Medical AI technology is hot, but where does its business model end?

These are exciting times. With the huge wave of information coming to us from medical fields such as genomics and medical imaging, we will be able to use artificial intelligence to analyze this data and provide medical insights.

However, with the surge in innovative products in the field of medical AI, some old-fashioned business problems have also begun to emerge. For example, how can startups achieve profitability in this field? And how can healthcare companies use AI to reverse the continued increase in medical costs? Most importantly, how can medical AI products gain the trust of government regulators, insurance companies, doctors and patients?

Xconomy, a well-known overseas media, has conducted in-depth reports on medical artificial intelligence, including the related work of large companies such as GE and IBM on medical AI, the genomics programming marathon (hackathon), and the impact of medical AI on patients and doctors.

The Problems Behind the Technology Wave

These questions came to the fore during a recent dinner discussion hosted by Xconomy, which was attended by some of San Diego’s most prominent tech and life science experts. The discussion focused on the opportunities and risks of combining AI with healthcare.

“What I love most about this as a healthcare investor is the technology aspect,” said Kim Kamdar, a partner at venture capital firm Domain Associates, in her San Diego office. “It opens up a whole new avenue for potential co-investors in our company.”

Regarding medical AI, the general consensus is that it is too early to apply machine learning and related technologies in the medical field, and it is difficult to foresee how these innovative achievements will play a role. This is just like the many questions raised by Jeff Engel, senior editor of Xconomy, in the article "The Many Impacts of AI on Doctors and Medical Institutions".

However, there is no doubt that the wave of transformation in the medical field is surging, and both small startups and industry giants such as IBM and GE are scrambling to gain a foothold in this emerging field.

If there’s one industry that’s in dire need of an infusion of change, it’s healthcare. In the U.S. alone, healthcare costs more than $3.2 trillion a year, or about 18% of the nation’s gross domestic product.

For investors, the healthcare industry is lucrative but daunting, with patients, providers and insurers all competing against each other and regulatory issues so complex that it can take a decade or more for an investment to pay off.

There’s perhaps no better example of a company riding this wave of AI than Grail, a startup valued at more than $1 billion that’s a spinoff of Illumina, the world’s largest genome-sequencing company, that’s working to improve the sensitivity of diagnostics to detect cancer DNA fragments using routine blood samples.

However, there are many cases of companies capsizing in the tide. A typical example is Theranos, a venture capital-funded diagnostic technology company that was valued at $9 billion in 2015, but its value plummeted to less than one-tenth of that last year.

Medical AI is gaining popularity in San Diego, a city with a well-established life sciences cluster and home to two gene sequencing giants: Illumina and Thermo Fisher Scientific’s Life Sciences Solutions team.

At the same time, San Diego also has many experts in neural network technology. With the rise of HNC Software, a software developer dedicated to providing analytical tools to the financial industry, their software has now been used by FICO to predict credit card fraud, etc. (HNC Software was acquired by FICO in 2002 for US$810 million in a stock transaction).

What do the big names think?

The dinner organized by Xconomy invited local investors, data scientists, medical company CEOs, academic researchers, and digital health company executives, including Kamdar. The opening question of the dinner was whether there was a proven business model for startups dedicated to applying machine learning in the medical field.

For Larry Smarr, director of Calit2, a telecommunications and information technology research institute based at the University of California, San Diego, the business model that comes to mind is Illumina itself, a pioneer in gene sequencing technology and increasingly in genomic data analysis, the analysis of biological functions and gene variations hidden in the genetic code.

“The capacity of the cloud solution they’re using to analyze the human genome is quite impressive,” Larry Smarr said. “And the data really requires this level of analysis. We haven’t been doing this kind of analysis in the past, but the volume of data has increased exponentially. So, if we don’t use these algorithms, we can’t expect to get any medical insights from this data, especially in genomics and microbiome.”

With its excellent gene sequencing technology and data services, Illumina has found customers in many genome research centers, clinical research institutions, and biotechnology and pharmaceutical companies. But can such a business model be simply copied? For example, if another company started a microbiome data analysis business, would it be in the same situation as Illumina?

Smarr then turned to Rob Knight, who sits across the table and holds a joint appointment in pediatrics and computer science at UC San Diego. Knight is the director of the UC San Diego Center for Microbiome Innovation and a co-founder of the American Gut Project, a citizen science initiative that has collected more than 16,000 stool samples to help scientists better understand the role of microbes in human health.

"First of all, remember, I'm a nonprofit," Knight said. "I think this model is definitely going to be tough because generally speaking, companies that built their business around selling gene sequencing have not had great success. For example, Celera has moved its business model into diagnostics."

“I think we should somehow shift the paradigm to real-time feedback and develop an interface that lets users learn about their microbiome,” Knight said, giving the example of, “like, letting users know instantly whether the piece of bread they just ate is having a positive or negative impact on their health.”

Of course, this business strategy has already been implemented by some companies. Nutrino, a technology company based in Tel Aviv, Israel, has developed applications and data platforms to help users understand how the food they consume affects their own physiology.

"They can provide real-time guidance on the impact of a user's 'dietary footprint' and their blood sugar performance," said Annika Jimenez, senior vice president of San Diego-based DexCom, a company focused on continuous glucose monitoring technology and diabetes management.

“It’s similar to insurance premiums, but over time they’ll shift their business model to target businesses and other potential customers,” Jimenez said.

The key advantage of AI in the healthcare industry lies in its extremely powerful data information capture ability, which can obtain useful medical insights from exabytes to zettabytes of data, a scale that far exceeds the ability of humans to understand.

"To me, finding a viable business model seems to be a long-term ultimate goal," said Rick Valencia, president of Qualcomm Life, who seemed skeptical of the current revenue model in this field. "In the short term, I think the answer to your question is 'no'. At least as far as I can see, I haven't found any viable business model. I think it's too early now."

Navid Alipour, co-founder and managing partner of San Diego-based Analytics Ventures, said CureMatch, in which his firm is an investor, is implementing a direct-to-patient model, in which patients pay CureMatch directly and the company recommends a top three chemotherapy drug combination for each cancer.

These recommendations are based on the patient's own medical records and are intended to help cancer doctors choose treatment options. CureMatch said its supercomputers processed millions of combinations of three chemotherapy drugs and evaluated the drug interactions of each combination separately, then integrated it with genetic data to come up with drug combination recommendations for specific patients.

CureMetrix, another company invested by Analytics Venture Capital, is working on using AI to analyze mammary radiology images for breast cancer. Of course, their technology still needs to be approved by the FDA before it can be marketed in the United States.

“Software as a service is going to be a business model,” Alipour said. “One of our institutional investors in Mexico is introducing us to their government at a high level. Breast cancer is a common problem in Mexico, and there aren’t a lot of experts in mammography. We are licensing it to them nationwide because they have a national health care system. So sometimes we have to think outside the U.S. and our insurance system.”

CureMetrix is ​​one of a number of companies applying machine learning to diagnostic images to identify abnormalities, and this image-based analysis model seems to be the ultimate application of AI technology, Jimenez said. "But you have to go to the Strata Data Conference, which is arguably the biggest event in the field of big data and data science, and the speakers always emphasize how complex this use case actually is. You can imagine that we may have to wait more than 10 years."

Replacement is not the goal, utilization is!

So, when will AI replace radiologists?

Smarr said he is skeptical that AI will replace radiologists. Instead, he believes the technology will complement human doctors, enabling even the worst radiologists to make more accurate diagnoses than their best human counterparts.

"So the application of AI technology in medicine is actually to use unprecedented massive amounts of data to arm humans and improve their intelligence," Smarr added. "This can really improve productivity in the short term, but this short term is also within decades."

While companies like DexCom focus on prevalent diabetes, Holy Grail is working to reshape patient behavior, Jimenez said: "That means integrating data streams from blood sugar monitoring, insulin measurement, patient behavior and diet, and using machine learning to generate medical insights so that the software can issue alerts and recommendations to patients and their doctors in a timely manner."

“But we’re still at the stage where we can just provide some numbers,” Jimenez added. “So we’re just telling the patient what their blood sugar is, which is critical for type 1 diabetes, of course. But for type 2 diabetes, they need to interact with the app and be able to respond to the medical insights. That’s where the real need for app development lies.”

Perhaps the ultimate goal of this technology is to develop a user interface that can truly meet the needs and use the medical insights obtained through machine learning technology to fundamentally change the behavior habits of diabetic patients.

This view is echoed by Jean Balgrosky, who has served as CIO of several large healthcare organizations for 20 years, such as Scripps Health in San Diego. She said: "At the end of the day, all machine learning technology should be absorbed by humans and used to help humans in the medical field."

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.