Cryogenic electron tomography (cryo-ET) is currently the golden standard for imaging cellular tissue at nanometer resolution. The current workflow from cultured cells to final image is however very time-consuming, labor-intensive, expensive, and has a low yield. Moreover, many of the steps in this workflow require practice. Even after practice, experienced users often lose samples during the cumbersome cryo-ET workflow.
Clipping of the Autogrid is one of the manual steps in the current procedure that is known to have a low yield. During this procedure, the fragile 20 µm thick sample carriers (TEM-grids) can get damaged or contaminated causing the loss of valuable samples. After clipping, the Autogrid increases the stiffness of the sample carrier, enabling automatic handling of the samples. Clipping of the Autogrid can therefore be considered as one of the missing links for full automation of the cryo-ET workflow. This thesis was aimed at automating the clipping procedure, thereby improving the yield, reducing the time required for practice, and reducing the amount of manual handling steps of the current procedure.
A problem analysis on the current procedure for clipping the Autogrid was performed, based on which multiple solutions were designed and tested experimentally. Designed gripper fingers were used with a six-axis industrial robot arm to automatically handle sample carriers and Autogrids. Such procedures are done manually using tweezers in the current procedure. Damage induced on the sample carriers during automatic handling was quantified experimentally.
Automating the clipping procedure allows for adjusting the orientation of sample carriers during the procedure, which is currently (almost) impossible. For this purpose, a machine learning-based marker detection algorithm was used to automatically detect markers that are present at the bottom of an Autogrid. This detection algorithm was used with a stepper motor and a designed mechanism to automatically obtain a specified orientation of the Autogrid. Finally, recommendations were given on how the proposed designs could be used in a final automated solution.