Patel Jaydeep, Round Adam, Bielecki Johan, Doerner Katerina, Kirkwood Henry, Letrun Romain, Schulz Joachim, Sikorski Marcin, Vakili Mohammad, de Wijn Raphael, Peele Andrew, Mancuso Adrian P, Ab Bey Brian
School of Computing, Engineering and Mathematical Sciences, La Trobe University, Melbourne, Victoria, Australia.
La Trobe Institute for Molecular Science (LIMS), La Trobe University, Melbourne, Victoria, Australia.
J Appl Crystallogr. 2022 Aug 1;55(Pt 4):944-952. doi: 10.1107/S1600576722005891.
Liquid sample delivery systems are used extensively for serial femtosecond crystallography at X-ray free-electron lasers (XFELs). However, misalignment of the liquid jet and the XFEL beam leads to the X-rays either partially or completely missing the sample, resulting in sample wastage and a loss of experiment time. Implemented here is an algorithm to analyse optical images using machine vision to determine whether there is overlap of the X-ray beam and liquid jet. The long-term goal is to use the output from this algorithm to implement an automated feedback mechanism to maintain constant alignment of the X-ray beam and liquid jet. The key elements of this jet alignment algorithm are discussed and its performance is characterized by comparing the results with a manual analysis of the optical image data. The success rate of the algorithm for correctly identifying hits is quantified via a similarity metric, the Dice coefficient. In total four different nozzle designs were used in this study, yielding an overall Dice coefficient of 0.98.
液体样品输送系统在X射线自由电子激光（XFEL）的串行飞秒晶体学中被广泛使用。然而，液体射流与XFEL光束的未对准会导致X射线部分或完全错过样品，从而造成样品浪费和实验时间损失。本文实现了一种算法，利用机器视觉分析光学图像，以确定X射线束和液体射流是否重叠。长期目标是利用该算法的输出实现自动反馈机制，以保持X射线束和液体射流的恒定对准。讨论了该射流对准算法的关键要素，并通过将结果与光学图像数据的人工分析进行比较来表征其性能。通过相似性度量——骰子系数，对算法正确识别命中的成功率进行了量化。本研究总共使用了四种不同的喷嘴设计，总体骰子系数为0.98。
