R²-Gaussian is a novel framework for sparse-view tomographic reconstruction, leveraging 3D Gaussian splatting (3DGS) to achieve high-quality and efficient results. The paper identifies and addresses an integration bias in standard 3DGS, which hinders accurate volume retrieval. To rectify this issue, R²-Gaussian introduces tailored Gaussian kernels, extends rasterization to X-ray imaging, and develops a CUDA-based differentiable voxelizer. Experiments on synthetic and real-world datasets demonstrate that R²-Gaussian outperforms state-of-the-art methods in both reconstruction quality and training speed, achieving results in 4 minutes, 12 times faster than NeRF-based methods and comparable to traditional algorithms. The method's effectiveness is validated through visual comparisons, quantitative metrics, and ablation studies, highlighting its potential for real-world applications in medical diagnosis and industrial inspection.R²-Gaussian is a novel framework for sparse-view tomographic reconstruction, leveraging 3D Gaussian splatting (3DGS) to achieve high-quality and efficient results. The paper identifies and addresses an integration bias in standard 3DGS, which hinders accurate volume retrieval. To rectify this issue, R²-Gaussian introduces tailored Gaussian kernels, extends rasterization to X-ray imaging, and develops a CUDA-based differentiable voxelizer. Experiments on synthetic and real-world datasets demonstrate that R²-Gaussian outperforms state-of-the-art methods in both reconstruction quality and training speed, achieving results in 4 minutes, 12 times faster than NeRF-based methods and comparable to traditional algorithms. The method's effectiveness is validated through visual comparisons, quantitative metrics, and ablation studies, highlighting its potential for real-world applications in medical diagnosis and industrial inspection.