How do I perform a DWT on a cluster?

I need to find out if a given brain MRI image contains a tumor or not. In order to do that, I am using a clustering algorithm (SLIC), from which I obtained the picture presented below. After this, I pretend to get a number of features from each cluster using a Discrete Wavelet Transform. This features should then be analyzed by a Support Vector Machine, that was previously trained to identify clusters that correspond to tumors.

Question 1: How can I feed an irregular shape image (cluster) into the DWT function (dwt2)? I'm admitting that I can't pad the cluster with zeros till I get a rectangle, because the zeroes will affect the results (cA,cH,cV,cD) given by the dwt2.

Question 2: Do you advise me to feed each cluster individually into the SVM or do you think it's wiser to train the SVM with the tumors themselves (that would represent class 1, whereas the rest of the image would represent class 2), and then give the SVM an unknown whole image for it to find out if there is or is not a tumor in it. Question 1 will still need to be addressed if this approach is followed...

Thanks for your time