Attention is altered even in subacute schizophrenia and can easily investigated with an e.g. auditory oddball paradigm. Very often you'll find a smaller and delayed peek amplitude of the P300 for the "rare" condition.

... should be spelled "peak" and not "peek" of course ...

Why not run an unsupervised model first and identify potential features? You can then cross-reference what you will find with common-sense and literature.