I might be missing the details here, but everything outside the breast should be pretty much black and will get compressed very efficiently (high signal, low noise, easy to predict).

Foveated compression sounds like a super cool idea I've never thought about before.

Before that, you would need to get radiologists onboard with the idea of something being "diagnostically lossless" (think visually lossless), which is currently a hard sell (some promising research does exist on this)!