Why medical AI is taking off in SE Asia

Today's episode is about trust; trust in technology and trust in each other.

Very few startups experience what LPixel went through and far fewer survive it.

Today we welcome Yuki Shimahara, founder of LPixel, back to the show. The last few years have been a roller-coster for LPixel, and despite the chaos LPixel managed to created Japan's first certified medical AI device and roll it out into hospitals around the country.

And despite his success in Japan, Yuki also explains why smart medical AI startups are all looking to Southeast Asia.

It's a great conversation, and I think you'll enjoy it.

Show Notes

How LPixel was certified as Japan's first AI medical device The transition from diagnostic support to full medical diagnosis Why it's not technology holding back medical AI The nature of trust in Japanese business Japanese health insurance is now paying for AI diagnosis What happens when an employee steals all your funds? The advantages (and disadvantages) of full transparency How investors reacted and their new demands Why more doctors are founding startups Why research is easier at startups than at universities Why developing countries will see more advances in medical AI than the developed world Going global does not mean going to the US (yet) How the Japanese government should (and should not) foster Japanese innovation

Links from the Founder

Everything you ever wanted to know about LPixel LPixel's medical diagnostic support system Eirl

Transcript Welcome to Disrupting Japan, straight talk from Japan’s most successful entrepreneurs. I’m Tim Romero and thanks for joining me. Japan is often described as a high trust society, but it's hard to explain exactly what that means and why it matters. Well, today we sit down to talk about trust and about medical AI with Yuki Shimahara, CEO of LPixel. Now, a lot has changed since Yuki was on the show four years ago. And by all metrics, LPixel is a stronger and more successful startup today. But one unfortunate event really put that level of trust to the test. Well, Yuki will give you the details, but the level of trust that existed between investors and clients and employees resulted in saving a startup that no one could reasonably expect to be saved. And we also talk about why medical AI is going to be adopted so much faster in Southeast Asia, why more and more doctors are starting startups in Japan and why Yuki thinks it's more productive to do deep research at a startup than at a university. But you know, Yuki tells that story much better than I can. So, let's get right to the interview.

Interview Tim: I'm sitting here with Yuki Shimahara, the CEO of LPixel. So, welcome back to the show. Yuki: Yeah, thank you for having me. Tim: LPixel a cloud-based AI image analysis for life sciences and medical research. And well, you can probably explain it much better than I can. Yuki: I'm very honored to be back here. LPixel is a startup company from Research Lab of Tokyo University, which is a pioneer bio image informatics. We combine life science and imagine analysis including AI, but also we do are the two main business. So, we developed the AI for medical misdiagnosis and then developing AI for accelerating the pharma research. Tim: And wow. Last time we talked, I think you were still a PhD candidate at that point. Yuki: Probably. Tim: Yeah. Because I do remember we were running around into different rooms at the University of Tokyo campus at Hongo trying to find a room that didn't like echo. So, much has changed since then. You're a lot bigger and more successful than before. So, how many people do you have working at LPixel now? Yuki: Now, 60 or 70. Tim: Tell me about your customers. So, last time most of your customers were research institutions, people working on medical research and it seems like you've expanded a lot since then. Yuki: I think the last time is just developing the AI but three years ago we got the AI as a medical devices. It is said that it is a fast AI got approval as medical devices using deep running. So, our stage changed the last time. Maybe I focusing on R&D, but we do the sales and marketing now. Tim: And just to clarify, so when you say it's a medical device, does that mean it is a diagnostic tool? What is the certification of being a medical device mean? Yuki: Our AI is supporting the diagnosis, the medical doctor even the professional mistakes sometimes. So, misdiagnosis is very big problem for patient. The AI support the diagnosis. For example, by using our AI, the accuracy increase 10%. So, we can claim that it is good for not misdiagnosis and improve accuracy. Tim: Okay. So, just let's walk through a typical example. So, LPixel technology works on, for example, imaging from a CT scan. And it would analyze the image. And what would it suggest to the doctor that there's a high chance of this disease? How does it help the doctor? Yuki: So, very simple AI highlight the candidate of the diseases. It is like mis-spelling check using something. Tim: I like that analogy. Yuki: Very simple. It's just highlight. Tim: When you say highlight, does it, for example, does it show an area of the scan that needs special attention? Yuki: Yes, exactly. The detection is the most important part I think. But the next we do the classify the diseases and suggest the treatment or something. But for now focus on detection. Tim: So, it's not yet a diagnostic tool. It's a diagnosis support tool. Yuki: Right, exactly. Tim: That's a subtle but very important distinction. How many years away are we from having AI being able to actually diagnose? Yuki: For now, legally AI cannot diagnose, it is just supporting tool. But little by little we improve the AI. Probably next step is we change the double check system. The US are the same. So, US and Japan, the government recommend double check because human misdiagnosed sometimes. So, we can change human plus human to human plus AI. So, it is the next step. Tim: So, but not yet AI plus AI? Yuki: No, not yet. Tim: Is the biggest challenge one of technology or is it one of perception? Yuki: Good question. So, because the technology is very important, but I think that society development is very important. So, we show that evidence human plus AI is better or same as human plus human. And then so we do lobby activity with public sector or academic society. I think it takes seven or eight years, but it is very soon I think. Tim: When both technology and society get to the point where AI plus AI counts as the second opinion. I think that's going to transform medicine completely. Yuki: Right. Tim: And eight years sounds like it's far away, but I mean it was only what, five years ago you were on the show. It's not that far away. Let's talk a bit about your go to market. So, last time we talked you were just starting to work with the manufacturers of CT machines, MRI scanners. How has that played out? Yuki: I think the five years ago, so we do joint the research business mainly, but from five or six years ago we had just started product development. Three years ago we got approval. So, our policy is vendor neutral. So, investors like Cannon Medical and FujiFilm and Olympus and also GE and Siemens, not investors. But our policy is vendor neutral. Tim: I think it is interesting because well one of the things I want to talk about is how trust works in Japan, the importance of trust in Japan. And how the level of trust enables different kinds of relationships between startups and enterprises and startups and the investors. You've had research relationships with these companies for many years. And last time we talked, you were saying that you weren't that concerned about having a direct relationship with the doctors, which as an American founder I found like shocking and like kind of made me nervous. But has that strategy been successful? Have the partners, the OEMs supported you and helped you get to market without those direct relationships? Yuki: Good questions. I think that timing is good. So, we was first paying in this market and then medical doctors present our AI or good AI is coming. So, medical doctor trust medical doctor sales, right? Tim: Yes. Yuki: Right. So, academic presence is very important. So, health insurance coverage is very important. So, this fiscal year government support, insurance coverage for AI, it is very limited, but it is good trend. Tim: That's very interesting. And that's very new. When you say insurance is paying for AI, what does that mean? Yuki: For example, if medical doctor diagnose MRI or CD images, they can get 3000 yen a year per scan and then from this year using AI 400 yen. Tim: And so until this year, hospitals couldn't even bill for it? Yuki: Right. But it is very limited, but it is good first step. Tim: Well I think that's an amazing first step, especially since it's like one-tenth the price. Yuki: Yes it is. Good change. Tim: The other thing I really want to talk about is what you and LPixel went through back in 2020. Well, actually why don't you share what happened? Yuki: So, it is very hard things for us. So, two and half years ago, one of our members director and embezzled about 25 million. So, we lost all money and then cannot pay the salary. Tim: Yeah. Like $25 million and it was gone and not recoverable because he spent it on like FX trading or something like that. Yuki: Exactly. Yeah. Tim: And that was all the money your company had? So, what's going through your mind at this time? Yuki: Yeah, so it is a tough situation. The first thing is we explain to the investors and very sorry, so why it happens or like that. And then this is a new brand, so that's why we need this kind of resources. So, very tough communication, few months. Tim: What about your employees? Because I think like customers and investors you can have more rational conversation.