Normalization in PET group comparison studies—The importance of a valid reference region