Using persistent homology to reveal hidden information in neural data