Enhanced Image Segmentation Now Available for Your ResearchGood morning, everyone! The spatial transcriptomics platform has made significant advancements, now providing accurate single-cell-level profiling for revolutionary insights. To further elevate our capabilities at ASOC, we have established a new Xenium In Situ analysis pipeline that integrates cutting-edge image analysis powered by a third-party segmentation tool, i.e., StarDist (https://stardist.net/). Our new pipeline enables more accurate identification of individual cells by mapping nuclei with high fidelity. Importantly, we can now optimize the segmentation tissue-by-tissue, ensuring tailored accuracy and improved data quality for different sample types. This enhancement allows for more detailed spatial insights into gene expression and cellular organization, empowering you to uncover deeper biological understanding. Check out our new workflow and the enhancement in the attachment.
We believe that the 10X Genomics segmentation algorithm performs effectively with specific tissue types due to its optimization. However, our new pipeline is designed to significantly enhance your research outcomes, particularly when your cells of interest are rare or when presets are unavailable through 10X Genomics. We can find a set of optimal parameters tailored to your tissue type to ensure the most accurate detection of nuclei, helping you to effectively address your research questions. If you're interested in learning more about how the Xenium analysis pipeline can benefit your work, please feel free to reach out. FAQs: A. Does a third-party tool outperform the 10X Genomics segmentation? The short answer is no. At ASOC, we evaluated various tissue types to compare their results. The effectiveness of a tool depends on the region of interest—whether it's tumor or adjacent normal tissue—and the cell morphology, such as cell size and shape. In some cases, one tool may perform better than another. B. Is segmentation effective in analyzing post-Xenium H&E stained images?Theoretically, yes! 10X Genomics has provided a practical guideline on how to register an H&E image to a DAPI image (https://www.10xgenomics.com/analysis-guides/he-to-xenium-dapi-image-registration-with-fiji). However, this method does not perform well with a complete mouse brain, as image deformation still introduces an offset. We believe it may be effective in a smaller region and are actively working on enhancing this capability for our customers. To learn more about image deformation, please refer to the description of the bUnwarpJ plugin at https://imagej.net/plugins/bunwarpj/. C. Do you offer 3D segmentation?Nucleus 3D segmentation is feasible using Cellpose (https://www.cellpose.org/) and 10X Genomics has provided analysis guides on 3D segmentation (https://www.10xgenomics.com/analysis-guides/performing-3d-nucleus-segmentation-with-cellpose-and-generating-a-feature-cell-matrix). However, it requires significant resources, specifically GPUs and memory, and we were unable to segment a single mouse brain at full resolution. While it is possible to perform segmentation with a downsampled DAPI image, the results did not exceed those of 2D segmentation at lower resolutions. Thank you very much for your ongoing trust and collaboration. We look forward to using these exciting new capabilities to support your research! Spatially yours, The ASOC team |
||||||
