Title: Machine Learning for Cryo-Electron Microscopy

Session Chairs: Tommi White (University of Missouri) and Alberto Bartesaghi (Duke University)



Cryo-electron microscopy, cryo-EM, Single particle analysis, SPA, single particle reconstruction, SPR, cryo-electron tomography, cryoET, particle detection, particle picking, image regularization, semantic tomogram segmentation, continuous heterogeneity, transformation invariant classification, latent space representations.



Using higher resolution imaging modalities, such as cryo-electron microscopy, scientists are able to discern subcellular structures at the molecular level leading to discoveries in basic and translational sciences as well as applications in drug discovery and precision medicine. These methods include cryo-electron microscopy for single particle analysis and cryo-electron tomography for sub-volume classification and averaging.  These imaging methods, as well as many others, rely on computational techniques to align and reconstruct high resolution images in 2D and 3D for visualization and further quantitative analysis. Advances in machine learning, particularly in deep learning, have been shown to facilitate low signal, noisy image analysis, particularly in the particle detection, denoising and continuous heterogeneity steps of reconstruction.

This special session brings together experts from computational and imaging fields to focus on cutting-edge algorithmic approaches for reconstruction and analysis of data obtained from high-resolution imaging modalities. Special emphasis will be on deep learning and other machine learning methods applied to cryo-electron microscopy images.  Original contributions in applications of deep learning and other machine learning methods in high-resolution microscopy including but not limited to denoising, detection, segmentation, transformation invariant classification and reconstruction, as well as new approaches in continuous heterogeneity analysis are welcome.