Detecting microtubules in high noise Cryo-EM micrograph

Microtubules (MT), a fundamental component in cell structure, have recently been intensively studied in structural biology. Cryo-Electron Microscopy (Cryo-EM) is an effective method to gain insight into MTs' structure, while detection and segmentation of MTs are still mainly done by human analysis, because of the extremely high noise and massive irrelevant cell structures in Cryo-EM micrographs. Here we present an automated method to extract microtubules from micrographs. Our method is able to: 1. eliminating large contaminants according to their morphology features 2. filter the picture using our newly designed nonlinear filters; 3. finally select microtubules through a threshold of length. Having been tested with a large number of Cryo-EM micrographs, our method is proved to be effective and capable to replace human segmentation work.