Design Choices in Splitting-Based Self-Supervised Sparse-View CT Reconstruction

2026-07-12

AI Summary

This work addresses the challenge of self-supervised learning for sparse-view CT reconstruction in the absence of fully sampled ground truth by introducing a unified framework that decouples existing methods into partitioning strategy, preprocessing, and inference components. The study systematically investigates how the structure of measurement noise affects reconstruction performance and proposes a multi-partition splitting scheme alongside a novel inference strategy. It reveals a critical dependence between partitioning design and noise correlation, thereby challenging the conventional assumption of noise independence. Experiments on the LoDoPaB-CT and 2DeteCT datasets demonstrate that grid-based partitioning is well-suited for independent noise, whereas angular masking exhibits greater robustness under correlated noise and real-world conditions. The multi-partition approach consistently outperforms single-projection partitioning, with its superiority confirmed through perceptual and structural metrics including LPIPS and HaarPSI.

Abstract

Self-supervised data splitting has emerged as a promising paradigm for sparse-view CT reconstruction, enabling training from incomplete measurements without fully sampled ground truth. However, the influence of key design choices, including partitioning strategy, preprocessing, and inference, remains insufficiently understood. In this work, we introduce a unified framework that decomposes splitting-based reconstruction into these three components, enabling controlled comparison of existing methods and two incremental extensions: multi-partition splitting and an alternative inference strategy. Experiments on simulated LoDoPaB-CT data under independent and correlated noise, together with validation on the real-world 2DeteCT dataset, show that the optimal partitioning strategy strongly depends on the measurement noise structure. Lattice-based splitting performs favorably under independent noise, whereas angular masking is more robust under correlated noise and real measured data. Multi-partition splitting consistently improves over pure projection-wise splitting in several settings. Complementary perceptual and structural metrics, including LPIPS and HaarPSI, reveal differences between masking strategies that are less apparent from PSNR and SSIM alone. These results provide practical guidelines for designing self-supervised sparse-view CT reconstruction methods and highlight the limitations of common independence assumptions in realistic imaging environments.

Problem

• sparse-view CT reconstruction
• self-supervised learning
• data splitting
• noise structure
• partitioning strategy

Innovation

• splitting-based reconstruction
• self-supervised learning
• sparse-view CT
• multi-partition splitting
• noise-robust inference