Where Does the Business Model of Medical AI Truly Anchor?

This is an exciting era. A tidal wave of information from medical fields such as genomics and medical imaging is now crashing upon us, enabling us to leverage artificial intelligence to analyze these data and deliver medical insights.

However, as innovative products in the medical AI sector proliferate, some long-standing business challenges are beginning to emerge. For instance, how can startups achieve profitability in this field? And how can healthcare enterprises leverage AI to reverse the trend of rising medical costs? Most importantly, how can medical AI products earn the trust of government regulators, insurance companies, physicians, and patients?

Xconomy, a renowned overseas media outlet, has published an in-depth report on artificial intelligence (AI) in healthcare. The coverage includes the AI-related initiatives of major corporations such as GE and IBM, genomics hackathons, and the impact of medical AI on patients and physicians. VCBeat (WeChat ID: vcbeat) has compiled and translated this article for our readers.

Not long ago, Xconomy hosted a dinner discussion attended by San Diego’s leading figures in technology and life sciences. The conversation focused on the opportunities and risks associated with integrating AI into healthcare, during which the aforementioned issues came to light.

“As a healthcare investor, I am most drawn to the allure of technology,” said Kim Kamdar, a partner at venture capital firm Domain Associates, from his office in San Diego. “This has opened up new avenues for our firm to attract potential co-investors.”

Regarding medical AI, the current consensus is that it is still premature to apply machine learning and related technologies in the healthcare sector, and it remains difficult to foresee how these innovations will function. This aligns with the many concerns raised by Jeff Engel, a senior editor at Xconomy, in his article “The Many Impacts AI Will Have on Doctors and Healthcare Institutions.”

However, there is no doubt that a transformative wave is sweeping through the healthcare sector, with both small startups and industry giants such as IBM and GE vying to secure a foothold in this emerging field.

If there is one industry in urgent need of fresh blood to drive transformation, it is healthcare. In the United States alone, annual healthcare expenditures exceed $3.2 trillion, accounting for approximately 18% of the nation’s gross domestic product (GDP).

For investors, the healthcare industry, despite its lucrative profits, can be daunting. In this sector, patients, healthcare providers, and insurance companies each have their own interests, and regulatory issues are so complex that an investment may take 10 years or more to see returns.

When it comes to promising players in the current wave of artificial intelligence, few examples are more compelling than Grail. This startup, a subsidiary of Illumina, the world’s largest gene sequencing company, is valued at over $1 billion. It is dedicated to enhancing the sensitivity of diagnostic technologies with the aim of monitoring cancer-derived DNA fragments using routine blood samples.

However, cases of failure amid the surge are also common. A typical example is Theranos, a diagnostics technology company backed by venture capital that was valued at $9 billion in 2015 but plummeted to less than one-tenth of that value last year.

Medical AI is gaining significant traction in San Diego, a city with a highly developed life sciences cluster and home to two gene-sequencing giants: Illumina and the Life Sciences Solutions division of Thermo Fisher Scientific.

Meanwhile, San Diego is also home to numerous experts in neural network technology. With the rise of HNC Software, a software developer dedicated to providing analytical tools for the financial industry, its software has been adopted by FICO for applications such as credit card fraud prediction (HNC Software was acquired by FICO in 2002 in an all-stock transaction valued at $810 million).

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

Calit2 is a telecommunications and information technology research institute headquartered at the University of California, San Diego. For its board member Larry Smarr, the business model that comes to mind is Illumina itself. Illumina is a pioneer in gene sequencing technology and has increasingly ventured into genomic data analysis, which involves analyzing the biological functions and genetic variations hidden within the genetic code.

“Their company’s cloud-based solution for analyzing the human genome has substantial capacity,” said Larry Smarr. “Such data indeed require this level of analytical capability. Although we did not employ such analyses in the past, data volumes have now grown exponentially. Therefore, without these algorithms, we cannot expect to derive any medical insights from these data—a point that is particularly evident in genomics and microbiomics.”

Leveraging its superior gene sequencing technologies and data services, Illumina has established a robust customer base across numerous genomics research centers, clinical research institutions, and biotechnology and pharmaceutical companies. But can this business model be easily replicated? For instance, if another company were to enter the microbiome data analytics market, could it achieve a position comparable to that of Illumina?

Subsequently, Smarr turned the conversation to Rob Knight, seated across the table, who holds a joint appointment in Pediatrics and Computer Science at the University of California, San Diego. Knight currently serves as the Director of the Center for Microbiome Innovation at the University of California, San Diego. He is also a co-founder of the American Gut Project, a citizen science initiative that has collected more than 16,000 fecal samples to help scientists gain further insights into the role of microbes in human health.

“It is important to remember that this venture is non-profit,” said Knight. “I believe this model will face significant challenges. Historically, companies built around selling gene sequencing services have largely failed to achieve favorable outcomes. For example, Celera shifted its business model toward diagnostics.”

“I believe we should somehow shift the paradigm toward real-time feedback and develop an interactive interface that helps users understand their microbiome,” Knight cited as an example, “such as enabling users to instantly know whether the piece of bread they just ate had a positive or negative impact on their body.”

Of course, some companies have already begun to implement this business strategy. Nutrino, a technology company based in Tel Aviv, Israel, has developed an application and data platform that helps users understand how the foods they consume affect their physiology.

“They can provide real-time guidance based on the impact of users’ ‘dietary footprints’ and their blood glucose patterns,” said Annika Jimenez, Senior Vice President at DexCom, a San Diego-based company specializing in continuous glucose monitoring technology and diabetes management.

“This model is similar to insurance-based payment, but over time, they will shift their business model to target enterprises and other potential customers,” said Jimenez.

The key advantage of AI in the healthcare industry lies in its exceptionally powerful data extraction capability, enabling it to derive actionable medical insights from datasets ranging from exabytes to zettabytes—a scale far beyond human cognitive capacity.

“For me, identifying a precise and viable business model seems to be a long-term ultimate goal,” candidly stated Rick Valencia, President of Qualcomm Life, who appears skeptical about the current revenue-generation models in this field. “In the short term, I believe the answer to your question is ‘no.’ At least within my scope of observation, I have not discovered any effective business models; I feel it is still too early.”

Navid Alipour, Co-founder and Managing Partner of San Diego-based analytics venture capital firm Analytics Ventures, stated that their portfolio company, CureMatch, is implementing a direct-to-patient model. Under this model, patients pay CureMatch directly, and the company provides them with recommendations for the top three targeted chemotherapy regimens tailored to their specific cancer conditions.

These recommendations are based on patients’ own medical records and are designed to assist oncologists in selecting treatment regimens. CureMatch stated that its supercomputer processed millions of combinations involving three chemotherapy drugs, evaluated the drug–drug interactions for each combination, and integrated genomic data to generate personalized drug combination recommendations for individual patients.

CureMetrix, another company backed by VCBeat Ventures, is dedicated to using AI to analyze mammograms for breast cancer. Of course, its technology must still receive FDA approval before it can be marketed and used in the United States.

Alipour stated, “Software as a Service (SaaS) will become a business model. An institutional investor of ours in Mexico is introducing us to their senior government officials. In Mexico, breast cancer is a prevalent and troubling issue, and the country lacks many experts in mammography. We are also pursuing nationwide licensing with them, given their nationalized healthcare system. Therefore, at times, we must think beyond the United States and our insurance-based system.”

Numerous companies, large and small, are applying machine learning to diagnostic imaging to identify abnormalities, with CureMetrix being just one of them. This image-analysis-based approach appears to be the ultimate application of AI technology. However, Jimenez noted, “You still need to attend the Strata Data Conference, a major event in the fields of big data and data science, where keynote speakers consistently emphasize just how complex these use cases actually are. It goes without saying that we may have to wait not just another 10 years.”

So, when will AI replace radiologists?

Smarr expressed skepticism that AI would replace radiologists. Instead, he believes this technology will serve as an aid to human physicians, enabling even the least skilled radiologist to make diagnoses more accurately than the most accomplished human peers.

“Therefore, dedicating efforts to the application of AI technology in healthcare is essentially about arming humanity with unprecedented volumes of data and elevating human intelligence,” Smarr added. “This can indeed boost productivity in the short term, although by ‘short term,’ we are referring to a timeframe spanning several decades.”

“Companies like DexCom focus on mainstream diabetes management, whereas Holy Grail is dedicated to reshaping patients’ behavioral habits,” Jimenez stated. “This means that by integrating data streams from blood glucose monitoring, insulin measurements, patient behaviors, and dietary patterns, and leveraging machine learning to generate medical insights, the software can provide timely alerts and recommendations to both patients and their physicians.”

“But our technological maturity is still at the stage of merely providing numbers,” Jimenez added. “Thus, we only inform patients of their blood glucose levels, which is certainly critical for those with type 1 diabetes. However, patients with type 2 diabetes need to interact with the app and are required to respond to medical insights. This is where the true demand for app development lies.”

Perhaps the ultimate goal of this technology is to develop a user interface that truly meets needs, leveraging medical insights derived from machine learning to fundamentally transform the behavioral habits of patients with diabetes.

This view is endorsed by Jean Balgrosky, who has served as CIO for multiple large healthcare organizations, such as Scripps Health in San Diego, for 20 years. She stated, “Ultimately, all machine learning technologies should be absorbed and leveraged by humans to play a supportive role in the healthcare sector.”

References:

http://www.xconomy.com/san-diego/2017/07/03/whats-the-business-model-for-artificial-intelligence-in-healthcare/?single_page=true