Yes it is possible, we have several single-case studies done this way in our lab. It's just that of course you will suffer from a class imbalance, so the results should be interpreted very carefully, but it should work. Also it would be a preferable to use non-parametric statistics as they were updated with this patch to work with such a design, and arguably a permutation test would be less prone to error rate inflation than parametric statistics (although because of the class imbalance and the lack of information - a single subject cannot represent a population - there is still a bias that cannot be corrected).

The patch for the non-parametric test to work correctly with single case studies is here:
https://www.nitrc.org/forum/forum.php?th...

Best regards,
Stephen