Differentiable SV-FBP Algorithm: 10x Faster CT Scan Reconstruction
maier_ak · x · 2026-08-12
This article explores recent advancements in CT scan reconstruction, specifically addressing challenges in non-ideal scan paths (e.g., metal implants or deviations from perfect circular trajectories).
The study introduces a physics-informed AI algorithm called differentiable SV-FBP. In sparse-view tests, it matches the iterative AIR baseline in MSE, PSNR, and SSIM using 300-400 projections while running about ten times faster. However, with fewer than 200 projections, its quality degrades faster than AIR, requiring data-consistency constraints.
Related event: Differentiable SV-FBP Algorithm Accelerates Cone-Beam CT Reconstruction 10x(2 posts)→
More from Research
- OpenCap Monocular: Open-Source Tool for 3D Human Kinematics from a Single Smartphone Video — rsasaki0109 · 2026-08-12
- Novel Power Law Graph Attention Mechanism Replaces Scaled Dot-Product Attention — FromtheskyResearchLabs · 2026-08-12
- Terminal-Bench 3.0 Shakeup: Evaluating Combined Model and Agent Systems — teortaxesTex · 2026-08-12
- Bottleneck in Tabular Data Generation? Study Shows Probabilistic Circuits Beat Diffusion Models — tetraduzione · 2026-08-12
- Compressing 2.7B Pixels into a 790k-Parameter Neural Network — thisdudelikesAI · 2026-08-12
- New Paper Proposes Object Co-occurrence Framework to Interpret Semantic Representations via LLM Embeddings — TimKietzmann · 2026-08-12