FastTomo: A Hybrid Approach to Speedup Tomography Data Collection¶

Co-Author: Albert Xu <albert.t.xu@gmail.com>
Co-Author: Chen Xu <chen.xu@umassmed.edu>
Date-Created: May 24, 2020
Last-Updated: Feb 21, 2022

Abstract

Due to the imperfection of stage movement, a tomography target positon (X, Y, Z) usually keeps changing with tilting angle. In order to obtain useful data, some kind of engineering control has to be in place to ensure the target postions are within the acceptable range through the entire tilting series. This engineering control might involve complete prediction or/and frequent tracking for X, Y and Z postions. Unfortunitely, these extra control actions cost time for each tilting series collection. Here, we propose a hybrid method to make the data collection as fast as possible without sacrificing too much data quality.

We use a precalibration for each and every tilting series for Z, and we use proportional control to constantly compensate X, Y potions. We try to eliminate tracking and auto-focusing shots as much as possible.

We present a SerialEM script using this Hybrid method for all three tilting schemes: uni-directional, bi-directional and dose-symmetirc. We slightly modify the original Hagen scheme of dose-symmetric with added options to 1) switch to bi-directional beyond certain defined angle and 2) change exposure time at high angle ranges after switching to bi-directional.

Some Brief background Information¶

Our tests on stage tilting behavious mainly show two things. 1) Statistically, it is hard to get specimen height to desirable range of eucentricity. There seems to be always some error to get to eucentric height, large or small. 2) the relationshiop between tilt angle and X,Y,Z position is not quite linear when the specimen is not close enough to eucentric height; even at very close to eucentric position it is no longer linear at high tilt range such as above +/- 45 degree.

Most of time consuming actions are the multiple tracking shots and auto-focusing precedures.

This makes us to think if we could have a way to eliminate most of, if not completely, the Focus and Tracking Shots, we might save time for tilting series data collection.

FastTomo Script¶

For Z, we found that a SINE function can describle the Z change fairly well. We only need a few tilt points to pre-calibrate such a SINE curve for each target postion. Such pre-calibration does cause some time, but the prediction based on the calibrated SINE curve seems to be very robust and we then completely eliminate Focus shots in the collection step. The Objective lens is used to compensate.

For X,Y, we do not perform any precalibration. Instead, we just use the returned Record images to perform the preportional control. Namely, we align the feature from Record image at each tilt to its previous one and we use Image Shift to compensate. We completely eliminate Tracking T shots in the procedure except in dose-symmetric scheme which tracking shot is used only at switching angles.

The SerialEM script - FastTomo and its usage can be found from the github.com project page.

A Few Points to Discuss¶

0. This is no replacement for SerialEM controller to collect tilting series. It can save some time, by skipping most of Focus and Trial shots, if the condition is proper. For example, the eucentricity is working very well, beam center shifted to tilting axis nicely so there is no much lateral displacement, the stage is working well mechanically and the pixelsize is not too small. When all these conditions are met, this script might work well and fast.

1. The same idea and method should be also applied to a FISE collection, provided the frames from each tilt can be grabbed and available for scripting.

2. A single, long exposure with shutter opening and closing during tilting series saves time for startup delay and returning time, compared to conventional way of multiple exposure one after another. But the significant time saving is from minimizing the numbers of F and T shots aleady! Using our script FastTomo, the total time for bi-directional tilting series from -51 to + 51 degree with 3 degee step is about 6 minutes including pre-calibration for Z prediction curve.

reference: “FastTomo: A SerialEM Script for Collecting Electron Tomography Data” Albert Xu, Chen Xu bioRxiv 2021.03.16.435675; doi: https://doi.org/10.1101/2021.03.16.435675