Automatic/interactive bead alignment in electron tomographic image processing

Abstract

In the electron tomographic (EM) reconstruction process the mutual alignment between projections of different view angles is a crucial step. The routinely used alignment method is based on fiducial markers': a single-axis tilt projection series is collected with gold particles distributed on the specimen, the positions of high density gold beads on the projections are found, and the relationship between the specimen and the digital projection coordinate systems is determined from least-square fitting these found bead positions. There are four alignment parameters for each projection: two in shifts, one in in-plane rotation, and one in magnification. In the threedimensional studies of subcellular biological structures, we routinely collect data sets of more than 100 projections in the tilt range of ±75 ° with our automated EM set-up. Normally around 10 bead positions are used on each projection to achieve the alignment. Bead alignment used to be a laborious task since approximately 1000 bead positions need to be hand-picked for each data set.