Video Transcription
Russ Yoshinaka 00:00
Russ, good afternoon. My name is Russ Yoshinaka. I'm the co-founder and CEO of Ceevra. We help surgeons operate better through advanced surgical visualization. So I'd like you to imagine for a minute a pretty somber thought: imagine that your spouse has been diagnosed with early-stage lung cancer. Now, this is a particularly bad prognosis. As you may know, lung cancer is the most deadly form of cancer, killing more people each year than the next three types of cancers combined. Now, most surgeons out there will be planning your spouse's operations using a CT scan. It looks like this. They're using this to plan the operation, the best type of operation to perform, the best way to perform it. This is a series of 2D slices that the surgeon has to scroll through to try to create the mental 3D reconstruction of your spouse's lung anatomy. And they're making literally life-and-death decisions based on this technology, which was developed in 1974 and really hasn't advanced much since then. Now imagine that your spouse's surgeon had this to plan the operation, and the surgeon using this technology, using a Ceevra model, which is generated from that same CT scan that you just looked at, can selectively show and hide each anatomical structure to really zero in on understanding the location of the tumor, where it sits relative to segmental boundaries and other structures, such as blood vessels and airways, and can plan a very, very precise operation to help them decide what's the best way to get this tumor out while taking out the least amount of lung tissue possible. Well, this is what we do. We use machine learning algorithms to transform standard CT scans and MRIs into 3D digital models, which surgeons use for preoperative planning, patient counseling, and intraoperative decision-making. We have supported over 3,000 surgeries to date, mostly in the areas of lung cancer, kidney cancer, and prostate cancer. We are FDA cleared and CE marked under MDR to support any type of surgery, and we have completed two multi-site randomized control trials demonstrating significant patient outcome benefits. What our 3D models are used for is planning, and one of those was actually published yesterday, and I'm excited to tell you about that today. So a bit on our architecture, our workflow. We are cloud-native and mobile-first, so all the processing is done in the cloud. We make it very, very easy for surgeons and hospitals to implement and use our product. They simply push the CT scans or MRIs directly from the PACS system to our cloud, where we process it into the 3D digital models through a combination of our software and our in-house imaging technicians. When the 3D model is done, the surgeons can view them from their phone, from a tablet, or from any computer. They can bring their phones into the operating room, connect them into the back of the surgical robot, so the surgeons can see them inside the surgical robot in a side-by-side view, or they can plug it into any sort of OR monitor, vision tower, or ablative system. Because of our cloud-based architecture, we make it very, very easy to integrate with larger intraoperative surgery platforms. So here's again, we are FDA cleared and CE marked for any type of surgery. Here are the main types of surgeries that we've been supporting today. So I wanted to do a little bit of a deep dive on our machine learning. So when we originally started the company, we were using non-machine learning algorithms to do the process, what's called image segmentation, which is essentially taking a CT image and creating a 3D representation of that. Over time, we amassed a database of enough data to develop machine learning algorithms to do this automation. The reason that using machine learning is really critical is because it can be trained to be more accurate. And the cool thing about machine learning, as we all know, is the more data that you feed it, the more accurate it gets. The reason that accuracy is important is because it allows us to generate more 3D models in the same amount of time with the same level of personnel. So over time, it will drive down unit costs, increase throughput, increase gross margins, and within a few years, we expect to have gross margins in the range of 85% or higher. We also have specific expertise in this area. Last year, we became the first company to be cleared by the FDA to use machine learning algorithms in generating 3D models for urologic and thoracic surgical planning. This database that we have amassed is actually unique, proprietary, and valuable. It's actually closer to 3,000 images now. We can use this database for ongoing machine learning development, and we're doing that in two areas. We're doing that to automate more and more surgeries, going beyond urologic and thoracic surgical planning, and we're using it to develop intraoperative augmented reality. This is what surgeons are doing today in the operating room. This is a view of a connection of a phone into the back of the Da Vinci console, where they can see the live endoscopic view of the patient anatomy right next to the Ceevra 3D model. What we are developing is intraoperative augmented reality, where the 3D model will be overlaid on the live view of the endoscopic anatomy, allowing the surgeon to have a virtual view of what lies beneath the operating field, essentially giving them X-ray vision. There are a number of companies that are working on this; we think that we're very, very well positioned to actually bring this to market, to put it into the hands of surgeons because of our team's expertise in machine learning, our proven experience with regulatory authorities, and because of this database I mentioned earlier, which is constantly growing. Anytime we generate a 3D model for one of our customers, it goes back into our database. So just briefly on our business model, we have a SaaS recurring revenue business model. We sell annual subscriptions that are based on the number of expected 3D models over the course of a year. Here are logos of some of our customers, which include notable academic medical centers such as Memorial Sloan Kettering and UCLA. We also have as customers two of the largest IDNs in the United States, Advent Health and Providence. The US government is also a customer. Increasingly, we have industry partners who are buying our 3D models and incorporating them into their surgical platforms. That's going to be a big part of our go-to-market strategy going forward because, whether it's a surgical robotic system, an ablation system, or a laparoscopic system, again, because of the way we've architected our solution, it's very easy for us to integrate into these larger third-party platforms. Okay, so the last subject I wanted to touch on is clinical evidence. There is a lot of great technology out there, but where the rubber really hits the road is: does that technology change clinical decision-making, and does it improve patient outcomes? We've completed two multi-site randomized control trials that prove yes to both of those questions. In our studies, we've shown consistently that when surgeons use our models for planning, they change their preoperative plans about a third of the time. We did a randomized study in partial nephrectomy, which is kidney cancer, and we showed that when our 3D models are used for surgical planning, as compared to just standard CT or MRI, the surgeon is far more likely to achieve a strong outcome as it relates to the metrics of operative time, blood loss, clamp time, and patient length of stay. Now onto the study that was published yesterday. This is prostate cancer again. Patients were randomized to a control arm, which only had the MRI for planning, and an intervention arm, which also had the Ceevra 3D model. As you can see, large teaching centers and high-caliber research partners published yesterday. We measured a number of endpoints. The most important one is cancer control. Did the surgeon get out all the cancer? Our endpoint was measured by detectable rates of detectable PSA after 18 to 24 months. In the control arm, as you can see, 18% of the patients had detectable PSA. In the intervention arm with the Ceevra model, zero. Adjuvant therapy was another metric we looked at. The patients in the control arm had almost a third of those patients go on to some sort of radiation, androgen deprivation, or some sort of follow-on therapy, whereas the patients with the model had only 3% of those patients needing some form of follow-on therapy. We then looked at functional outcomes. Functional outcomes in a prostate cancer case are urinary continence and erectile function. Both groups returned to pretty strong functional outcomes, so no significant difference related to urinary continence. But then when we looked at erectile function, we once again saw a very significant difference. As you can see, the preoperative SHIM scores for both arms were about 18.5, close to 19. The control group, who took the same 25-point survey (higher number being better), got back to 9.8 after 24 months. Our intervention group got back almost to complete continence. Lastly, Trifecta outcomes, which take everything and put it all together. A Trifecta outcome is essentially cancer-free (no detectable PSA), with no adjuvant therapy, full return to continence, and return to erectile function, which was in the same category as the preoperative SHIM score. We saw that the patients who had a 3D model had almost a 5x rate of receiving this Trifecta outcome. In closing, I'll leave you with two thoughts that we firmly believe. One is that 3D imaging is the future of surgical planning. We believe it's not a matter of if, but when.
Russ Yoshinaka 09:55
Second, we firmly believe that 3D imaging is an essential component of intraoperative augmented reality, which will represent a quantum leap in surgical planning. If you share either of those beliefs, I'd love to talk to you. Here's my contact information, or you can message me through the LSI app. Thank you. Applause.
