Artificial Intelligence in Health Care - Stat€¦ · Artificial Intelligence The editorial content...
Transcript of Artificial Intelligence in Health Care - Stat€¦ · Artificial Intelligence The editorial content...
Artificial Intelligence
The editorial content in this presentation was written and produced by STAT with no participation from sponsors.
in Health Care
Investment in AI health startups
● mainly used in the back
● Clinical use is increasing
● Drug makers are using AI
● Financial costs:
● Privacy and ethical costs:
● Reputational costs:
● Imaging, imaging, imaging:
● A recent example:
● The benefit:
● pump the breaks
● AI might play a role in quality assurance and help surface research
●
● AI comes in many forms
● Two main tasks:
● A third use:
Machine learning:
Deep learning:
Neural network:
Natural language processing:
●schedule its operating rooms
● detect the onset of sepsis
●
target interventions to particular patients
● improving the R&D process
●examine the mechanisms of
disease
●
● helping with recruiting
processing and cleaning of data
● incorporate real-world evidence
● significant roadblocks to overcome
● The goal:
●
● The possibilities are enormous,
●unrealistic expectations
●skepticism and
anxiety
● slower adoption
●built in incentives to aggressively market
the use of AI
● sell their products
●work is cutting-edge to attract patients
●co-branding arrangements
●
● needs to be grounded in facts
● dispel hype, not peddle it.
The Gartner Hype Cycle
Source: www.gartner.com
POLL QUESTION
● two extremes
● replacing doctorscuring terminal disease
● the truth becomes much harder to suss out
● AI is designed to augment human capabilities
● doctors can focus on tasks that matter
● Human oversight will continue to be needed
● Progress is slow and incremental.
● Use in clinical practice remains largely experimental.
●
21st Century Cures Act
●products can be exempt from
regulation if users can “independently review
● has not clarified what it means
● left to determine on their own
● commercialization of IBM’s leading cancer product, Watson for Oncology,
●required to do prospective studies
POLL QUESTION
●not enough to enable AI systems to
generate new insights
● A greater diversity of data is needed.
● Providers use different EHR systems
● harder to link data sets
● Solving this is crucial
●not perfect
●
● data must be properly obtained
● hype is a continued threat
● need for sound science and
prospective studies.
● meaningful partnerships
● long-term vision
rapid increase in AI use
reason for optimism
vigilance will be necessary
clearer picture of winners and losers