In the code ($a =~ /(\d+)/)[0] <=> ($b =~ /(\d+)/)[0] the regular expression is used in a list slice with a single index which returns a scalar value which can be compared with the <=> operator.    Without the list slice the regular expression would return TRUE or FALSE in scalar context.

In the code my($aa) = $a =~ /(\d+)/; the parentheses around $aa impose list context on the regular expression which in list context returns all the contents of capturing parentheses in the regular expression.