Projection-Volume Fidelity Divergence: Diagnosing and Controlling Optimization Drift in Sparse-View 3D Gaussian Tomography

2026-06-21

AI Summary

This study addresses the "optimization drift" problem in sparse-view CT reconstruction, where persistent improvements in projection-domain optimization paradoxically degrade reconstructed volume quality. The work formally defines and diagnoses this issue as a projection-volume fidelity divergence (PVFD). To mitigate PVFD without requiring ground-truth volumes, the authors propose LADES, an optimization controller integrating linear-annealed Dropout with a structure-aware early-stopping strategy. The approach further incorporates an explicit 3D Gaussian splatting representation, geometric and voxel-level diagnostic metrics, and a mechanism to detect saturation in Gaussian density growth. Experimental results demonstrate that LADES significantly enhances volume fidelity, effectively suppresses structural degradation, substantially reduces training time, and maintains high projection accuracy.

Abstract

Sparse-view computed tomography is a severely ill-posed inverse problem, where recent 3D Gaussian Splatting methods offer an efficient explicit representation for tomographic reconstruction. However, we find that projection-domain optimization can be misleading in this setting: the rendered projections may continue to improve while the reconstructed volume deteriorates. We identify this failure mode as Projection-Volume Fidelity Divergence (PVFD), a representation-level optimization drift caused by anisotropic Gaussian deformation and view-specific primitive co-adaptation under sparse Radon constraints. To characterize this behavior, we introduce geometry- and volume-level diagnostics that measure needle-like Gaussian degeneration and the stability of the voxelized density field. Based on these observations, we propose LADES, a ground-truth-free optimization controller for sparse-view Gaussian tomography. LADES combines Linearly Annealed Dropout, which applies strong stochastic masking in early training to disrupt premature primitive co-adaptation and gradually restores full capacity for structural consolidation, with Structure-Aware Early Stopping, which terminates densification according to the saturation of Gaussian population growth rather than validation PSNR. Experiments on sparse-view CT reconstruction show that LADES improves volumetric fidelity, suppresses structural degeneration, and substantially reduces training time while maintaining competitive projection accuracy. These results suggest that robust Gaussian-based tomography requires monitoring and controlling volumetric structure, rather than optimizing projection fit alone.

Problem

• sparse-view tomography
• Projection-Volume Fidelity Divergence
• optimization drift
• 3D Gaussian Splatting
• volumetric fidelity

Innovation

• Projection-Volume Fidelity Divergence
• Gaussian Splatting
• Sparse-View Tomography
• Optimization Drift
• Structure-Aware Early Stopping