Magnetic Particle Imaging (MPI) is an emerging tracer-based medical imaging technique measuring distributions of superparamagnetic nanoparticles using field-free point trajectories, often following Lissajous curves. The resulting tracer concentration frequently exhibits sharp edges between regions, making them ideal candidates for VSDK reconstruction. We present a shape-driven algorithm combining edge detection via Support Vector Machine classification with VSDK interpolation. The approach is fully automatic, requiring only threshold selection for binary classification, and significantly outperforms standard polynomial schemes in both visual quality and quantitative accuracy metrics.