There isn't any existing work, but basically all you need to do is the following:
modify the pdf metadata reading code (in calibre.ebooks.metadata.pdf)
calibre contains a nice library for pdf reflow that converts pdf to xml use that and then search for the ISBN in the XML
Basically:
Code:
with CurrentDir(temp_dir):
pdfreflow.reflow(stream.read())
will create index.xml in the current directory