About the only method that I've seen that doesn't require spike-ins is the one described in this paper, which is actually fairly similar to what I think a number of us had been informally recommending to people (nice to see that someone made a real paper out of it).
By significant you mean genes that are expressed or differential expression between samples?
I means the genes that are expressed.
You might want to add that as a comment instead of an answer as now this post is listed as "Answered".
On the other note, do you have any ERCC control within your samples? If not, people usually just arbitrarily select a threshold for "expressed"
Thank you for the reminder :) , I dont have any ERCC control within my samples and I will select a threshold for "expressed".
You might want to read this post: How Do You Justify Your Rna-Seq Expression Threshold (Fpkm/Rpkm) ?