How can the mantissa of float, double be determined: is their a standard definition?

Is there a standard manner to determine the mantissa of a double? I'm aware that IEEE-754 format of a double stores the mantissa in 53 bits, but I'd like to avoid the magic number in my code. On ...