DeepSparse: A Foundation Model for Sparse-View CBCT Reconstruction

2025-05-05

AI Summary

To address critical challenges in sparse-view cone-beam CT (CBCT) reconstruction—including poor image quality, limited generalizability across scanners/protocols, and high computational cost—this work introduces the first foundation model tailored for low-dose CBCT sparse-view reconstruction. Methodologically: (1) we propose a Dual-dimensional Cross-scale Embedding (DiCE) network that jointly integrates multi-view 2D features and multi-scale 3D geometric priors; (2) we design a Hybrid View-sampling Pretraining (HyViP) framework with a two-stage fine-tuning strategy, enabling self-supervised pretraining and task-adaptive transfer. Evaluated on multiple public CBCT datasets, our method consistently outperforms state-of-the-art approaches, achieving ≥2.1 dB PSNR gain and +0.035 SSIM improvement. It delivers high-fidelity reconstructions while exhibiting strong cross-device and cross-protocol generalization, all with reduced inference computational overhead.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Cone-beam computed tomography (CBCT) is a critical 3D imaging technology in the medical field, while the high radiation exposure required for high-quality imaging raises significant concerns, particularly for vulnerable populations. Sparse-view reconstruction reduces radiation by using fewer X-ray projections while maintaining image quality, yet existing methods face challenges such as high computational demands and poor generalizability to different datasets. To overcome these limitations, we propose DeepSparse, the first foundation model for sparse-view CBCT reconstruction, featuring DiCE (Dual-Dimensional Cross-Scale Embedding), a novel network that integrates multi-view 2D features and multi-scale 3D features. Additionally, we introduce the HyViP (Hybrid View Sampling Pretraining) framework, which pretrains the model on large datasets with both sparse-view and dense-view projections, and a two-step finetuning strategy to adapt and refine the model for new datasets. Extensive experiments and ablation studies demonstrate that our proposed DeepSparse achieves superior reconstruction quality compared to state-of-the-art methods, paving the way for safer and more efficient CBCT imaging.

Problem

• Reducing radiation exposure in CBCT with sparse-view reconstruction
• Overcoming high computational demands in sparse-view CBCT methods
• Improving generalizability of sparse-view CBCT across diverse datasets

Innovation

• DiCE integrates 2D and 3D cross-scale features
• HyViP pretrains with sparse and dense projections
• Two-step finetuning adapts model to new datasets