Few days ago I landed to this page https://www.poliambulanza.it/news/intestino-in-poliambulanza-la-prima-videocapsula-con-intelligenza-artificiale
This is a nice example on what can be done when physicians and engineers talk together.
Seems easy but is not at all.
Different discipline not only have different languages but have totally different approach on problem analysis and solving.
The difference it’s huge in this case because doctors works on human and each case is different.
Engineers works on systems, collects requirements, need numbers,… try deduce a rule to control the system.
How much “red” should be a spot to be considered bleeding, how big is a bowel lesion…
Don’t ask those questions because the answer would be: look, this is bad, this is just a change of tissue texture.
Of course rules exist, literature and scientific papers can guide to identify what’s most relevant.
But that would require huge effort to Engineers with sub-optimal result because we simply can’t learn about a disease just reading papers.
When Engineers learn to talk to non-engineers and start asking right questions and Physicians realize that any automatic system is built with numbers they can start producing exceptional results.
That’s why interdisciplinary communication is the success key.
I had similar experience working on image processing from #videocapsule for semi automatic diagnoses about 10yrs ago.
Now AI really helps because you don’t have to ask such annoying questions to doctors anymore.
The inference engine can deduce what’s relevant with a “learn by examples process” analyzing the manual classification done by an expert doctor.
It’s easier ? not at all again.
The design of the data set (examples) requires huge effort of syntheses and most of the time a new discipline jump into the game, typically a mathematician.
This simply means communication is being more complex.
Again, interdisciplinary communication is the success key.
Talk less, Listen more