F120 A glimpse into Japan and how to introduce AI to clinicians (Adrian Sossna)
“I think the first thing we need to collectively communicate and understand when it comes to technologies such as AI, is that an AI model or an AI tool is not Microsoft Word or Facebook Messenger. It's not a strict piece of software that you build, you test, you ship, and then it's finished. AI needs to be seen as a tool for automation in specific fields and use cases. You can't say that an AI model is a finished product. It's probably never a finished product,” says Adrian Sossna, VP of Global Sales at the Japanese company Hacarus.
Hacarus is a Japanese company developing AI Solutions for Manufacturing and Medical Industries. Their Salus platform for medical and life sciences uses Medical imaging data such as CT & MRI scans, time-series data, such as ECG data, and medical record to create precise, complex tools, that aid caregivers and researchers to provide better, faster and safer treatment, based on data-driven insights.
In this episode, Adrian Sossna, who is originally from Sweden but has been living in Japan for several years now, shared his insights into life in Japan, the tech ecosystem, and the challenges in developing AI for healthcare and medicine.
Adrian Sossna has been living in Japan for six years but has been in Asia for nearly a decade.
As he describes his first memories of Japan from 2008, he was disappointed by the state of technology, since he was under the impression that Japan was a more digital and connected country than it turned out to be. “Interesting things here are not digitized yet. For instance, we use a type of stamp to sign documents and various official documents. You need to have a personal seal. But I would say that in the past decade or so, we've seen a dramatic realization on the side of Japanese society, of the need to digitize and the need to adopt new technologies,” says Sossna.
What he appreciates about business in Japan is attention to detail craftsmanship and deliverance of stellar products. His role is to bring these products to the global market, currently as the VP of Global Sales at Hacarus.
Elderly care and technology in Japan
Japan received a lot of attention in the international media due to its use of robots to aid caregivers and keep the elderly company. The Japanese culture perceives robots first as friends not as enemies as is an often pop-culture reference in the West.
“In the West, the Terminator in the late 90s created a scare around AI robots. Compare that to Sony, which made a lot of news for making the first robot dog. I think that gives you a good indication of the different perceptions of robots. In Japan they are not killing machines, but the man's best friend.”
Japan is the most rapidly aging society in the world. There's a growing shortage of people that are able to provide care for the elderly as well as the rest of society. As commented by Adrian Sossna, in addition to importing labor, deploying robots deploying automation solutions, deploying ways to digitize the healthcare system is seen as a necessary thing, a solution to a very acute problem.
Diagnostics, Imaging, and AI
Japan has a very high number of diagnostic imaging instruments, such as CT and MRI devices. But according to the data published in 2015, 60 to 70%, are done in the absence of a radiologist. What does that mean in terms of the accuracy of reading the tests done with these machines?
Hacarus is working on digitizing human expertise to aid the accuracy of diagnostics. For example, Hacarus is working with Kobe Universit on early-stage liver cancer detection using MRI imaging, as well as projects with companies looking at brain stroke classification in MRI images. “The problem tends to be the same. You have expertise that's needed to be able to detect, classify, look at these images. This expertise exists, but it's very limited. Large hospitals in large cities will have it, but you won't find it everywhere. Our work is largely about how do we take specialist knowledge and create AI models that are able to replicate that expertise to provide insights to doctors or caregivers that are at sites where you don't have that level of specialist knowledge?”
Training doctors in AI to build trust in technology
AI is complex and when introduced to the clinical practice, adoption is anything but easy. “On the one hand, there's the hype machine that provides these very extravagant articles about magical things that you can do it with AI. Then there's the reality of what can be achieved and what can be done. That gap can sometimes be a problem in terms of managing expectations from users, doctors. On the other hand, it's also a lack of knowledge of how these things actually work. At Hacarus we’re addressing this with a training program for the medical industry, which we call the Hacarus AI Academy. This is a full course program, where we train doctors, medical professionals, researchers in pharmaceutical companies, the basics of how to work with AI. We take them from the very beginning through several month course. At the end, we do a real project with real medical data, and participants are the ones that are doing it, our team assists and guides.”
AI accuracy concerns
One thing the AI field does not lack is concerns. Algorithms are very narrow. Most AI models are based on retrospective studies, lacking real-world validation, which is needed before wider adoption in clinical practice. Two other issues challenging AI development and introduction in clinical practice are AI brittleness and concept drift.
AI brittleness refers to the changing accuracy of AI models when they are transferred from one dataset to another. An AI model trained on a data set from one hospital can fail when transferred on a different data set in a different hospital, because of different machines used, different protocols doctors follow, etc. Additionally, what can occur even when a model works, is concept drift - a decrease in accuracy of an AI model over time due to imaging diagnostics changes (different device, new protocols, etc.).
“I think the first thing we need to collectively communicate and understand when it comes to technologies such as AI is that an AI model or an AI tool is not Microsoft Word or Facebook Messenger. It's not a strict piece of software that you build, you test, you ship, and then then it's finished, with the same interface and the same way to interact with it, no matter where you are, no matter what you're doing. AI needs to be seen as a tool for automation in specific fields and use cases. If you transfer a model from one piece of equipment to another one, so a different type of data format or a different type of conditions, then you're going to get very different results. And that's because what you've trained the algorithm to do is is detect in that specific format for those specific things. So it's not interchangeable. But rather than seeing that as a problem, you can also see that as a strong opportunity for really customized healthcare solutions. You can create a framework where the top-level AI algorithm is the same, but the implementations are unique to a hospital, the geography, the ethnicity. When you think about it like that, some of the things actually become advantages. As far as decreasing performance over time goes, I think that that boils down to the same question, or the same point: you can't say that an AI model is a finished product. It's probably never a finished product. It's continuously in need of updates, it continuously needs to get better with more data or new data.”
Tune in for the full episode:
Some questions addressed:
Let’s start with something easy: You are originally from Sweden, but have been living in Japan for 8 (?) years by now. How would you describe the country based on the time you’ve been living and working there? What strikes you in the culture, how advanced is digital health, what got you there etc?
At the recent MEDICA Asia Fair you participated in the panel: How the Pandemic Changed the use of Blockchain, AI & Cloud in Healthcare. How exactly did that change in Japan?
Hacarus AI models are used in genome analysis, regenerative medicine to aid in developing methods to regrow, repair or replace damaged or diseased cells, organs, and tissue. You also offer your AI to aid Pharma with drug discovery. How exactly is your technology applied to these areas?
How do you see your technology in action in the future? SALUS is HACARUS’ platform for Medical & Life Sciences AI solutions.
You work a lot with medical imaging and designing decision support systems for radiologists. Fortunately, Japan has a very high number of diagnostic imaging instruments, such as CT and MRI devices. I assume many companies are looking for AI-supported solutions to increase the accuracy of imaging-derived diagnosis?
In May 2020 The Journal of Intensive Care published an article about medical AI. The advent of medical artificial intelligence: lessons from the Japanese approach. “The demand for an AI-literate workforce has outpaced training programs and knowledge in Japan. This is particularly observable within medicine, where clinicians may be unfamiliar with the technology. How does AI development in Japan compare to other countries in the world in your view?
At the moment, there are still a lot of AI-related concerns in the use of solutions. How are you addressing these issues and what do you expect of AI development and it’s impact on healthcare in Japan in the upcoming years?