Training Weka segmentation gradually

Hi everyone,
I am currently using the Weka plug in, and it works pretty well actually. Nonetheless, in order to perform a robust analysis of my data ( drosophila embryos happily growing) , I want to run the same classifier on each of my images stacks. But even if the segmentation is still ok when I usemy classifier on a different embryo, I was wondering if I could re-train my classifier on a batch of other embryos in order to gain in robustness. If I do so, will I lose the work from my first training? Or are the info from the first training set kept in the classifier file?
Thanks in advance for the help,
Baptiste Tesson