#027 AI in MedTech: Challenges & Opportunities w/ Adam King

LISTEN : Podcast | Youtube | LinkedIn

LINKS

• Connect with Adam King on LinkedIn
• AI in MedTech courses mentioned
  • Harvard AI in Healthcare (Harvard Medical School)
  • Technology and Innovation Acceleration Program (MIT xPro)
  • AI For Everyone (coursera)
  • AI in Healthcare Specialization (coursera)
• Get show updates at mannybernabe.com

NOTES

• Ambulatory surgery centers—known as ASCs—are modern healthcare facilities focused on providing same-day surgical care, including diagnostic and preventive procedures. (Ambulatory Surgery Center Association)
• Use Cases
  • AI to route appropriate care for cardiovascular-related ASCs.
  • AI to optimize for total cost of care by leveraging ASCs and other lower cost options. Route patients to the best healthcare service.
  • Using AI to review imaging to find new insights to recommend interventions early enough to make a difference.
• You’ll need more to more efficiency optimize case management as you leverage lower cost options like ACSs.
• This data that is already being generated and captured. Challenges exist with data silos, combining structured and unstructured data.
• Digital twins (”simulation”) may be used to make better decisions, right interventions, and recommendations. Potential use for pharmacological (drugs) and anticipating outcomes from treatments and recommendations (i.e. diet changes, nutrition, etc.).
• Wearables (whoop, Oura Ring, Apple Watch) present a great opportunity. Great way to complete a picture of a patient. Challenges existing in integrating into healthcare systems seamlessly and safety.
• Key challenge in MedTech AI is measuring the effectiveness of treatment. Wearables might be helpful here.
• Key question is can we identify the right intervention at the right time. I.e. You have a very limited window to alert on a stroke.
• Consider how we are going to manage precision versus accuracy. Getting the right balance is key.
• Brittle models are another limitation of healthcare. Will need to model post production.
• For enterprises leaders looking to get started with AI, here are some options:
  • Massive Open Online Course (MOOC’s)
  • Executive Grad Programs (i.e., MIT executive class)
  • Internal company programs (Medtronic’s “AI Considerations for Management”)
  • See links above
• Regulatory is a major challenge. AI is new to industry. It’s REALLY new to the FDA.
• Annotation (or generating labels) for data, in particular imaging can be expensive and time consuming. Easy to under appreciate this part of the AI work flow.
• AI is about augmenting existing medical professionals. Needs to be framed as such. Key question is how much do they trust AI recommendations? Measuring trust will be important to new tools.
• Put the AI recommendation in the context of what is generating the prediction. “Show your work.” This will help build trust with the end user.
• The guiding principle of AI in MedTech, “What is the ultimate impact we want to have on the patient?”