EMBO Practical Course Retrospective
In September, Franklin colleagues, Sam Kersley, Dolapo Adebo, Piper Fowler-Wright, Maud Dumoux and Michele Darrow, joined international colleagues to take part in an EMBO practical course on Advanced analysis and processing of volume electron microscopy data. Sam, Dolapo and Piper have written about their experiences here.
What was the purpose of the EMBO course?
Sam: The course gave participants the opportunity to learn how to process microscopy images, from understanding the basic concepts to getting hands-on experience with different software. They were supported by people who use these tools and workflows in their own work. The course covered a wide range of topics for both beginners and more experienced users, while also giving participants opportunities to meet others and build future collaborations.
Dolapo: The course was designed to give researchers and scientist hands-on training in processing and analysing volume electron microscopy (vEM) data, with the aim of making vEM analysis more efficient, reproducible and accessible, so that participants with prior imaging or data analysis experience can confidently handle, analyse and interpret complex datasets.
Piper: As a research software engineer who has supported other EMBO events, I see these courses as bringing experts together to deliver training in advanced techniques and the latest domain-relevant software, to participants at various career stages across the community.
What was your role and what did it entail?
Sam: My role was to be both an in-person teacher and to help create material for tutorial. I helped create the hands-on tasks and tutorials with Dolapo and deliver the in-person tutorial/hands-on sessions regarding Napari and the Volume Segmentics (Vol-Seg) toolkit. Helping answer questions about the toolkits use and potential regarding future projects was not only helpful for the participants but it also allowed us to rather feedback ahead of the toolkits release early in the new year.
Dolapo: I was a teacher and instructor on the course, leading the quantification component for the online participants. I developed reusable Jupyter notebooks showing how to quantify segmented volumes, including skeletonization, pore and throat analysis, surface area and volume measurements. I also supported online participant during the hands-on sessions, guiding them, through the exercises and troubleshooting any issues they encountered.
Piper: My role was to help provision compute workspaces for the online participants. Volume EM tools and workflows can be computationally demanding, often beyond what a personal computer can handle. Asking participants to install the software themselves would also have been a barrier, since the complexity of installation and the availability of documentation can vary considerably between tools and operating systems.
The setup proceeded in two stages. First, in collaboration with Yi Sun at EMBL, I packaged the vEM software and image analysis tools into portable software bundles called container images. Second, I coordinated deploying those images onto the workspaces run by the Science and Technology Facilities Council’s Ada Training service.
What was the biggest success for you?
Sam: The smoothness of the courses flow; the hands-on sessions worked without major hitches where everyone had access to materials and was able to experience what was made specifically for each part. If participants had questions or issues, we were there to help, answer them promptly and also give additional information relative to their needs; which is essential to learning.
Dolapo: Creating iterable Jupyter notebooks and Python scripts that turned quantification from an abstract topic into something participants could run themselves. They covered skeletonization, pore and throat analysis, surface area and volume measurements, and gave participants a reusable toolkit they could adapt to their own datasets after the course. Another highlight was helping remote participants get through the hands-on sessions, troubleshooting their issues in real time so they could follow the exercises as fully as those in the room.
Piper: Every online participant connected to a usable, preconfigured workspace with the main datasets pre-loaded, and there were no major connection problems despite the distance of some European participants from the workstation servers in Oxfordshire. Personally, however, the standout was the coordination with Yi Sun at EMBL, who ran the BAND cluster for in-person participants. We worked hard to ensure consistency between the two platforms, so that online and in-person participants had the same experience. Despite the platforms having very different configuration management and initial software bases, I think this was achieved!
What was the biggest takeaway from your experience?
Sam: It was great to able to connect participants and offer advice for their own work that hopefully gave them the confidence or new pathway that they need to complete their work. The feedback was positive and indicated that everyone learned what they needed to from the experience.
Dolapo: The bottleneck in vEM has shifted from acquiring data to analysing it, and that AI-assisted segmentation, deep learning and high-performance computing are closing that gap. It taught me that reproducible, well-documented workflows matter as much as the imaging methods, and that teaching is one of the best ways to deepen your own understanding.
Piper: The importance of real-world testing. Research software is often deep and takes expertise together with large datasets to operate fully. You can deploy it and run automated tests, but until you have experts run real workflows against it, you can’t be sure it is fit for purpose.