Trans${^2}$-CBCT: A Dual-Transformer Framework for Sparse-View CBCT Reconstruction

2025-06-20

AI Summary

Sparse-view cone-beam CT (CBCT) suffers from severe undersampling, leading to prominent artifacts and incomplete spatial coverage, thereby limiting its utility in low-dose and rapid imaging. To address this, we propose Trans²-CBCT—a dual-Transformer collaborative reconstruction framework. Its backbone integrates multi-scale CNN and Transformer features (Trans-CBCT), while a neighborhood-aware point cloud Transformer enforces voxel-level consistency. Key innovations include 3D positional encoding, k-nearest-neighbor attention, a lightweight decaying prediction head, and a multi-view feature query mechanism. Evaluated on the six-view LUNA16 dataset, Trans-CBCT achieves a PSNR of 34.21 dB (+1.17 dB) and SSIM of 0.9231 (+0.0163) over baseline methods; further refinement yields additional gains of +0.63 dB in PSNR and +0.0117 in SSIM, significantly enhancing reconstruction quality and geometric fidelity.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Cone-beam computed tomography (CBCT) using only a few X-ray projection views enables faster scans with lower radiation dose, but the resulting severe under-sampling causes strong artifacts and poor spatial coverage. We address these challenges in a unified framework. First, we replace conventional UNet/ResNet encoders with TransUNet, a hybrid CNN-Transformer model. Convolutional layers capture local details, while self-attention layers enhance global context. We adapt TransUNet to CBCT by combining multi-scale features, querying view-specific features per 3D point, and adding a lightweight attenuation-prediction head. This yields Trans-CBCT, which surpasses prior baselines by 1.17 dB PSNR and 0.0163 SSIM on the LUNA16 dataset with six views. Second, we introduce a neighbor-aware Point Transformer to enforce volumetric coherence. This module uses 3D positional encoding and attention over k-nearest neighbors to improve spatial consistency. The resulting model, Trans$^2$-CBCT, provides an additional gain of 0.63 dB PSNR and 0.0117 SSIM. Experiments on LUNA16 and ToothFairy show consistent gains from six to ten views, validating the effectiveness of combining CNN-Transformer features with point-based geometry reasoning for sparse-view CBCT reconstruction.

Problem

• Reconstructing sparse-view CBCT with severe artifacts
• Enhancing global context and local details in CBCT
• Improving spatial consistency in volumetric CBCT reconstruction

Innovation

• Hybrid CNN-Transformer model for CBCT reconstruction
• Neighbor-aware Point Transformer for volumetric coherence
• Multi-scale features with view-specific queries