CryoEM is an imaging technique that allows the determination of structures at the atomic level and cryoET refers to the tomographic version of this method. These highly noisy images have been extensively studied and the detection of single particles (i.e., isotropic molecules in a homogeneous sample) has been automated. However, for more complex structures such as fibrils, or for samples where more than one particle is of interest, the annotations must be manual. In this project, we aim to apply deep learning methods to improve the annotation pipeline by searching for geometric similarities along with image features. Furthermore, these methods are generalisable to other scientific imagery fields: a tool (scivision) is being developed to facilitate and bridge between them.
Organized by: Paulo Brás, Paulo Silva, Jaime Silva