Been someone working on the development of healthcare AI (mammography) on a developing nation (Brazil), I do get your pain.

The truth is that healthcare is too rigid to openly accept AI, but it is also a really noble area to work on for publicity sake. What big companies look for is the publicity of being "changing the world", yet, you rarely see any deployed AI doing real health work. Likewise, Google poses as ethical and ethics-oriented, but fired one of their most famous ethics researcher for trying to publish a paper that could harm their reputation.

Finally, this endless chase for beating SOTA is as real on healthcare as it is on ImageNet. People are not trying to solve issues, they solve datasets. No wonder why most algorithms, when deployed, serve no one or are too opaque to be trusted by any medical staff.

If you need more proof, take all the effort towards a COVID-related AI. Where did it go, what did it accomplish? COVID is almost two years old now.