Antony Rix, Lucida Medical - Pi™ Prostate AI with Expert Level Performance | LSI Europe '24

Antony Rix 00:02
My name is Dr. Antony Rix, and I'd like to present our work on cancer imaging today. What inspires me every day is the challenge that we have: about half of all cancer patients are diagnosed late, at stage three or stage four. There are many reasons behind this; cancer diagnosis is very inexact. The tests are costly, they're slow, and they can have a devastating impact, affecting millions of patients each year. What we've been doing is developing AI-based software that analyzes medical imaging and clinical data to accurately help us rule cancer in or out, but also to localize it precisely for biopsy, staging, and treatment. Our initial focus and traction is in prostate cancer, the most common male cancer, with one and a half million diagnosed and 400,000 dying each year. Our predictions are that by 2030, 16 million men a year will be undergoing prostate diagnostic or monitoring work.

So who are we? I've been lucky to build an absolutely fantastic team. I bring 25 years of experience in artificial intelligence, including a previous startup, a PhD, and my first exit in this space, Cytechnics, and then I had a further exit at TTP. My co-founder, Professor Evis Saleh, is one of the world's top cancer imaging radiologists. She's now chair of Radiology at Gemelli in Rome, Italy's top hospital, and we came together with this shared vision to use artificial intelligence and medical imaging, combining our forces to target cancer. Supporting us is a tremendous team. David Tuck, our chair, is a serial entrepreneur in this space and has recently exited his previous startup, Lightpoint Medical, to TLIX for a nice multiple. Marc Hinton is a serial startup Chief Technology Officer and builder who previously spent 10 years working at a medical imaging CRO. Finally, our medical advisory board includes some of the people who are writing the practice guidelines today, such as Professor Anwar Padhani, Dr. Francesco Gigante, Professor Jochen Wolt, and others, working with Professor Saleh. Our shareholders include two large cancer charities based in the UK, as well as eight leading clinicians in our space.

The software platform that we've developed finds cancers like this that doctors missed. Let me tell you about this patient, and this is what makes me passionate about this work. This man had his prostate imaged using MRI, and the radiologist didn't report anything in this area. What you're seeing here are the outputs from our software. This is a really common occurrence. Unfortunately, there's about a 30% recurrence after surgery in prostate cancer as a result; part of this tumor was left behind, effectively leaving this guy at stage three as a result of that surgical procedure. You can see that our software is highlighting this tumor here. If that software had been used before this procedure, this guy could have had curative treatment then and there, the same day. And this is the potential of AI in medical imaging and cancer diagnosis.

For it to work, we need to do four key things. AI has got to be automated; in a sense, that is obvious, but a previous generation of artificial intelligence algorithms has generally required a lot of human input. We've been investing heavily in automation. It's got to be scalable. This is something that needs to run on any old hardware; we can't rely on farms of GPUs. In fact, our software runs on my laptop. Historically, cancer diagnostics have been either sensitive or specific, but we need both. And finally, we need to be able to do localization for staging and treatment. Purely detecting cancer is not enough.

So let's show with some data what this means. The best-in-class blood test diagnostic for clinically significant prostate cancer in a diagnostic population is represented by this marker as a comparison to an expert radiologist against extensive biopsy ground truth, also in a diagnostic population. That diamond shows around 50% sensitivity and about 85-90% specificity, with around a 50% false positive rate. The blue line plots the standalone accuracy of our AI software, which, as you can see, is certainly good in comparison to these benchmarks. And that's the product that we now have in the market today. Finally, if we layer in additional clinical data, the research that we're presenting at the Radiological Society of North America at the beginning of December, we can start to achieve completely unprecedented levels of accuracy in prostate cancer, hitting that crucial point of both 90% sensitivity and specificity.

Because of this, we can offer really valuable benefits that touch the clinicians who treat the patients, the radiologists, payers, and providers. Obviously, a productivity increase: AI speeds the whole process up and allows you to process more cases, referring up to two times more significant cancer detection compared to a standard of care trust biopsy. Looking further into the future, to tackle that problem of late diagnosis, we need to screen more—that's up to a five times volume increase that will rely on AI like this to deliver. Finally, there's a clear pathway for us to achieve premium reimbursement that's been pioneered by companies like Perspectum, HeartFlow, and many others in the space. So we know how we can get this reimbursed in both the US and in Europe.

The PI platform itself is used by radiologists like this through their standard imaging equipment. It enables them to review cases, comment, and report quickly and accurately, and it's underpinned by six years of investment that we've made in intellectual property, which spans data, significant know-how in the artificial intelligence algorithms, and how we train them. We have one patent already granted, with further in the pipeline. Finally, we've invested very heavily also in the software for automation and quality assurance because we realize that in the future, this is going to be crucial. Because of this outstanding performance, PI is now the go-to solution, both for vendor partners and for hospitals. Somerset Foundation Trust: this is Dr. Paul Bern on the left of the picture there, one of our first users in the UK. They're writing and presenting at conferences about how the software has high positive predictive value. So they're using it for triaging patients and also highlighting the ability to speed up the process, both the time that a radiologist spends and the time the patient spends waiting for a diagnosis.

A short note on the market: we all appreciate cancer is a huge market. It will be valued at $165 billion by 2030 in this space of cancer diagnosis that we play in. Our initial focus and our current certification is in prostate cancer itself. That will be a $13 billion market, with 400,000 deaths worldwide in 2022, and the burden is set to double over the next 15 or 16 years. We've already proven this platform in bone, liver, and lung cancers, using whole body MRI. Over time, we'll build this pipeline of clinical indications to further extend our value.

So really, to summarize, with PI, we've developed a unique technology platform for cancer diagnosis that stands out in detecting cancers that others miss. Its expert-level performance is opening multi-billion dollar applications for precise screening, diagnosis, and treatment. We're now at the commercialization stage with a part; we have key vendor partners like Sectra and GE, and the technology is now deployed in seven countries. We're seeking investment to unlock substantial value through completing FDA approval and growing sales, building on the traction that I already talked about. Please come and see me or get in touch with us to find out more. Thank you very much. Applause.