Some termites produce noise when chewing. The noise becomes detected by dogs, which can be trained to listen to them and report insect attacks. Possibly they could also be trained to smell them!
With modern speech recognition systems I can imagine that they could be detected and the typical noise of those insects when eating could be recognized (by its frequency, its modulation ...)

seems an ultra sound would detect the voids created.. albeit maybe to late.. 

There are some patents in that sphere, but even they require scraping off a huge chunk of bark for the coupling to work. Ideally we would like to have something that does not expose the tree to infection when we are screening. (Also, as you say, probably hard to detect until it is too late.) One of the things that is challenging is that unless you have a baseline for each tree it is hard to detect small changes...

this Berkeley study used hot glue to detect termites , https://nature.berkeley.edu/upmc/documents/Sensor Performance in Acoustically Detecting Drywood Termites.pdf

and yes, to late tree in trouble

Have you tried looking at it with a thermal camera? Or a UV camera, too. It's possible the signs might be a little more obvious when viewed in different spectra.

 Joan Horvath

wrote a few seconds ago
Hi-  We tried one of the PublicLabs IR webcams,  and the bottom line is that IR doesn't seem to help particularly, at least at the level you can achieve with a low-cost IR  webcam. Our collaborator also borrowed a UV flashlight and is in the process of 
trying out various things with that. Because the holes are so tiny (a little over 1 mm) and the bugs burrow so deep so fast, anything that is  just surface observation, we think, is not likely to detect them directly. It would however be awesome to be proven wrong on that one.