Volcanic risk escalates significantly during unrest. In late 2021, the Italian island of Vulcano entered into a phase of unrest featuring Very Long Period seismic events, which are considered to be markers of magma and gas flowing across the volcanic plumbing system. Here we show how Neural Network Nodal Ambient Noise Tomography generates a high-resolution shear-wave velocity model for investigating the causative drivers of Vulcano's unrest. Using a deep learning model we harvest seismic dispersion data from a dense nodal seismic network deployed during the early unrest's phase. The inverted 3-D model reveals a high-resolution tomography of the shallow part of a volcanic system in unrest. If deployed and rapidly processed in (near) real-time during periods of unrest, Neural Network Nodal Ambient Noise Tomography can lead to dynamic and adaptive evacuation plans. Such advances would contribute to more effective, source-dependent risk mitigation schemes in volcanic regions, potentially saving lives.