#127 Reinventing Prostate Cancer Testing with AI: From Development to Regulation to Production with Elliot Smith – CEO & Founder
Elliot Smith is the CEO and founder of Maxwell Plus. A visionary company that has been using machine learning and artificial intelligence to help treat and diagnose patients with prostate cancer. The way they bring in the data, help analyze it and help people along their journey is fascinating and makes them leaders in this specific field.
In this episode, Elliot recounts how after getting a PhD in Biomedical Engineering, he realized that, as much as he loved academia, he wanted something with a faster pace. With his knowledge in numerical modelling, optimization and how that connected with AI and medical imaging, he decided to take another shot at building a start-up. That’s how Maxwell Plus was born.
Elliot is very passionate about helping others find diseases at a point where they can be treated. To know where things are, if they should be treated and how best to treat them is his ultimate goal in terms of applying data science and AI to the medical world.
His is a very inspirational entrepreneurship story and the way he has managed to grow his company for the past four and a half years is admirable.
Stay tuned to know how his desire to build something that could be useful in the real world turned into a company with the potential of impacting the lives of over 300 million men that have or are being diagnosed with prostate cancer.
Enjoy the show!
We speak about:
[0:35] How did you get into the world of data and AI?
[06:06] How did you decide to start a company?
[09:08] Can you explain to us the complexities of the problem your company tries to solve?
[13:18] How did you go through the process of gathering the necessary data?
[17:00] Did you find greater success when approaching doctors, medical centers, or hospitals, etc… when trying to gather the data?
[20:11] How did you decide to focus in prostate cancer?
[23:41] What does the company look like now?
[25:55] Tell me about the approval process for Maxwell Plus as a medical device.
[30:25] Regarding the approval process, why did you look into being registered in multiple geographical areas where there are separate processes and was that something you did for the company to have long-term success?
[32:35] When you started the company, did you know that you would have to be approved as a medical device?
[34:59] How beneficial do you think it was for you to get that prior experience before starting Maxwell Plus?
[38:50] Do you have any particular things that helped you change your mindset?
[41:45] Why did you have to be approved as a medical device if you are providing software?
[42:35] Tell me more about the AI side and how it works.
[48:53] How standardized are the outputs of the MRIs that you get from different machines or different hospitals?
[50:13] What effect, if any, will My Health Records have on your company?
[54:00] What’s coming up next for Maxwell Plus?
Resources:
Elliot’s LinkedIn: https://www.linkedin.com/in/smitec/
Elliot’s Twitter: https://twitter.com/elliot_c_smith
Maxwell Plus on LinkedIn: https://www.linkedin.com/company/maxwell-plus/
Quotes:
Had an opportunity to participate in some research at the end of my undergraduate degree, looking at kids that have cystic fibrosis. So cystic fibrosis is a quite rare disease that affects a number of different systems in the body, but one of them in particular is it creates a thick mucus in the lungs and if that mucus doesn't get cleared out it can get infected. So there is a lot of respiratory physiotherapy, so breathing exercises, to get that mucus out of the lungs. The problem we were trying to tackle was that kids, especially kids under 10 years old, they hated the physiotherapy, as you can imagine. It’s like an hour a day kind of stuff and we were working on a project to build a Bluetooth connected physiotherapy device that allowed them to play a game on their phone, as they were doing their physio, so their physio would be the controller for that game and that was a really interesting, really wonderful project. That was kind of the first time I was looking at collecting a large database of medical data and trying to do something useful with it. I was trying to track the performance and trying to predict the infection rate on these kids. So that project went well, we ran some trials, built a few devices, but ultimately it turned out too expensive to commercialise, but it was a great exposure to the world of medical data and data science in general.
When I finished that I sort of came to the realisation that I loved academia, but it was a little bit too slow moving for me. I wanted to be out there building things, putting stuff out into the world and decided to have another run at building a start-up and at the time I happened to know a lot about numerical modelling, optimisation and how that translated into things like AI and also medical imaging and without jumping too far ahead that's the point where all of this came together and the company that I now run Maxwell plus was born in that collision of worlds.
It was an interesting one I think with any start-up origin story there is like that healthy mix of serendipity and I guess ignorance or arrogance, whichever way you want to look at it. So the story we tell and it's how it happened was that I was having drinks with a friend of mine, Steve Backster, a local Brisbane angel investor and I was talking to him about my PHD and what we were doing in image guided radiation therapy. He said to me, “you know I wish we had a way to lay down, get checked and know from head to toe what was going on and then be able to do something about it”. This is where the arrogance comes in, three beers into this conversation, I turned to Steve and I said I could build you one of those.
As I said early days, we were looking at any kind of whatever data we could get our hands on, we were small and excited and just wanted to build AI. But probably 1 to 2 months into really getting serious with the project that became the business, we met a local clinician here in Brisbane named Dr Peter Swindle. He is a local urologist. We were looking for clinicians to work with and it happened that his passion and interest was MRI of the prostate. He taught radiologists how to read MRI and he really just loved this idea of building technology to help make that process better. As a urologist they do surgery as well as diagnosis. You see a lot of people with that same problem coming in too late and having to have those conversations, I wish you had got here earlier!
We looked at their data, we saw something in their data that wouldn't have been picked up otherwise and our doctors said we need to act and with the hindsight of knowing their outcomes we acted at the right time and those people have what is for prostate cancer is 98% 5 year survival rate because we acted in time to avoid that conversation.
I think it was very beneficial. I think you're absolutely right. There is learning from experience and there is learning from, I guess, other people’s experience. Two things that I think when I was first time founder in that original product; I thought I could work it out all by myself and you know people out there can but it's going to take you 3 times longer than if you find somebody that knows what they are doing, ask them for help and learn it.
So, there are definitely differences. Which can be good and bad. So you know for us we have aimed to collect international, multi institution, multi reader widespread demographic dataset. Because, and the reason that is bad I should say is it takes a long time, it’s really hard to get a diversity in data because we have seen that in the medical world some AI models just trained on, just 1 or 2 institutions worth of data end up becoming great predictors of not the underlying disease state but which institution the data came from. If you know that this one is a late stage and this one is an early stage centre, you can get pretty good performance just by making that guess. So, for us it’s really making sure that we're not biasing our dataset to any one of those kinds of factors.
If you have had treatment or you choose to go into what is called active surveillance, which is let’s just watch this and determine when to treat, the algorithm that we apply to do initial diagnosis apply just as well to monitoring your diagnosis over time. How can we then go to provide ongoing support to really see these men through from day one to day infinity and continue to provide real clinical value to these men.
We are now on YouTube! Watch the episode here: https://youtu.be/QqblPG98krU
And as always, we appreciate your Reviews, Follows, Likes, Shares and Ratings. It really helps new data scientists find us.
Thank you so much, and enjoy the show!