Concurrent Workshop (Imaging): Light Microscopy and Flow Cytometry Research Groups
Accuracy and reproducibility of image segmentation by microscopist end-users using widely accessible programs
Segmentation—the partitioning of objects in an image from each other and from the background—is fundamental to image analysis and the starting-point for many automated measurements. As microscopy datasets increase in complexity and size, automated and semi-automated segmentation has likewise become essential in the analysis of these data. While increasingly sophisticated segmentation strategies continue to be developed, segmentation is still typically performed by microscopist end-users using pre-packaged routines available in common imaging and analysis programs. Numerous studies have been published in which single research groups compare the performance of different segmentation algorithms, which may or may not be easily accessible to the broader public. However, to our knowledge, there is a lack of studies that address the effects of user-to-user variability on consistency in image segmentation. Our goal is to compare and evaluate the efficacy and reproducibility of segmentation analyses performed by microscopist end-users using a range of widely accessible programs.
Because different segmentation programs involve varying levels of user input, we will test the consistency with which multiple, independent users analyze identical data sets with identical programs. We will quantify reproducibility and identify whether certain types of user interaction are more likely to increase variation. Because the efficacy and variability of segmentation may vary with image quality, we will utilize a range of datasets. To have complete control over image quality (e.g. signal-to-noise, object clustering), we have produced a library of simulated fluorescence microscope images (z-stacks) and their ground-truth references using the CytoPacq virtual microscope framework. We are currently developing protocols for each of the programs to be evaluated, and the infrastructure to deploy the study this year. The results of our study will inform the selection of programs best-suited for specific tasks and user abilities, and aid in the refinement of programs to minimize user-to-user variation.
FCRG - Fixation and Sorting for RNA
Increasingly, Flow Cytometry Shared Resource Facilities are asked to sort cells for RNA isolation either in bulk or at the single cell level. In many cases, the ability to fix the cell prior to sorting is desirable. With so many fixation methods in the literature the Flow Cytometry Research Group (FCRG) decided to perform a systematic evaluation of the reported fixation methods to assess how the different fixatives affected the quality of RNA isolated from sorted cells. Based on the literature, five different common chemical fixatives were analyzed using the cell line HL-60. The assessment included a paraformaldehyde fixation, alcohol fixation (methanol and ethanol), formaldehyde fixation, zinc fixation and two commercial fixations, BD Cytoperm/Cytofix (cat #554715) and eBiosciences Intracellular Fixation and Permeabilization Buffer (cat#88-8824-00). Each method was tested at two separate shared facilities and for each method different variations of the fixation procedure ie, time, temperature, dilution were also tested. The protocol involved fixing the cells first then proceeding to sort those cells into lysis buffer (RLT) and measure the amount, quality, and purity of the RNA. Four samples were used for each fixation condition: unfixed not sorted, unfixed sorted, fixed not sorted, and fixed sorted. Nanodrop was used for purity, Ribogreen for yield, and a bioanalyzer/qPCR for quality. Results will be presented that will aid researchers and shared facilities in determining optimal fixation processes for experimental design involving cell sorting.