Robotic ultrasound-guided interventions, such as breast biopsies, suffer from target inaccuracies due to probe-induced deformations. Real-time estimation of this displacement is essential to improve the accuracy of these procedures.
Therefore, this research developed a hybrid mass-spring Position Based Dynamics (PBD) Tissue Deformation Model (TDM). This model supports force-based, position-based and mesh-based perturbation inputs and can be initialised either from a predefined geometry or from stereoscopic depth camera point cloud data. Performance was evaluated through both in silico parameter sweeps and an experimental setup utilising a Franka Research 3 robotic arm, a Telemed ultrasound system, stereoscopic depth cameras, and a tissue-mimicking phantom.
In silico validation demonstrated that the proposed TDM produces stable deformation simulations across varying perturbation strategies, with PBD constraints successfully introducing the necessary non-linearity and stability. However, a reliable ground truth for lesion displacement could not be established due to the physical phantom's structural constraints, hindering experimental validation. While quantitative physical validation is missing, this research successfully establishes a modeling framework with potential for predicting probe-induced lesion displacement in robotic ultrasound procedures.
