Earlier quoted context omitted.
Or just crowdsource the reading! It would be really interesting to check out the aggregated quality of this kind of technique.
If by crowdsource you mean have multiple dentists look at every image set, then maybe. But if you're suggesting that a random selection of people look at dental x-rays and guess where the cavities are... that would be an interesting experiment, but I wouldn't expect too much. In either case, I doubt you'd find much traction since looking at dental x-rays isn't something most people are going to want to do on a regula…
First, however, you would have to find out whether lay people are any good at this task.