Medical devices would be just on the upper bound of what I think YC could do. The right way is probably to do an unregulated "fitness" or "convenience" product, which later has regulated-medical-device functionality. Defer the compliance parts until the product itself is proven.
There are plenty of "small business" ($1-5mm in personal-recourse debt) type medical device entrepreneurs. US and Israel seem to be the two big markets for developing them. Surprisingly, a lot come from people outside the medical professions.
A medical device which had applicability to the military (specifically, IEDs, TBI, PTSD, eye/hearing protection or remediation of damage, traumatic amputations, or prosthetics) would largely avoid the FDA, too. There is going to be so much money in improving the lives of wounded veterans from Iraq/Afghanistan for the next 50 years, and it's going to be one of those cases where money you earn is also related to a societal benefit.
If I were in biotech/biomed or robotics, that'd be what I'd want to do; the only comparable-scale problems are reducing cost of care for everyone and dealing with an increasingly elderly population.
The ubiquitous computing plus sensor environment is making healthcare hardware startups the norm—we're seeing more and more applicants in this space at Rock Health. We don't think the FDA process is that onerous, and just published a report outlining the process for entrepreneurs who are new to the space.