CryoTriplane: Triplane-based Neural Radiance Field Representation for High-resolution Cryo-EM 3D Homogeneous Reconstruction.

CryoTriplane achieves 4.6% and 5.5% sequence coverage improvements over SOTA methods on two cryo-EM datasets, enabling higher-resolution 3D reconstructions.

• Why it matters: Accurately reconstructing high-fidelity 3D structures from noisy cryo-EM projections is crucial for understanding biomolecular functions, but existing methods struggle with noise and high-frequency details.
• What they did: The authors developed cryoTriplane, a real-space 3D reconstruction network combining fixed and structure-aware positional encoding, evaluated on five simulated and four real datasets.
• The result: CryoTriplane significantly outperforms existing approaches in resolving fine structural features and recovering high-quality density maps, supporting improved atomic modeling.