The study of radio-mechanical AGN feedback provides powerful insights into the energetics of early-type galaxies and galaxy clusters. The total energy released during individual AGN outbursts has been imprinted into jet-inflated bubbles (X-ray cavities) and can be derived by estimating the total cavity extent. To this day, studies of X-ray cavities focused mainly on individual objects, and the cavity size estimates were ultimately based on visual inspection of noisy X-ray images. We are systematically studying an extensive sample of >150 galaxies and clusters containing X-ray cavities. We will present preliminary results of this analysis showing correlations with other galaxy properties (SMBH mass, thermal state of hot atmosphere, etc.) and comparing individual cavities of multi-cavity systems. Furthermore, we will present a novel machine-learning method called the Cavity Detection Tool (CADET), developed to allow an automated and reproducible study of X-ray cavities.