The 5534 is very good for an opamp, 3.5nV/root-Hz, but for cutting edge noise performance you want more like 1nV/root-Hz, and there are only a handful of op-amps that can manage that - individual low-noise transistor differential input stage would be cheaper. Make the input stage have lots of gain so that the next (noisey) stage can't add more effective noise to the signal (30x voltage gain perhaps).
And no resistors in the input signal path, they just add noise (all resistors generate noise). Feedback network needs to be very low impedance to reduce noise (perhaps 1k / 33ohm resistor divider) - the noise in the 33 ohm resistor will add to the signal but 33 is low enough (noise voltage proportional to square root of R, in fact about 0.13 sqrt(R) in nV/root-Hz, thus 33 ohms is 0.75nV, 1k is 4.1nV)
To get best low-noise performance from a transistor you have to set the collector current at the sweet spot for the device in question, often around 0.1mA.
Google for suitable circuits and look for noise performance near the 1nV/root-Hz range.
Incidentally 1nV/root-Hz is about 0.14 uV RMS across the audio band. 4nV/root-Hz is 0.56uV RMS... For 60dB S/N ratio the Vrms of the signal needs to be 1000 times the noise Vrms.
If you do want op-amp design I can recommend the AD8656 dual opamp - 2.7nV/root-Hz (better than NE5534 and low-voltage rail-to-rail so can use from single 5V supply).