I'm working on features extracted from heartbeat intervals based on Fantasia database (http://physionet.org/physiobank/database/fantasia/) and Congestive Heart Failure dataset (http://physionet.org/physiobank/database/chf2db/). My question is regarding whether these features extracted from the Fantasia database (which comprises a specific experimental protocol) could be biased when compared to same features extracted from Congestive heart failure (whithout any protocol)?