The additional detection of sleepiness waves and PUI is a demanding task.

We could perform a spectral analyses using short-time fourier transformation (STFT) on pupil diameter.

To avoid aliasing-effect it would be best to perform STFT on raw data (120 Hz sampling rate).
Lowering the sampling rate is quite tricky and would require an appropriate low pass filter for downsampling of pupil diameter. Otherwhise aliasing-effect would severely distort the frequencies.
Considering potential frequency distortion and aliasing-effect I would recommend to avoid downsampling.

Due to the room illumination of 0.10-0.12 lux during MWT, eye movements, respectively movements of focal point to darker or brighter areas, could potentially affect the pupil diameter.
To distinguish between light induced pupil reactions and sleepines waves, it could be helpful to get information on the eye movements.

What would you think is the best way, to handle potential aliasing-effect and light induced pupil reactions? And which data are avaible?