VT recurrence after ablation remains high because current mapping relies on surface activation alone and lacks a principled way to identify minimal targets to terminate reentry. In our earlier I-O-U model, rotational centers in simple simulations adopt I-, O-, or U-shapes, with shortest ablation lines connecting boundaries of opposite rotation. This approach fails for realistic scars, where infinitely many surface layers must be considered. We address this using a torus-based framework grounded in algebraic topology. In this scheme, VT is modeled as rotation around 3D obstacles (scar tissue, the surrounding medium, or wave interactions), with boundaries defined as surfaces of healthy tissue between healthy tissue and obstacles, thereby including the epi- and endocardium. All rotational activity is captured on these boundaries, eliminating the need to analyze for additional layers. In this work, we show that critical boundaries must have a torus topology. As a result, realistic scars form collections of boundaries, each with g holes. This property is shown in three simulations of increasing complexity. In each simulation, minimal ablation surfaces are identified via the shortest homology basis, yielding g tunnel loops per boundary. Critical tunnels hosting reentry define minimal ablation targets. We hypothesize that MRI-derived geometry in clinical VT can generate candidate loops, with activation maps of the epi- or endocardium identifying those sustaining VT.