NeRF-based CBCT Reconstruction needs Normalization and Initialization

2025-06-24

AI Summary

In neural radiance field (NeRF)-based cone-beam CT (CBCT) reconstruction, training instability, slow convergence, and severe artifacts arise from misalignment between local features encoded by hash tables and global features learned by the neural network. To address this, we propose a normalized hash encoder and a mapping-consistent initialization strategy: the former enforces feature normalization to improve local encoding consistency, while the latter leverages geometric priors from a pre-trained NeRF model to semantically align and initialize network weights. Integrated within the NeRF framework, our approach synergizes hash-based acceleration with transfer learning, requiring only minimal code modifications. Evaluated on 128 CT volumes across four datasets—spanning seven anatomical regions—our method significantly enhances training stability and convergence speed. Quantitatively, reconstruction quality improves by an average of 2.1 dB in PSNR and 0.032 in SSIM, while artifact suppression is markedly strengthened.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Feature Construction/Reformulation
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Cone Beam Computed Tomography (CBCT) is widely used in medical imaging. However, the limited number and intensity of X-ray projections make reconstruction an ill-posed problem with severe artifacts. NeRF-based methods have achieved great success in this task. However, they suffer from a local-global training mismatch between their two key components: the hash encoder and the neural network. Specifically, in each training step, only a subset of the hash encoder's parameters is used (local sparse), whereas all parameters in the neural network participate (global dense). Consequently, hash features generated in each step are highly misaligned, as they come from different subsets of the hash encoder. These misalignments from different training steps are then fed into the neural network, causing repeated inconsistent global updates in training, which leads to unstable training, slower convergence, and degraded reconstruction quality. Aiming to alleviate the impact of this local-global optimization mismatch, we introduce a Normalized Hash Encoder, which enhances feature consistency and mitigates the mismatch. Additionally, we propose a Mapping Consistency Initialization(MCI) strategy that initializes the neural network before training by leveraging the global mapping property from a well-trained model. The initialized neural network exhibits improved stability during early training, enabling faster convergence and enhanced reconstruction performance. Our method is simple yet effective, requiring only a few lines of code while substantially improving training efficiency on 128 CT cases collected from 4 different datasets, covering 7 distinct anatomical regions.

Problem

• Addresses local-global training mismatch in NeRF-based CBCT reconstruction
• Improves feature consistency with Normalized Hash Encoder
• Enhances stability via Mapping Consistency Initialization strategy

Innovation

• Normalized Hash Encoder enhances feature consistency
• Mapping Consistency Initialization improves training stability
• Simple yet effective method boosts reconstruction performance