Parameter-Efficient CT Reconstruction via Deep Graph Laplacian Regularization

ArXi:2605.25348v1 Announce Type: cross Low-dose computed tomography (LDCT) reconstruction faces a critical tradeoff between reconstruction quality and resource requirements. While recent deep learning methods achieve state-of-the-art performance, they typically rely on over 500,000 parameters trained on large-scale datasets exceeding 35,000 scans. This work investigates whether graph-based regularization can provide meaningful noise reduction under strict resource constraints.