Improved cryo-EM Pose Estimation and 3D Classification through Latent-Space Disentanglement

arXiv:2308.04956v2 Announce Type: replace-cross
Abstract: Due to the extremely low signal-to-noise ratio (SNR) and unknown poses (projection angles and image shifts) in cryo-electron microscopy (cryo-EM) experiments, reconstructing 3D volumes from 2D images is very challenging. In addition to these challenges, heterogeneous cryo-EM reconstruction requires conformational classification. In popular cryo-EM reconstruction algorithms, poses and conformation classification labels must be predicted for every input cryo-EM image, which can be computationally costly for large datasets. An emerging class of methods adopted the amortized …

abstract arxiv challenges classification cryo-em cs.cv eess.iv electron image images latent-space low microscopy noise popular projection q-bio.bm q-bio.qm signal space through type