Partially-automated feature (region of interest) detection

I have a reference image with 4 human-discernable features (regions of interest). These occur in several hundred test images with varying scales, orientations, and deformations.

Looking for ways to define these areas of interest from the master image and automatically detect them in the test images.

I was thinking of using SURF, but I already know what I'm looking for, so I wasn't sure this would be the most efficient method.

I am new to MATLAB/computer vision, so thank you in advance for your patience.

[Chloe Matthews]

Hi Nina, although i cannot help you with this question, i myself am stuck on something very similar at the moment. I have multiple magnetic resonance images and want to crop a region of interest around the muscles in the lower back but some of the images are positioned in different ways (due to patient moving between images). I need the cropped regions of interest to be of the exact same part of the anatomy in every photo but do not want to manually do it for each picture!

Hope that made sense! I am wondering whether you found an answer to this question and if so if you could somehow give me some help?

thank you in advance!