What Should I Do with My Serial Block-Face Data? Suggestions for Preparing, Analyzing, and Presenting Volume EM Datasets
Year: 2023
Type: article
Abstract: Volume electron microscopy (volume EM) is undergoing an exciting evolution. Advances in instrumentation and methods to automatically acquire large image datasets have brought volume EM to the forefront of cell and tissue biology, and it is rapidly becoming a tool of choice for meso-scale imaging at many research institutions. Investigators trying volume EM for the first time quickly realize that acquiring the images is only the first step in unleashing the potential of this imaging technique. The most common question we receive from a client after receiving their serial block-face data is…” What do I do with this now?” The sheer volume of volume EM data can be intimidating to those unfamiliar with it. Investigators often struggle to catch up and learn just how to work with these large volume EM datasets. What can we do with the data? How can it be presented? Where do we start? Presented here is the stepwise approach we take in our core facility to introduce investigators to their first volume EM datasets. Our goal is to help them see the answers to their research questions and to reveal the potential of new perspectives for discovery science. We generate volume EM data in our facility using a serial block-face SEM, but these steps could be applied to data from any volume EM modality, including focused ion beam (FIB-SEM), electron tomography, automated tape-collecting ultramicrotome (ATUM-SEM), array tomography (AT), and manual serial section TEM. There are numerous software packages available for image processing. Our approach is to first introduce new clients to volume EM using as much freeware as possible and incorporate more advanced packages and processing techniques as the project develops. We use programs such as Fiji, Reconstruct, and AmiraTM (Thermo Fisher) for most of our work [1, 2]. Step 1. Prepare the raw images for use. Before any image manipulation, we first archive a copy of all original data to a backup location. We next assess the entire image stack, removing images that have debris or did not capture the correct region of interest. Depending on the acquisition parameters, the final images may benefit from some quick initial processing, such as Gaussian blur and unsharp mask to denoise and sharpen while adjusting contrast and brightness to enhance the aesthetics of the final image. If the images need to be aligned, then this is the point in the workflow to do so. Step 2. Create and observe a flipbook movie. After the images are ready, we drop the stack into a viewer and create a flipbook movie that plays through the entire stack at a reasonably slow speed. The creation and subsequent viewing of this movie is perhaps the most critical step in processing volume EM data. It provides the investigator that first look at the full volume and a tool to envision the cellular and tissue architecture and interactions in 3 dimensions. It is here where those first questions begin to be answered. We highly recommend to our clients to spend considerable time taking in these movies and acclimate to viewing EM data in 3D, because it is often at this stage where initial discovery of unanticipated architectural relationships become realized. Step 3. Prepare a volume rendering. Generating a volume rendering can be quickly done using the rendering functions in Fiji or Amira, and immediately provides a 3-dimensional look of the data that can be rotated on any plane. For many renderings it is helpful to adjust the opacity or color to highlight specific features within the volume. The rendering also provides a visually agreeable image that can be immediately incorporated into a figure. Step 4. Segmentation. Usually at this time the investigator has viewed and selected key features and often wants to further highlight or analyze them by measuring size, volume, or spatial proximities. To address this, these key features need to be segmented from the full image. Whether manually segmented or segmented using processing algorithms, this step will take some time. For simple manual segmentations of relatively small areas, we start our clients with the program Reconstruct, as it is very easy to use freeware and does not require considerable processing power, allowing clients the convenience of working on their own computers anywhere. If more extensive segmentation and analysis is required, we have several high-end workstations available with licensed image processing packages such as Amira. Step 5. Present the data. Options for presenting volume EM data are limitless and driven by how the final product will be used. Volume renderings, segmentations and data analysis presented in 2-dimensional format; movies and 3D printed models are all examples we have used in our facility. In our experience, displaying as many of these examples to our new clients has been immensely successful in helping them to see what is possible and motivate them to explore the exciting and evolving world of volume EM. Examples of how volume EM data can be presented. Grayscale volume renderings of blood clot (a) and collagen fibers (b) highlight arrangement of fibrin and collagen fibers and distribution of cells within. A single 2D image of a budding yeast next to the colorized volume rendering from the same area highlights the numerous transport vesicles present (c). Manually segmented and rendered fibrin strands in proximity to red blood cells within a blood clot (d). Manually segmented portion of a kidney tubule highlighting the primary cilia of the proximal tubular epithelial cells (e). Manually segmented neurons from mouse brain rendered transparent to highlight the mitochondria within (f). Manually segmented axons and neurons displayed atop a 2D image from the area used for the rendering (g). AI generated segmentation of a large platelet plug formed at a wound site on a jugular vein (h). Examples of 3D printed neurons that allow the investigator to physically hold and study tissue architecture – great for teaching and presentations.
Source: Microscopy and Microanalysis
Institution: Mayo Clinic
Language: English
Subfield: Structural Biology
Domain: Life Sciences
Open Access status: bronze
OpenAlex ID: W4385072594