XTransCT: Ultra-Fast Volumetric CT Reconstruction using Two Orthogonal X-Ray Projections for Image-guided Radiation Therapy via a Transformer Network

Chulong Zhang,Lin Liu,Jingjing Dai,Xuan Liu,Wenfeng He,Yinping Chan,Yaoqin Xie,Feng Chi,Xiaokun Liang

2023-11-24

Abstract:Computed tomography (CT) scans offer a detailed, three-dimensional representation of patients' internal organs. However, conventional CT reconstruction techniques necessitate acquiring hundreds or thousands of x-ray projections through a complete rotational scan of the body, making navigation or positioning during surgery infeasible. In image-guided radiation therapy, a method that reconstructs ultra-sparse X-ray projections into CT images, we can exploit the substantially reduced radiation dose and minimize equipment burden for localization and navigation. In this study, we introduce a novel Transformer architecture, termed XTransCT, devised to facilitate real-time reconstruction of CT images from two-dimensional X-ray images. We assess our approach regarding image quality and structural reliability using a dataset of fifty patients, supplied by a hospital, as well as the larger public dataset LIDC-IDRI, which encompasses thousands of patients. Additionally, we validated our algorithm's generalizability on the LNDb dataset. Our findings indicate that our algorithm surpasses other methods in image quality, structural precision, and generalizability. Moreover, in comparison to previous 3D convolution-based approaches, we note a substantial speed increase of approximately 300 %, achieving 44 ms per 3D image reconstruction.

Image and Video Processing,Computer Vision and Pattern Recognition,Medical Physics

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

This paper aims to address the problem of real-time 3D CT image reconstruction from ultra-sparse 2D X-ray projections in image-guided radiation therapy (IGRT). The specific objectives include:

1. **Reducing Radiation Dose**: Lowering the radiation dose received by patients by reducing the number of required X-ray projections.
2. **Improving Real-time Performance**: Achieving faster reconstruction speeds to enable real-time navigation and positioning during surgery.
3. **Simplifying Hardware Requirements**: Achieving high-quality CT image reconstruction using only two orthogonal X-ray projections, thereby simplifying equipment requirements.

### Main Contributions

1. **Proposed a Novel Voxel Space Search Transformer Model**: This model can reconstruct CT images in real-time without the need for 3D convolution, significantly improving reconstruction speed.
2. **Superior Algorithm Performance**: Validated on multiple datasets, including data from 50 patients in hospitals, the large-scale public dataset LIDC-IDRI, and the LNDb dataset, demonstrating excellent performance in image quality, structural accuracy, and generalization ability.
3. **Significant Speed Improvement**: Compared to traditional 3D convolution methods, the reconstruction speed is improved by approximately 300%, with each 3D image reconstruction taking only 44 milliseconds.

### Conclusion

The XTransCT algorithm has potential applications in clinical practice, providing high-quality CT images while reducing radiation dose, suitable for real-time radiation therapy navigation in various medical environments.