Matching of any medical images between patients or mapping patient's medical images to the simulated atlas is key to an automatic registration as well as a useful tool for indexing and searching among multiple patients or visual aids that could be used for training of medical staff.

Here we review existing feature extraction methods and quantify their usability to match medical images and describe preprocessing steps which help in this process. To be able to compare each methods and their sensitivity to different image degradations (rotation, scaling, noise) we present simple metrics alongside.

For now we have tested feature extraction/detection methods like AGAST, AKAZE, BRISK, GFTT (Good Features To Track), HardNet and ORB on MRI brain images.