Cardiac shockwave therapy (CSWT) is a promising non-invasive treatment for patients with refractory angina, aiming to stimulate myocardial regeneration by delivering acoustic energy to ischemic heart regions. However, the therapeutic CSWTprobeusedin this work has limited spatial resolution due to their low element count and large crystal sizes, making it difficult to detect and avoid bone structures such as ribs that can obstruct wave propagation towards the therapy target. This study presents a proof-of-concept algorithm designed to detect bone structures using low-resolution RF signals acquired with the therapeutic probe. The approach was implemented through ex-vivo phantom experiments using pork ribs, signal preprocessing, feature extraction, and machine learning algorithms. Multiple models were evaluated, achieving test accuracies above 82%, with Support Vector Machines and Ensemble Trees demonstrating strong performance. The standing-out models were able to reliably identify blocked elements in most probe positions, offering a potential method to guide probe orientation and optimize therapy.