Who can see your viewing activity?
Aloha from Maui everyone!
Break the silos!
Perfect day with AI and live Maui backdrop! Thanks to all the organizers and participants for getting this together.
Rising cost of healthcare is a big threat!
The 2 papers I mentioned
A question to the panel. Over the last 2-3 years we set out to create a cross-specialty data infrastructure to capture every data point possible from all our labs, specialties and Cerner. The purpose is to create a digital twin and make curated data available for AI research. COVID helped us get the ethics approval required, perhaps the most important milestone. The issue we're deliberating on now is the choice of technology and future proofing. I am very keen on open source everything and maximizing interoperability ad am pushing in that direction. Any wisdoms on how best to do that?
Good question Zeidon.One word answer: Persistence!
Trying different approaches and addressing each barrier.
Is any centre presently using ‘silent implementation’ of AI solutions?
Yes,many are Avneesh in trial mode or for validation.
Suppose you have an AI algorithm you want to implement in clinical workflows. What general infrastructural components must be in place to bring it to the bedside, and how can we systematically provide such infrastructure to both the ivory towers and to underserved populations?
Great question Tim,I firmly believe there needs to be translational group which helps with implementation and is resourced adequately.Human computer interface expertise is very much needed there.
Its one of the most underestimated problem of AI in healthcare.
For the second part, we are not there yet…..
For clinical AI, if AI is clinical decision support, and considered by FDA as a device if its not highly understandability, what is best practice to deploy. Do you use disclaimers liberally?
Yes maybe that’s why I read some awesome academic papers but go back to facing similar bedside problems everyday. We need better implementation on ground.
I always like the pacemaker analogy :)
A biochemistry analyser is a blackbox too. We put a blood sample and it generates a set of numbers that we trust without questioning how it works. We just trust that it has been "calibrated"
Validity is more important
If it has been determined that it is a device, you should conmply with the appropriate standards
Thanks for the answer Piyush. It sounds like implementation will be very resource-intensive as long as we are all using different systems. You can’t just write a new algorithm and a new interface into the EHR system and push it out like an iPhone update if everyone is using different systems, right? So is the underlying problem that we are all running such different systems? Is there any way around that, or will we eventually have to converge on the same systems in order to democratize AI implementation?
Great analogies 👍
Doesn't the FDA care if its a black box though?
ABAIM and MIS workshops were really monumental in shaping my understanding of AI in Healthcare.
Our brains are also black boxes.
I totally agree with you Fatme!
Great questions Tim.I think the issue is that we are too early.Just like 20 EHRs and 100s of deployment and IT centers, this will also evolve.The cost-effective models will survive.
Also,Darwin will hold true and it will be survical of the fittest
Great insights Dr Piyush
Even though it is full of cognitive biases, we trust it
Clinician needs to communicate better and more with data scientist.
@Arlen So should we hold AI algorithms to higher standards, somewhere near that of devices? I personally do think so to an extent
Aren’t we all biased!
@Alfred - I agree
Eric- great question.The answer lies with development of CPT codes.I can elaborate on it being a member of AMA’s CPT editorial committee
There is now a CPT code for AI in Radiology but who will pay for it?
Stay safe everyone.Another amazing session.Learnt a lot.
Thank y‘all! . Stay safe!
Thank you MIS and all the great speakers and panelists. Had a great time!
Thanks to attendees, presenters and moderators for a great first day
Mohammed Hassan Mohammed
Thank you for great talks and discussions
Its 2am here in UAE. Have a good nite. Stay safe!
Thank you all. Great discussions.
Another incredible day with the MIS crowd. Thank you all for your insights and camaraderie!
Great course programming, very diverse perspectives, its very inspiring to see you all!!
Stay neural networked everyone!
Thank you!! Will see you again tomorrow.
excellent session ! thank you
Eagerly looking forward for the Day 2 tomorrow
You all made my day :) Thank you so much! Looking forward for tomorrow
Great meeting today! Thanks Julie and David for all of your work behind the scenes.
I attest to that Anthony! Thank you for the opportunity
Thank you and looking forward tomorrow.