Diffusion Low Rank Hybrid Reconstruction for Sparse View Medical Imaging

2025-10-05

AI Summary

To address the severe ill-posedness and substantial texture degradation in low-dose, extremely sparse-view CT reconstruction (e.g., 2–8 views), this paper proposes TV-LoRA: a novel reconstruction framework integrating diffusion model priors—built upon NCSN++ architecture and stochastic differential equation (SDE) modeling—with multiple regularization constraints, including low-rank approximation (LoRA), anisotropic total variation (TV), and nuclear norm regularization. The optimization is efficiently solved within the alternating direction method of multipliers (ADMM) framework. To accelerate computation, fast Fourier transform (FFT)-based acceleration and tensor parallelism are incorporated. Experiments on multiple public datasets demonstrate that TV-LoRA significantly outperforms state-of-the-art methods: it achieves higher structural similarity (SSIM), markedly reduces artifacts, preserves edge sharpness and fine-texture fidelity, and exhibits strong robustness and promising clinical applicability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Matrix & Tensor Methods
• Search and Optimization: Non-convex Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

This work presents TV-LoRA, a novel method for low-dose sparse-view CT reconstruction that combines a diffusion generative prior (NCSN++ with SDE modeling) and multi-regularization constraints, including anisotropic TV and nuclear norm (LoRA), within an ADMM framework. To address ill-posedness and texture loss under extremely sparse views, TV-LoRA integrates generative and physical constraints, and utilizes a 2D slice-based strategy with FFT acceleration and tensor-parallel optimization for efficient inference. Experiments on AAPM-2016, CTHD, and LIDC datasets with $N_{mathrm{view}}=8,4,2$ show that TV-LoRA consistently surpasses benchmarks in SSIM, texture recovery, edge clarity, and artifact suppression, demonstrating strong robustness and generalizability. Ablation studies confirm the complementary effects of LoRA regularization and diffusion priors, while the FFT-PCG module provides a speedup. Overall, Diffusion + TV-LoRA achieves high-fidelity, efficient 3D CT reconstruction and broad clinical applicability in low-dose, sparse-sampling scenarios.

Problem

• Reconstructs CT images from sparse-view low-dose data
• Addresses ill-posedness and texture loss in medical imaging
• Combines diffusion priors with multi-regularization constraints

Innovation

• Combines diffusion generative prior with multi-regularization constraints
• Integrates 2D slice strategy with FFT acceleration optimization
• Uses ADMM framework for efficient 3D CT reconstruction