The medical industry is one of the most challenging areas for startups to succeed in.

“Move fast and break things” just doesn’t work in medicine.  So you might be surprised to learn that right now there are quite a few innovative medical startups coming out of Japan.

Today we talk with Yuichi Tamura, founder of Cardio Intelligence, who has developed Smart Robin, an AI platform that reads EKGs, has been certified as a diagnostic device, and is being used in clinics and hospitals all over Japan.

We talk about the challenges of bringing medical AI to market, their plans for global expansion, and the most important thing that venture capital can offer medTech startups.

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

Show Notes

  • The importance and challenge of the current EKG-reading workflow
  • Why is is so hard to bring a new medical innovations to market
  • Yuichi’s transition from medicine to business.
  • A go-to-market strategy for medical startups
  • How Cardio Intelligence acquired enough EKG training data
  • Why automatic EKG diagnostic innovation stopped in the 1970s
  • The importance of explainable AI for medical devices and diagnostics
  • The role startups need to play in medical innovation in Japan
  • What venture capital firms can really contribute to medTech startups (besides the capital)

Links from the Founder

Transcript

Welcome to Disrupting Japan. Straight Talk from Japan’s most successful entrepreneurs.

I’m Tim Romero and thanks for joining me.

Genuinely new medical technology is one of the most difficult things for a startup to bring to market. Regulations are complex and capital needs are high, and yet Japan has a surprisingly large cluster of innovative medical startups who have new technology both approved for and actually in use in clinical practice.

There are a number of reasons for this, and today we sit down with Yuichi Tamura, MD and founder of Cardio Intelligence, a startup using AI to read EKGs and detect atrial fibrillation.

It’s a technology that not only makes work faster, but it opens up a whole new range of important inexpensive diagnostic tests that were simply impractical before. It’s AI technology that is doing genuine good.

Yuichi and I dive deeply into that, and we also talk about how AI is going to change the face of telemedicine and rural hospitals. Why EKG innovation stopped in the seventies and exactly when technical founders need to step out of the CEO role.

But, you know, Yuichi tells that story much better than I can. So, let’s get right to the interview.

Interview

Tim: So, I’m sitting here with Yuichi Tamura, the founder and CEO of Cardio Intelligence and maker of Smart Robin, who’s using AI to detect atrial fibrillations from EKGs. Thanks for sitting down with me today.

Yuichi: Sure, my pleasure.

Tim: Well, I gave a really brief introduction to what Cardio Intelligence does, and I’m sure you can explain it much better than I can. So, what do you guys do?

Yuichi: So, Cardio Intelligence provides the AI medical software, which enables physicians and the technicians to lead the long-term electrocardiogram more easily.

Tim: And you’re focused on detecting atrial fibrillation. So, what exactly is atrial fibrillation and why is it bad?

Yuichi: Atrial fibrillation is a very, very big problem for cardiac health. It brings not only heart failure, but also brain stroke because an atrial fibrillation make a paralyzing the atrium, the upper chamber in the heart, which brings some very, very small thrombin. And finally, it drives into the brain arteries which brings a brain stroke. So, in such a case, the patient suffer from very severe symptom, half of the body paralyzing and sometimes make sudden death.

Tim: And from what I understand it, it’s very hard to detect the risk of atrial fibrillation when someone’s not having a seizure or not having an incident.

Yuichi: Yeah. And some patient feel palpation with a very regular or very high beat with a heartbeat, but half of the patient does not feel anything.

Tim: And so Smart Robin can read EKG output, just normal EKG output where they’re not having a seizure and not feeling palpitations and predict when atrial fibrillation is likely.

Yuichi: Yeah. Very important thing is, some half of the patients suffer from atrial fibrillation attack, but this kind of an attack occurs some patients suffer from once a day, some patients from the once a week, some patient once two weeks. So, it’s very important to make a long term electrocardiogram to detect the atrial fibrillation attack.

Tim: And is Smart Robin certified as a diagnostic device as well?

Yuichi: Yes.

Tim: So, tell me about your customers. Who’s using Smart Robin? Is it clinics, hospitals, individual doctors?

Yuichi: Yeah. Many doctors and technicians use Smart Robin to make a final diagnosis to find an atrial fibrillation.

Tim: So, really briefly, let’s walk through the workflow in a clinical setting. How does Smart Robin change the doctor’s workflow?

Yuichi: So, please remember your health check. During the health check timing, you performed an electrocardiogram, but it’s covered only five to 10 seconds. So, in case the patient suffers from wrong term atrial fibrillation attack, we can capture during the very, very tiny seconds. However, half of the patients suffer from attacks by-weekly or by day. So, in such a case we cannot capture just a few seconds. So, in such a case, we put the patient the long term electrocardiogram, which called a patch type ECG, or H type ECG. But it is very, very hard burden to read in every waves more than one day, more than one week.

Tim: Oh, I see. So, the patients can have a wearable EKG, generate a week, two weeks of data, and instead of a human specialist reviewing it, you can have an AI review it and say, oh, at three o’clock last Friday you had this problem.

Yuichi: In these days, the physicians and the technicians by themselves manually read the electro cardio by themselves. And there is not so much confident support within software. So, it’s very, very hard work.

Tim: Right, and expensive work too. And so Smart Robin, it’s an actual diagnostic device, it won’t just say, oh, by the way, this looks interesting. It’ll say, this is fibrillation.

Yuichi: Yeah. From our data, the technicians usually read the 24 hours ECG in 30 minutes to one hour in general. But after taking the Smart Robin take them only three minutes.

Tim: Right. I want to get back to a deep dive on both your go-to market and how it’s used in a clinical setting. But before that, I want to talk a little about you. So, you graduated from KO Medical School back in 2004 and started teaching and practicing at International University of Health and Welfare. So, what made you decide to start a startup after more than a decade of teaching and practice?

Yuichi: Yeah. Initially we start to make a basic techniques to make an AI diagnosis for the long term electrocardiogram and start with my friend and start with my CTO. And after successful for the manufacturing we want to make an some relationship with device companies, but it is very difficult to make new software device in the market.

Tim: Why is that? Is this just the medical industry is very conservative or it’s just very difficult for startups to produce these devices?

Yuichi: Yeah. They do not want to make new softwares because they cannot understand their meaning for the AI devices at the time with five or six years ago.

Tim: Wow. That’s certainly changed, hasn’t it? Everyone is talking about AI now. So, are you still teaching?

Yuichi: Yes. And I make a clinical practice in the hospital.

Tim: So, do you ever think you’ll completely leave academia and stop practice and focus on just being a CEO?

Yuichi: Yeah. Sometimes I consider with it, but it is very important to see this patient by myself and to see the clinical situation because every pain and every question always in the clinical situations.

Tim: Well, I could see that from two points. I mean, one, you need to be really close to understand the needs of the doctors. But also on the sales side, I mean, doctors only listen to other doctors.

Yuichi: Yeah. And I actually use a Smart Robin by myself in the hospital so I can identify the problems and goals and the target to make a Smart Robin more efficient.

Tim: So, as Cardio Intelligence grows, do you see you stepping back into a more product leadership and market leadership role and hiring a full-time CEO?

Yuichi: Yeah, and actually I work as a CEO to make a more good market. And because then we launched and Smart Robin for more than two years and the number of the users getting more and more, so.

Tim: One of the biggest challenges that deep tech startups in Japan have is making that leadership change. It’s a hard thing to do.

Yuichi: Every situation is done faster for me. It’s like a resident for the doctors.

Tim: Yes. And very similar hours too. It’s very hard workload.

Yuichi: Yeah. Yeah.

Tim: Well let’s talk a bit about your go-to market and the future of commercializing this technology. So, the technology’s proven in clinical trials. It’s certified as a diagnostic device. You’ve got some initial use in some hospitals. What’s the next step in commercialization here?

Yuichi: One thing is to make a global market. In the United States, the burden of the clinical situation with an atrial fibrillation is almost the same, maybe more big problem compared with Japan.

Tim: Japan. Well, I’d imagine the needs are the same everywhere. I mean, we all have the same body.

Yuichi: Yeah, yeah, exactly.

Tim: But are the regulations different to the certification process, are they similar? Are they really different?

Yuichi: Yeah, it’s similar. And we can use the same techniques. We make AIs for the universal waveforms. So, it’s very important to make globalization.

Tim: And in expanding the market. Are you partnering with equipment makers who, even though they’re conservative, everyone’s interested in AI now? Or are you going direct to doctors? What’s your strategy?

Yuichi: Yeah, it’s very important to make a partnership with device manufacturers, especially in the United States because then our software can be used in every devices. So, if there’s a good manufacturer for the devices exist in the United States or other countries like in India or Europe, they can use their own devices. So, it’s very good model to make a partnership.

Tim: And you mentioned when you started Cardio Intelligence back in 2019, the medical community was not taking AI seriously and not interested in partnering. I assume that’s changed a lot now, right?

Yuichi: Yeah. During the few years.

Tim: Okay. So, a lot of partnership opportunities.

Yuichi: Yes. Excellent. Especially in Japan.

Tim: In training this AI, how did you get the training data? You must have needed a huge amount of raw EKG data.

Yuichi: Yeah. We acquired from the university hospital. I worked in the — I still working the university hospital. And we also obtained the teaching data for the each brain forms.

Tim: Oh, okay because I was going to ask, because you’d also need to have this data like tagged. You’d have to have specialists say this is a…

Yuichi: Yeah, yeah, exactly.

Tim: Oh, and so the university already had effectively had tagged data.

Yuichi: Yeah.

Tim: Excellent. And so how much training data was required? So, how many hours or years of EKG data?

Yuichi: More than 2000 patients. We obtain the data and each patient have more than 24 hours.

Tim: Are you still training and refining? How do you develop an AI like that?

Yuichi: Because the Smart Robin is now commercially used so we can obtain the more and more data so it brings more efficient to cover each kind of devices. And we also already obtained more than 3% of the data from the patient using for the Smart Robin. Such data brings more upgrades for Smart Robin more reliability, more sensitivity.

Tim: Okay. And are you still having specialists tag this raw data and continuing to train and improve?

Yuichi: Yeah. For the purpose we have technicians inside our company to treat and teaching data.

Tim: I’m just kind of curious. So, the EKG has been around for like a hundred years. It’s an old, old technique. Why is it so useful?

Yuichi: Actually, the waveform is very simple. From 1970s, the device makers make an automatic diagnostic system that’s very, very early compared with the CT or other images. But the innovation stops more than 30 years. So, I became a doctor 20 years ago. The system is almost the same during the two decades. So, there is no innovations.

Tim: So, what did those early systems detect based on?

Yuichi: Detect based on the big waves.

Tim: Oh, okay. So, they were just looking for spikes in the wave form? Just looking for high amplitude. Well, that’s something,

Yuichi: Yeah. So, by using deep learning, we can observe the very small waves and the changes in the small waves. So, it’s very important to detect an atrial fibrillation, for example.

Tim: Okay. So, you’re mentioning that the patients would wear a kind of portable wearable EKG device for a day or two or longer I guess. How wearable are these devices? Is it practical for someone to wear this device and just like lead their normal life?

Yuichi: In usual case the doctors recommend to make a usual life.

Tim: It’s not something that they’d have to be in the hospital for. They could wear this and go to work, go to school.

Yuichi: Sure, sure. They can do it in the outpatient situation.

Tim: So, how does this fit into kind of the future of telemedicine, which is important in Japan with the doctor shortage, especially in rural communities, but it’s important everywhere.

Yuichi: Telehealth is not so common in Japan. Still not so common even after the situation of the COVID-19.

Tim: But the Japanese government is pushing hard to develop telehealth and rural hospitals are opening up to the idea because they just can’t get doctors. But I mean, around the world, telemedicine is becoming more and more important. How would this contribute to telemedicine?

Yuichi: Yeah. The big difference with hospitals and telehealth is to make an examinations. So, most of the cases with the telemedicine, we can make only conversations with Zoom or other kind of the telehealth system. But doctors want to know other things. For example, blood samples or electrocardiograms. So, by using the Smart Robin with a patch type ECG, so we can send the ECG directly to the patient they can put by themselves and AI system can lead it easily. So, we can promote the telehealth by using the AI system, by using a tele monitoring system. So, most of the doctors worried about, they fail to observe something only within the system.

Tim: I imagine doctors would be more anxious to use this kind of a technology because it’s something they don’t do themselves anyway. They don’t take the EKGs. Another thing I’ve noticed is Cardio Intelligence focuses a lot on using explainable AI models. Yeah. So, why is that?

Yuichi: Thank you for the focusing this point. So, I think it’s very important to make an explanation for the result from the AI system because then doctor is very anxious to desire only for the output from deep planning system because then the system is black box. So, doctors do not want to in the situation to rely on the AI or not. So in such a case, doctors can see and observe what is happening for the diagnosis.

Tim: Yeah, I think that’s true in a lot of complex decision making. Most people have trouble trusting the black box. Does explainable AI sacrifice accuracy or speed of development or…?

Yuichi: No, it’s a kind of equipment for the systems. So, it covers with high accuracy and high speed to run the deep running system as well as an explainable AI.

Tim: And so it makes a lot of sense for doctor confidence and explainability.

Yuichi: Yeah, exactly.

Tim: Does it also affect regulatory approval or partnerships with equipment manufacturers?

Yuichi: The regulatory system only covers the core algorithm input and output. So, regulation only covers what the devices present for the normal evaluations with the electrocardiogram. So, appendix system is just appendix system. It doesn’t cover what approval.

Tim: Let me ask you a few questions about Japan and startups in general. So, medicine is just such a specialized field. And I’m curious, what role do you think startups have to play in medical innovation in Japan? Because medicine is not something where you can move fast and break things.

Yuichi: Especially for the medical device and as a medical startups. So, it usually takes a lot of time compared with other project, compared with FinTech or smartphone games. It’s very easy to start, easy to get the market. But for manufacturing the medical devices or other kind of health tech, it takes the time to make a commercialization. So, it’s very important the startup covered such kind of the high risk project because the pre-scale device companies does not want to take a risk to make a new device, a new health tech project.

Tim: What is the best way for Japanese startups to overcome that hurdle? So for example, Cardio Intelligence has raised about 8 million US dollars.

Yuichi: Yes.

Tim: So, that’s not a lot of money by US startup standards or even Japan startup standards. It’s not like a 10 year development plan. So, what is the best way for startups to overcome that challenge of needing to have a long development cycle? Needing expensive clinical trials and approvals.

Yuichi: It is very, very difficult to find the collaborators, especially to make a health tech startup. The number of the players is so limited. Business readers, regulatory systems in every field there is less people to work in the startups. The number of the startups and number of the venture captures is quite different compared with the United States.

Tim: Yeah. That’s for medical startups in particular, do you think that in addition to venture capitalists and more founders, do you think that universities or government has a role to play in medical startups specifically?

Yuichi: During a few years, it just start with the support from the government, but the number of player still limited. So, please imagine the investigation in the United States, if a professor make a seed, there’s a lot of people to cover the seeds to make a business, but this kind of situation does not exist in Japan. Of course, there’s a lot of good scientists in the medical field in Japan, but the seed is still seeds.

Tim: Just a seed hasn’t grown yet. So, what is really missing for that? What is missing to get those seeds to grow?

Yuichi: The role pf the venture capital or the business readers is very important to find the seed and to catch and to make a cooperation with the universities and with scientists.

Tim: All right. Yuichi before I let you go, I want to ask you what I call my magic wand question. And that is, if I gave you a magic wand and I told you that you could change one thing about Japan. Anything at all, the education system, the medical device certification system, the way people think about risk, anything at all to make things better for startups and innovation in Japan, what would you change?

Yuichi: In the health stack to more and more activities in Japan, I guess I double triple the number of the venture chapters.

Tim: Okay. So, you think it really is just a matter of funding?

Yuichi: Not only for the amount of the funding itself, but also the number of the capitalists because the capitalists should cover the business, learn the business skills, and to bring the people to make companies. But the power of the venture capital is still not so strong in Japan.

Tim: I was saying I think you have very high expectations of venture capitalists.

Yuichi: That’s the problem with the chicken eggs. So, because there’s a lot of seeds under the ground.

Tim: No, I see this. So, I mean, really what you’re thinking of the role of a venture capitalist in Japan, or at least in in medical science, it’s not just providing capital and connections, but also helping with staffing and growing the business and the legal parts of regulatory, I mean the whole support package.

Yuichi: Yeah. But lowly is not so strong in Japan compared with United States.

Tim: Yeah, I’d say, I mean even in the US the percentage of venture capitalists who do those kind of things well is very small.

Yuichi: Yeah, I understand. Because there’s a lot other specialists and that they want to see the new project and the new good seed and make a lot of monies.

Tim: Of course. So, getting back to the role that venture capitalists can play in Japan and med tech, aside from the funding, which is the obvious role, what was the biggest challenge that they could have helped with? What is it staffing and recruiting or what would’ve been the one big value add?

Yuichi: Yeah. To make a package, to make a business from the seeds.

Tim: Okay. Yeah. So, to really just help commercialize the whole idea.

Yuichi: Yes, for sure. So yeah.

Tim: Okay. And I guess it’s just, it’s something you’ve had to learn on your own.

Yuichi: Yeah.

Tim: You’ve grown. Well, it’s hard, as you say, it’s chicken and egg. There aren’t many people with that kind of experience in Japan, there aren’t many people who’ve started startups. But I suppose as we get more role models, more success cases…

Yuichi: Yeah. That’s a very important thing. And it is my role, to make one success model being off from the university hospitals.

Tim: It looks like you’re well on your way to doing that. Well, listen, Yuichi, thank you so much for sitting down with me.

Yuichi: Thank you very much. My pleasure.

Outtro

And we’re back.

You know, I really admire Yuichi’s approach to knowing and planning for the point when he will step aside as CEO.

One of the most common problems in university spinouts and startups founded by professors in general, in fact, by far the most common problem is that too many of these startups stay technology focused and they never successfully move to full commercialization or to meaningful scale. Yuichi has thought this through from the beginning and has managed to commercialize the product and will do what’s needed to scale the company.

Yuichi’s point about the first mover advantage in AI is also extremely important for all aspiring AI startups. Now sure, almost all markets have a first mover advantage to some extent. But in AI it is especially powerful. The first mover in any vertical gets the usual PR partnerships and brand recognition boosts. But far more important, they get access to a lot of real world data. That new data is used to train and improve their product, which gives them greater market penetration, which in turn gives them access to even more training data.

It’s a virtuous cycle that makes it very difficult for new competitors to enter the market.

I also find Cardio Intelligence’s decision to use explainable AI, very forward thinking. Explainable AI is not getting the media attention it did a few years ago, but I think it explainable AI will be essential for widespread AI adoption in any application that has real world impact.

After all, if AI is going to be human’s partner in the future, we need to be able to understand each other.

 

If you want to talk more about AI in medicine, Yuichi and I would love to hear from you. So, come by disruptingjapan.com/show211 and let’s talk about it. And hey, if you enjoy disrupting Japan, share a link online or just tell people about it. Disrupting Japan is free forever and letting people know about it is the absolute best way you can support the podcast.

But most of all, thanks for listening and thank you for letting people interested in Japanese startups know about the show.

I’m Tim Romero and thanks for listening to Disrupting Japan.