Investigating depth-dependent transformation of vibroacoustic signals originating from tissue interactions during needle insertions

Abstract Clinical needle placement requires accurate information about tissue boundaries and transitions, which are often not reliably resolved by standard imaging modalities due to limited spatial resolution. Vibroacoustic signals can encode such information and could complement existing modalities with real-time feedback, but their signal composition and the influence of insertion depth remain unclear. By applying physical concepts from established force-based needle insertion models we assess whether tip-related excitations contribute characteristics relevant for tissue classification and whether shaft-interactions alter these components with insertion depth. A ResNet-18 model was trained on tip-dominated interactions and tested on the same tissue at the bottom of a stacked phantom, to assess the shaft contributions and depth-dependent effects. The model failed to classify similar tissues beneath an overlying layer. However, augmentation with shaft-dominated recordings elevated classification performance over 90%, indicating that tissue-discriminative information remained present. As augmented signals did not reproduce spectral characteristics of depth-affected signals, shaft interactions alone were insufficient to explain observed transformations, suggesting additional modification by depth-dependent propagation effects including attenuation and damping. This study proposes a conceptual model in which vibroacoustic needle signals are interpreted as a depth-dependent superposition of tip- and shaft-related excitations, which are modified during propagation along the shaft.