Taking A Closer Look at AI-Driven Biomarkers
By Diego Arias - Plug and Play Techcenter
What are biomarkers? Biomarkers include everything from blood pressure, heart rate, basic metabolic studies, x-ray findings to complex histologic, and genetic tests of blood and other tissues. From clinical decision-making support, new drug discovery, and underwriting risk assessment, biomarkers can help lead innovation towards early detection and significant cost savings for payers and large self-insured organizations.
At Plug and Play, we put together a panel of industry experts in insurance, pharma, health systems, and emerging startups to discuss trends, future implications, and applications of AI for biomarkers within their respective industries.
Debbie Lin is an executive director at Boehringer Ingelheim Venture Fund (BIVF), where she leads digital health investments. She received her scientific training at UCSF and has worked in various areas with BI for more than 13 years, including spending time internationally and leading the company's global initiatives in corporate strategy and development.
Sven Ebert studied mathematics, has a minor in economics, and holds a Ph.D. in Probability Theory. He's currently working as a Head of Product Development for SCOR, one of the world's largest reinsurers, and in his role, he's taking a close look into the technology-driven trends of the reinsurance sector. He is responsible for the development of activity-based products.
Dr. Zeev Neuwirth is the Chief Clinical Executive for Care Transformation and Strategic Services at Atrium Health. He has over 15 years of experience in client clinical operations quality and process improvement, population health, as well as care design and innovation. He is the author of “Reframing Healthcare: A Roadmap For Creating Disruptive Change” and produces and hosts the popular podcast series, “Creating a New Healthcare.”
Why Biomarkers Are Important
The conversation kicked off with a very general question: why are biomarkers of relevance for every insurance company?
Sven Ebert explained that SCOR has looked at biomarkers as a solution to keep people healthier in the long run. Biomarkers allow them to measure the progress and improvement of certain things like a decrease in the resting heart rate, or higher active calories burned, per day, or very simply the number of steps taken.
From a VC angle, Debbie Lin explained that BIVF’s investments are strategic to the interests of their pharma company. Having access to biomarkers can help pharmaceutical companies better develop drugs, understanding whether patients do well during clinical trials, and properly stratifying them. Pharmaceutical companies want to use biomarkers to find novel ways to better understand how patients do through their disease and to monitor them on medications.
Dr. Neuwirth explained that the importance of biomarkers lies in identifying bad outcomes before they happen. Using biomarkers for prediction, and being able to know if someone is at high risk of having an acute event ending up in an emergency room, should become the standard of care.
Training AI-Driven Biomarkers: Things to Keep in Mind
During the event, the speakers posed some things to keep in mind when training AI:
- Ebert touched on the importance of avoiding cherry-picking data and building rich models that also respect laws and regulations. Dr. Neuwirth agreed: it’s key to collect data from multiple institutions and multiple data sets.
- Diversity is essential. It’s important not to focus on 50-year old, white men, and make sure to cover different backgrounds.
- When drawing data from large data sets coming from multiple organizations, privacy and compliance are also an issue.
- Bias problems in training algorithms are a huge problem. Lin mentioned that an iPhone in a purse and an iPhone in a pocket track steps differently, and looking out for these biases will be essential in building the databases needed for AI development.
AI-Based Biomarkers: Relevant Solutions
The conversation moved forward to address the most promising solutions in the biomarkers space.
Lin explained that the epidemic has had a silver lining and has accelerated the acceptance of remote monitoring tools, wearables, and telemedicine. This generates a wealth of data, which increases the speed of biomarker development. For instance, Lin mentioned that cameras can be used to detect dementia and Alzheimer’s disease. Through a retinal scan, a camera can detect biomarkers in the eye, to predict whether someone's going to have dementia or Alzheimer’s disease or not. The FDA has lowered the barriers, granting breakthrough designation, which has sped up innovation.
With AI and ML solutions scaling and expanding in this area, we also wanted to hear from entrepreneurs that are tackling these issues across many use cases in the healthcare value chain. That is why the second half of the event featured four incredibly innovative startups:
Predictiv offers DNA-based Digital Twins for healthcare. Their solution analyzes the patient’s entire DNA and compares it to 16,000 genetic diseases. This predicts which genetic diseases the patient is most at risk for and enables them to take preventive measures to prevent and decrease the impact of the disease.
AI that detects and analyzes cough. Hyfe is a Cough Detection Algorithm that runs on a smartphone and tracks coughs. Hyfe uses machine learning algorithms trained on millions of labeled sounds to detect coughs from ambient sound and track them over space and time. The end goal is to improve it to the point where it can distinguish between different types of coughs and become a powerful diagnostic tool.
AI image recognition for cancer diagnostics. Mechanomind is making cancer diagnosis fast, accurate, and accessible everywhere with AI-based detection and classification of tumor types, and creating a global marketplace for consultations and second opinions. The goal is to solve the shortage of pathologists worldwide, improve accuracy and patient safety, scale the best diagnostic expertise to underserved markets, and reduce healthcare costs.
Virtual patients for data-driven clinical trials. Virtonomy is creating the first web platform for use by medical device developers that utilizes virtual patients for data-driven clinical trials, thereby shortening the time-to-market for life-essential medical devices, accelerating medical innovation, and significantly reducing costs by up to 50%.
After the startup introductions, there was a brief discussion that you can watch here.