Martin is the Deputy Head of Microscopy Prototyping in the Electron Microscopy Science Technology Platform at the Francis Crick Institute. His work focuses on developing new software and hardware tools for dealing with the deluge of data coming from modern microscopes. His background is in physics, with a DPhil from the University of Sussex in experimental quantum optics. After postdoctoral research and teaching fellowships in physics, he moved to the Vascular Biology Lab at Cancer Research UK's London Research Institute to work on microscope development and image analysis. From there he moved to the LRI's Electron Microscopy core facility, which subsequently moved to its current home at the Francis Crick Institute. Martin works closely with the Crick's Software Engineering and AI team to develop methods for analysing large complex imaging datasets. He is also the current chair of the Royal Microscopical Society’s Data Analysis in Imaging section.

Abstract: Extracting meaning from big data in biological electron microscopy

Volume electron microscopy techniques, such as serial block face SEM (SBF SEM), focused ion beam SEM (FIB SEM) and array tomography (AT), can produce datasets comfortably into the terabyte regime. In many cases, much of the analysis is still performed manually, which represents a major bottleneck in the workflow, restricting the amount of information that can be extracted from these rich datasets. Modern computational techniques, such as deep learning based segmentation, have shown a great deal of promise for analysing this kind of complex data. Another powerful technique for efficiently imaging and interpreting biological data is the use of correlative multimodal methods, such as correlative light and electron microscopy (CLEM). Once again, some key analysis steps in such pipelines are still frequently performed manually, such as the co-registration of volumetric images acquired in different modalities.

I will present two of our analysis workflows. First, I will show a segmentation pipeline based upon training a deep learning model on crowdsourced training data in our Etch a Cell project [1]. Second, I will present our recent work on leveraging existing segmentation methods to automatically align volumetric EM and fluorescence datasets via the use of point clouds [2].

[1] H. Spiers et al, Traffic (2021), https://doi.org/10.1111/tra.12789

[2] D. Krentzel et al, bioRXiv (2023), https://doi.org/10.1101/2023.05.11.540445