s learning to do a task that we, as humans, cannot do and furthermore, that we cannot teach. And it’s not that we can explain to the machine how to do it, we just give it enough examples and it can teach itself. It would take us ages to learn these things on our own, but the machine can learn it overnight.
What’re the challenges you have encountered with AI tomography on the technical side of things?
Years ago, when we first started working on our AI tomography, we were getting really strong results in simulations, but we couldn’t make it work properly in real-world data. There was a gap in the data. This was our biggest challenge to overcome: How do we build a bridge between these two worlds? It required our research and development team a long process of continuous evolution and refinement for many years, step by step, in order to get where we are today.
How do we get the data that AI creates to train our models?
We get the data to train our models in a few different ways. We have been collecting data during procedures for over seven years. We found ways to combine this data along with the data from hundreds of thousands of simulations that we generated to use for machine-learning training on powerful machines.
How do you envision our AI Tomography technology to evolve in the future?
I believe that Body Vision’s AI Tomography will remain the technology that truly sets us apart from other players in the field. After physicians use it, they won’t be able to go without it. It’s unlike anything else on the market, giving them real-time information in the format that they were used to seeing only on preoperative CT scans. Moreover, they can see soft tissue anatomical structures, but more importantly the spatial relationship of the lesion and endobronchial instrument during the procedure using only a conventional C-Arm. Doctors are so excited about our technology because it provides real-time imaging (as opposed to virtual). Moreover, CT-to-body divergence is no longer an issue because AI Tomography is not using preoperative CT. Ultimately, Body Vision is allowing physicians access to intraprocedural, CT-like imaging with only the use of a standard C-arm.