Abstract

Fred J. Sigworth, Hemant Tagare, Andrew Barthel Yale University Departments of Cellular and Molecular Physiology, Diagnostic Imaging and Biomedical Engineering

Carazo, Scheres and co-workers have demonstrated very substantial advantages to using likelihood maximization as an approach to 3D
reconstruction from noisy cryo-EM images of heterogeneous samples.
The Expectation-Maximization algorithm is however computationally costly, requiring an exhaustive integration over many hidden variables. We present a dual-resolution integration procedure that allows high-precision integration with reductions by more than an order of magnitude in the number of function evaluations.