we are using matched 100ohm platimum RDTs in a wheatstone and compensated for non-linearity in a separate circuit. Amplification is done through a high CMMR instrumentation amp. the rtds are pulsed using a precision current device to keep self heating below a couple 0.01mA.
I understand that 24 bits of dynamic range is a lot, however for an initial experiment it is well justified because then we have direct proof of the circuits ability to operate at a specified bit resolution. Shoot for the moon, end up in Chicago... i guess.