MPN Workshop of the Carolinas 2025 | Key barriers to the clinical adoption of AI tools for MPNs

Great question about the key barriers, and the truth is, there’s quite a lot. You know, as I mentioned before, we’re in a very, very early stage of the development of these tools. What we need to see is that these tools are valid across not only the institution that made these tools, but institutions that have never seen this tool beforehand. Because we know MPN patients are highly diverse and highly heterogeneous, both clinically and genetically, so there are many parameters that may make an AI tool biased, and that the accuracy that we initially see may not be the accuracy that’s in the real world. So we have to be very, very careful to make sure that what we believe is going to happen actually happens. So that’s number one about generalizability. 

I do worry about patient safety and privacy, particularly as these AI tools, especially in the generative AI space with chatbots, I want to make sure that patient privacy is maintained and that PHI or private health information is not freely available in the public space as we try to use these generative AI models. 

Finally, I think that the regulatory agencies, including the FDA in the United States and the EMA in Europe, they will have their final say about how they view artificial intelligence tools. That’s still a moving target right now. The FDA has released a draft about how they view AI tools, especially for commercial deployment, but the final recommendation is still to be decided. So there’s more to come on the regulation aspect. So there’s a lot of things that we need to consider before we use this in the real world for our patients.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.