- A Deep-ish Dive into AI Health
Just now·15 min read
Healthcare is one of the most promising — and also one of the most problem-prone — industries in the United States and other countries across the world. Considering what most Americans pay for their healthcare, quality and outcomes are extremely poor. Because of the supply constrain for physicians, Americans have far fewer doctors visits than peer countries. This, coupled with drug price increases alongside governmental strife has led to a perfect storm of operational inefficiencies for providers, suppliers, and exacerbated health issues for patients. To build on this more, the US now possesses one of the highest rates of hospitalizations from preventable causes like diabetes and hypertension, as well as the highest rate of avoidable deaths. There are two common themes here among the problems in the United States: a lack of accessibility, as well as slow execution for all parties, leading to the system’s current inefficiencies.
In order for proper execution to occur, the deeper problem, accessibility to quality and scalable care, must be solved first. To solve the seemingly insurmountable issue that is accessibility, healthcare requires two things:
- Integration with technology: While there is a clear national desire to adopt more advanced software into healthcare, costs are too high for there to be an incentive to adopt. Within this, there need to be decreased implementation and maintenance costs for incentives to arrive. Furthermore, on a micro-level, there needs to be ease of access to information regarding technological adoption for employee and administrative understanding.
Business Model Overhaul: There is incredible fragmentation within the industry. For instance, more than half of US physicians work in practices of three or fewer doctors, a quarter of the nation’s 5,000 community hospitals and nearly 17,000 nursing homes are independent. This problem of fragmentation extends into BioTech and MedTech; thousands of small firms exist within a space run by few large competitors. There needs to be a solution, some sort of common platform, to aid in the synchronization of all these independent players under one large hub. This hub will allow for horizontal integration and therefore generate economies of scale, allowing for true efficiency and unity when it comes to treating patients as well as sharing patient information and general knowledge sharing.
The amount of technological progress society’s made in the past few years alone is remarkable. Artificial intelligence is the catalyst that will aid in mitigating informational and business problems in healthcare through implementing algorithms and connectionist networks designed to imitate human actions and function flawlessly upon learning. This will allow for relatively instantaneous administrative improvements and easy maintenance as the industry begins to see the positive effects of integration. It doesn’t stop there; AI is the key to rapidly scalable automation and solving a supply-constrained issue. Robot-assisted surgeries will soon be performed entirely by robots, and R&D timelines will condense as digital twins manifest as a common practice in drug development. Accenture said it best: AI is becoming the new OS in health.
I’ll keep it the cliff notes version. AI is, as described here, “leverages computers and machines to mimic the problem-solving and decision-making capabilities of the human mind.” AI can be thought of as not just being a complement to humans like current computer algorithms are, but rather directly replicating aspects of being human in order to scale while simultaneously upping the quality of the task being performed.
This includes implementing sensing, comprehending, acting, and learning. These “human” aspects are performed already by computer algorithms, however, their design prohibits them from modulating their actions to the extent of true human thought. This is where AI comes in, where human activity is truly augmented.
AI doesn’t end there, though. Within this space are a number of subcategories, each built off of the other. As IBM puts it, think of AI, ML, Neural Networks, and Deep Learning as Russian Nesting Dolls:
Each concept within the umbrella of AI relies on fundamentals from the larger picture.
The fun doesn’t stop there (for me at least. And don’t worry, I promise this is relevant to healthcare and life sciences). Neural networks, derived from the connectionist models in cognitive psychology and neuroscience to represent how the human brain’s neural networks process information and concepts. Within these neural networks are “deep” neural networks that possess multiple layers and have input fed through multiple layers.
Additionally, neural networks utilize backpropagation techniques to feed input back through the layers in order to calculate the error with each neuron or node. As a result, by using mathematical models to analyze errors and modify at each node, the network begins to learn how to properly process massive data sets far faster than humans ever could. This leads to some remarkable capabilities such as highly accurate and scalable predictive analytics models.
To be more specific, ANN models have three layers of neurons: the input layer that receives information, the hidden layer that’s responsible for extracting patterns and undergoing most of the internal processing, and the output layer responsible for producing and presenting said outputs.
A typical Deep Neural Network model (credit to IBM) where feed-forward learning methods occur, but also backpropagation is used to attribute and calculate error at each “neuron”
While this deep dive is focused on AI and healthcare, I want to take a moment to focus on the relevance of ANNs. The connectionist model of an ANN allows it to solve highly complex problems by being able to generalize pattern info to new data, tolerate noisy inputs, and also produce reliable and reasonable estimates as more layers are added to the model and as the neurons in the ANN become more exact in their calculations via backpropagation.
All this high-level technical insight is great, but what does this actually mean for healthcare?
ANNs are — and will be — the key to superior decision-making in healthcare, as well as unlocking new healthcare innovations.
In 2019, Humana found that about $1 in every $4 spent in healthcare in the US is wasted. A majority is being wasted as a result of misguided decision-making at various levels within the healthcare system.
- These issues are due to other inefficiencies I’ve mentioned earlier, such as a lack of information sharing and connectivity, leading to the same poor decisions being made by poorly connected healthcare providers, suppliers, and patients.
ANNs allow for not just the seamless aggregation of information to inform better administrative decision-making, they imitate the decision-makers themselves by devising their own plan of action for administrators to implement at scale.
This is just one small example of the power of ANNs, and therefore the power of AI itself, in just one sliver of the healthcare industry on the provider side, not even getting into the suppliers in MedTech and Life Sciences.
Let’s get back to business… We know what’s wrong with this market, what the solutions are, and how each solution plays into each others. Through this, we’ve been able to get a pretty good idea of what can happen when healthcare and AI go together. Let’s get into the business side of things a bit more.
Innovation is a key factor when it comes to the viability of a market’s potential. What’s one of the best ways to measure innovation or the drive to innovate? R&D expenditures. Take, for instance, the United States’ R&D expenditure growth, a pretty clear indicator of the significant increase in interest in healthcare and life sciences. From 2013 to 2018, R&D expenditures increased by 36% (from $143B to $194B). It’s pretty remarkable what R&D by both public and private entities have allowed come into fruition.
Companies across the spectrum of AI health are churning out products and are basking in high-potential sub-markets. Let’s look at Intuitive Surgical (while it’s now public, the work they’re doing is incredibly cutting edge. We’ll get to some exciting startups shortly), a leader in developing minimally invasive care via advanced robotic systems. Since the development of their first robotic system in the early 2000s, their Da Vinci model is now in its seventh version, having over 3,000 systems manufactured and sold in the last three years. Total revenue is up over 30% since 2018, and the operating system is now prepared to perform surgeries ranging from hiatal hernia repair to hysterectomies to GI surgeries such as cholecystectomies. If this monstrous expansion is occurring in a publically traded company, what innovations and growth paths are smaller startups on?
Additionally, similar to most tech innovations, AI is incredibly applicable to a variety of current key gears in healthcare, such as drug pricing and discovery, mental health, or really any space requiring humans in the healthcare space (so every space in healthcare)! Looking beyond healthcare, this is what makes AI exciting and game-changing in just about any industry on earth.
- Drug Discovery: Drug discovery is a long and complex process depending on a laundry list of factors to be deemed effective and safe. When it comes to painstakingly long processes within drug development such as making decisions derived from mountains of data and engaging in drug discovery processes such as target validation, identification of biomarkers, and analyzing digital pathology data, AI/ML is the perfect tool to speed up a process that takes ages even by advanced computer algorithms. It’s no wonder that this is a $1.5B market opportunity.
There are immense and ever-changing data requirements for various stages of the drug discovery process. It currently takes an immense amount of time and resources to parse through the data. AI/ML can accelerate the discovery process through rapid learning and backpropagating at scale.
2. Virtual Health Assistants (VHAs): VHAs exist as chatbots in other industries such as retail, but are practically absent in the healthcare space. When it comes to quality diagnosis of a patient’s symptoms, there is the immense capability to scale while maintaining or improving quality through an AI/ML program. This can drastically reduce healthcare visits for mild conditions that can be treated by the patient and ease up patient traffic for healthcare providers, leading to a far higher quality of care at a far lower price in a no longer unnecessarily supply-constrained market. Currently, this is an estimated $20B market, however as AI/ML is further integrated into societal doings, it will become more commonplace and promoted to visit VHAs to take care of treat-and-release hospital visits, which on average make up 82% of all patient visits regarding flu-related symptoms from 2006–2016, one of the most common reasons to visit the hospital.
3. Digital Twins (R&D): This is a new one. Essentially, a digital twin is a clone of a process, product, system, or facility. In the case of AI/ healthcare and life sciences, digital twins can be patients, organs, or even individual cells. Even better, based on how each digital twin is designed, it can test a drug, product, or treatment at scale with near-infinite combinations of different accompanying conditions or disabilities. Imagine a drug being developed meant to help with breathing for those with asthma. Instead of doing years-long tests first on cells, then animals (a rightfully dying practice), and then humans, this asthma drug can be tested virtually on digital cells, animals with various accompanying conditions such as heart problems or blood diseases, and finally humans. Through implementing AI/ML to analyze such massive swathes of data and predict the accuracy of the drug and even quantify safety, can now drastically speed up the lengthy R&D process, as well as make testing on all types of life far safer and ethical. This is an exciting place to be, with a current global market size of roughly $3.1B, with the expectation that it’ll grow far upwards of $40B by 2026.
4. Robotics: We talked a bit about Intuitive Surgical and their advances with their Da Vinci Surgical System, but robotics and AI go beyond just surgery. As a result of traumatic incidents, individuals are left anatomically impaired, missing limbs or other organs essential to sensory and the ability to function in society. Enter neuroprosthetics, “devices that help augment the subject’s own nervous system, in both forms of input and output.” Recent advances in brain-machine interfaces (BMIs) have allowed subjects to store voluntary and goal-directed wishes (as seen through EEG) in the neuroprosthetic powered by AI through training the intelligent controller, the prosthetic. As a result, this training through the AI-powered prosthetic allows for truly personalized movement of a typically rigid prosthetic, adding personality and increasing the quality and function of the previously impaired patient. As seen through a number of key players such as Medtronic and Intuitive in this space, it’s clear this is a hot industry, valued collectively at over $40B, with the neuroprosthetics market currently valued upwards of $14B.
Below, I’ll investigate seven startups working to build at the intersection of AI/ML and healthcare/ life sciences to help healthcare providers, suppliers in life science/ medtech, and patients overcome problems such as administrative issues, access to quality care, scalability of said quality, diagnosis accuracy, robotic capabilities, drug development, and more.
As exciting as it is to explore this fascinating intersection at a high level, examine factors of innovation and markets of note, the best way to see the importance and effectivity this space has held and will hold in society is through investigating startups uprooting how healthcare is carried out for each of these three participants in the space.
All financials and investor info found via Crunchbase as of 11/24/2021!
To elaborate, Buoy Health is a virtual health assistant (VHA) leveraging ML techniques to take incalculable combinations of potential consumer symptoms to train and make symptom diagnoses without the initial need for patients to visit a hospital. As Buoy’s system interacts with the patient, it prepares a variety of specific medicines, services, and professionals for the patient to visit regarding their condition, leading to less time and money spent on inefficient trial and error.
- Founders: Andrew Le, MD and Adam Lathram
Funds Raised and Investors: $62.3M/ Series C; Notable Investors → F-Prime Capital, Optum Ventures, Cigna Ventures
HQ: Boston, MA
Prognos Health utilizes AI methods to compile and refine billions of entries worth of healthcare data on over 325 million patients to establish an effective and refined marketplace for healthcare providers and suppliers to meaningfully utilize as they make administrative and individualized healthcare decisions for both their patients, as well as how healthcare systems and life science companies should operate.
Diligent Robotics is increasing impact in high-stress environments while decreasing human input through building socially and emotionally aware robots that can aid in the workplace. For instance, one of their most recent models is Moxi, a robotic hospital assistant that can help staff with routine activities so that medical professionals can spend more time caring for patients. Nurses spend up to 30% of their time on routine, non-value-added tasks, and Moxi is there to take over so nurses can focus on what really matters: the patients. To date, over 35,600 tasks have been completed by Moxi models around the United States since its launch in 2019.
- Founders: Vivian Chu (CTO) and Andrea Thomaz (CEO)
Funds Raised and Investors: $15.9M/ Series Unknown; Notable Investors → True Ventures, The E14 Fund, Ubiquity Ventures, Promus Ventures
HQ: Austin, TX
Unlearn.AI is a one-of-a-kind, utilizing AI to simulate biology. Unlearn.AI specializes in developing digital twins, which as we saw earlier, are virtual representations of any object you can think of, as represented by data. This is massive for how R&D in healthcare and life sciences is done. It normally takes months or years at each phase to test first on individual cells, then organs, and then humans. However, with an incalculable amount of combinations of organisms with different extraneous factors such as genetic deficiencies, blood disorders, etc. scientists can rapidly observe the consequences — both positive and negative — of their innovation and rapidly make necessary changes that are rooted in safely aggregated and assembled predicted data points.
- Founders: Charles K Fisher (CEO), Aaron Smith (Head of Engineering), Jon Walsh (Head of Data Science)
Funds Raised and Investors: $15M/ Series A; Notable Investors → DCVC Bio, Epic Ventures, Alumni Ventures, 8VC
HQ: San Francisco, CA
Genesis Therapeutics is a leader in integrating in-house assembled deep neural networks with biophysical simulation onto an easily digestable and scalabe infrastructure for industry participants of all sizes from F500 companies to small biotech startups to use while examining molecular generation of their products as well as property prediction. This will be a huge leg up for the drug discovery industry and, like Unlearn.AI, aid in rapidly decreasing R&D time and expenditures so that higher quality solutions can be discovered safely in less time.
- Founders: Ben Sklaroff, Evan Feinberg
Funds Raised and Notable Investors: $56.1M/ Series A; Notable Investors → Menlo Ventures, Ulu Ventures, Jazz Venture Partners, Radical Ventures
HQ: San Francisco, CA
Curai Health is a VHA that is leveraging AI/ML techniques to go the extra mile with patient care, providing immediate actionable steps for the patient alongside connecting to further medical professionals and other solutions if needed. As the patient interacts with the VHA, the program rapidly utilizes ML techniques to determine next steps in terms of the most appropriate questions to ask to determine the most probable diagnosis. This will allow for more pinpointed and cheaper expenditures on the pateint’s part, as well as alleviating hospital traffic for more pressing conditions to easily obtain priority and swift help.
- Founders: Neal Khosla (CEO), Neil Hunt (CSO), Xavier Amatriain (CTO)
Funds Raised and Notable Investors: $38.2M/ Series A; Notable Investors→ Khosla Ventures, General Catalyst, Civilization Ventures, Morningside Ventures
Pricing Model: $7.99/ month subscription for individuals. Pricing varies for larger entities such as corporations
HQ: Palo Alto, CA
Q Bio is another digital twin startup with an incredibly unique mission of not just making digital twins more commonplace in healthcare and life science R&D, but also being the pioneer in developing an easily accessible location for individuals and large institutions alike to engage with utilizing digital twins for research. Through an hour-long exam at a “Q Center”, Q Bio gathers millions of data points from medical records, blood, saliva, vital, and urine, as well as a whole-body MRI, where data will then be stored and processed in a secure “BioVault” that, at the user’s discretion, can be easily transferred to any professional in the user’s healthcare network with ease. This will not only make understanding one’s vitals and health easier, but make it far more accessible to researchers and professionals when attempting to decode a patient’s health abnormalities.
- Founders: Garry Choy, MD, MBA, Dr. Michael Snyder, Jeffrey Kaditz
Funds Raised and Notable Investors: $84M/ Series B; Notable Investors → Founders Fund, Andreesen Horowitz, Khosla Ventures, Thirty Five Ventures
Pricing: $3,495 annual membership
HQ: San Carlos, CA
That’s a wrap on my first ever Deep-ish Dive! It’s quite a handful to read a lot from the fundamental problems of healthcare, how AI/ML assists healthtech as well as biotech, and even getting into the nitty gritty of AI/ML (and I didn’t even touch the math or computational side, just a high, high level of one of many subconcepts). This space is absolutely awesome and is the culmination point of decades of innovation in computers, psychology, mathematics, biology, neuroscience, business, and robotics. In terms of exploring the origins of AI and how innovations in this space will change how healthcare — and even how business in general is done — there is no shortage of exciting rabbit holes to dive down. This was a major eye opener for me. I’ve had a rough idea of how psychology, neuroscience, and biology played into this bigger picture of AI, and it was so rewarding to explore how these principles are being applied in sci-fi areas like robotics, digital twins, and virtual health assistants. Seeing the businesses emerge from these innovations was even more exciting, and it’s fired me up about this space.
This is my first ever significant industry exploration, and I would love feedback. I’m a 19 year old college student and have a long way to go when it comes to analyzing industries and companies, especially those who are run by PhDs in subjects I’ve just scratched the surface in. I’d love to see how I can improve, and the only way that comes about is through feedback. Feel free to email me at firstname.lastname@example.org or DM me on Twitter.
I hope you’ve gotten a bit more knowledgeable on this exciting intersection of healthcare and AI/ML, two areas that are and will become even more center stage in our lives!
Thanks for reading, and you’ll hear more from me soon.
- Date of publication:
- Thu, 11/25/2021 - 13:10
Click on the link - it will be copied to clipboard