but I have no idea where I should start.
1) start with reading what osteoporosis is and the known ways to detect it. include the 1st and 2nd order hypothesis error statistics
2) determine what kind of sensors are suitable / could do this kind of detection. range etc
3) determine what is the datastream from those detectors? how to analyse this stream? how much processing power do you need?
4) Can an Arduino do this?
Answer (1) might include: rontgen, blood testing, BMI, specific mass, statistics chance = f(age, sex, length, weight, form of feet, family, ... ) .....
Answer (4): I guess not (as I expect heavy processing), but you can allways add an Arduino to the system with a LCD displaying "YES/NO : uncertainty %"
Q: do you want to determine osteoporosis in living beings or check old bones for osteoporosis ?
NB dead material can be tested in a sort of break bench and determine at what force it would break - Arduino could control such device. (for living beings this is not ethical)
Q: what is your medical / programming / electro knowledge?