De novo exon quantification from BAM files

[nsalomonis]

I am wondering if there is software that can predict which exons are present (clustering of reads) from TopHat BAM files and return associated counts for those predicted exons. I know BEDTools will do this for supplied known exons and that Cufflinks predicts novel exons, but I'm not sure how to get novel exon predictions and associated counts. Without these, I wouldn't want to trust novel junction predictions from programs like TopHat (need confirmatory data that a novel exon is present and expressed).

Thanks in advance!