Search papers, labs, and topics across Lattice.
The paper introduces DSA-SRGS, a super-resolution Gaussian Splatting framework for reconstructing high-fidelity 4D vascular models from sparse-view dynamic DSA data. It incorporates a Multi-Fidelity Texture Learning Module that leverages a DSA-specific super-resolution model to provide high-quality priors, along with a Confidence-Aware Strategy to balance supervision signals. The framework also employs Radiative Sub-Pixel Densification to refine the 4D Gaussian kernels, resulting in improved reconstruction quality.
Achieve significantly sharper and more detailed 4D vascular reconstructions from sparse DSA data by injecting super-resolution priors into Gaussian Splatting.
Digital subtraction angiography (DSA) is a key imaging technique for the auxiliary diagnosis and treatment of cerebrovascular diseases. Recent advancements in gaussian splatting and dynamic neural representations have enabled robust 3D vessel reconstruction from sparse dynamic inputs. However, these methods are fundamentally constrained by the resolution of input projections, where performing naive upsampling to enhance rendering resolution inevitably results in severe blurring and aliasing artifacts. Such lack of super-resolution capability prevents the reconstructed 4D models from recovering fine-grained vascular details and intricate branching structures, which restricts their application in precision diagnosis and treatment. To solve this problem, this paper proposes DSA-SRGS, the first super-resolution gaussian splatting framework for dynamic sparse-view DSA reconstruction. Specifically, we introduce a Multi-Fidelity Texture Learning Module that integrates high-quality priors from a fine-tuned DSA-specific super-resolution model, into the 4D reconstruction optimization. To mitigate potential hallucination artifacts from pseudo-labels, this module employs a Confidence-Aware Strategy to adaptively weight supervision signals between the original low-resolution projections and the generated high-resolution pseudo-labels. Furthermore, we develop Radiative Sub-Pixel Densification, an adaptive strategy that leverages gradient accumulation from high-resolution sub-pixel sampling to refine the 4D radiative gaussian kernels. Extensive experiments on two clinical DSA datasets demonstrate that DSA-SRGS significantly outperforms state-of-the-art methods in both quantitative metrics and qualitative visual fidelity.
