2023 Danaher Summit Highlights: AI and Data
May 30, 2024
On data as a new kind of resource:
There is a common phrase: "Data is the new oil, AI is the new engine." I hate this phrase. Because as we know, especially in California, oil is grabby-grabby. Either I have this barrel of oil, or you have it. We can’t both have the same barrel of oil.
Data doesn’t work that way.
I might have a dataset and create something magical with it. You might have that same dataset and create something differently magical with that. So as another one of those TED talks says, "Data is the new soil." You plant your ideas in, and data helps them grow.
Atul Butte, Priscilla Chan and Mark Zuckerberg Distinguished Professor, Director, Bakar Computational Health Sciences Institute
On the challenges of working in a low data environment:
For us in pediatrics and in the emergency room we have the other side of the data problem: We don’t have enough data.
For instance, I might have a one-year-old that comes in who can’t tell me [or their parent] that they have a severe headache. Can't tell their parent the same thing. How do I know what exactly is going on with this child?
Or the EMS team calls in, or an ambulance calls in, and says, “I have an unresponsive ten-year-old.Heart rate of this, blood pressure of this, we’ll see you in two minutes.” So I’m working with no data or very limited data— this is where [bedside] diagnostics really come in.
Kemi Badaki-Makun, Professor of Pediatric Emergency Medicine, The Johns Hopkins Hospital
On the need for actionable insights at the point of care:
From a practical perspective, doctors are under an enormous stress and we shouldn’t lose sight of that. They’re asked to do more, absorb more and integrate more, with less time and more pressure to be productive...
Whatever data comes down, I think we need to increasingly rely on the electronic medical record to distill and provide actionable information [to doctors]. Because in the middle of a 15-minute follow-up visit, the doctor cannot possibly integrate all the diagnostic information alone, so we really need to rely on informatics to help distill the critical information and make actionable decisions at the point of care.
Scott Friedman, Dean for Therapeutic Discovery and Chief of the Division of Liver Diseases, Icahn School of Medicine at Mount Sinai
On AI’s growing role as a partner for providers:
What AI needs to be is a really, really smart partner...
An interesting study of breast cancer patients in Denmark found that at each diagnostic step, if you had two AI systems diagnose [the patient] versus two pathologists or radiologists diagnosing the patient, versus one and one—a computer and an expert—the computer and expert together were 20 points better than either two people or two computers.
Mara Aspinall, Partner, Illumina Ventures
On the benefits of using existing data to train AI models:
A huge driver of innovation will be generating new insights from data that’s currently being ignored.
[Physicians ignore data because they] aren’t trained to use it. Whereas new deep learning models—foundation models, can turn that [data] into new insights. Then, [the data is] meeting user needs and delivering higher quality products at lower cost.
Andy Beck, CEO, PathAI
On the need for better analytics:
I call [it] the "data-insight gap." Essentially, data has been growing exponentially; in biology, insight grows linearly. Every day looks worse: the percent data utilized is lower.
The answer is to invest a huge amount on the analytics side, rather than on the data generation side.
That’s the revolution we need to justify the generation of that data.
Shai Shen-Orr, Associate Prof. at the Technion – Israel Institute of Technology, Founder & Chief Scientist, CytoReason
On the importance of integrating new sources of data:
The biggest public health crisis we’re going to face is neurologic diseases and neurodegenerative diseases.
Early intervention and early detection will be key but the [current] tests are not great — they’re OK. And the drugs will be really expensive. So we need to consider wearables and other things: either cognitive tests online to detect early cognitive decline or movement disorders that can be detected with watches... these things exist.
As a diagnostic laboratory, we must consider the information [these wearables and other things] are creating, and how to put the data into an algorithm that could trigger a more confirmatory lab test.
Bill Morice, President and CEO of Mayo Collaborative Services and Board Chair, American Clinical Laboratory Association
These quotes were edited for length and clarity.
