ClusterMap Spatial Transcriptomics Cell Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for cell segmentation in spatial transcriptomics struggle with precise and automated assignment of RNAs into individual cells, often requiring manual curation or training datasets, limiting the integration of high-dimensional transcriptomic data into low-dimensional biological patterns.
Innovation Solution
The ClusterMap framework uses spatially resolved RNA patterns to segment cells and subcellular structures without fluorescent staining, employing a point pattern analysis that incorporates physical proximity and gene identity for unsupervised clustering across diverse tissue types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If fluorescent staining and conventional segmentation methods are used, then cell segmentation can be performed, but manual curation and training datasets are required, reducing automation
Solution Approach 1:
The system uses the spatial transcriptomics data itself to perform segmentation without requiring external fluorescent staining or manual training datasets. The RNA spatial patterns inherently contain the information needed to identify cell boundaries and structures, allowing the method to be self-sufficient and fully automated
Solution Approach 2:
The method extracts cell segmentation information directly from the spatial transcriptomics data by removing the need for auxiliary fluorescent staining procedures. The spatial distribution of RNAs is used to infer cell boundaries, extracting segmentation capability from the transcriptomic signal itself
2Loss of information
If high-dimensional spatial transcriptomic data is collected, then comprehensive gene expression information is obtained, but extracting low-dimensional biological patterns becomes challenging
Solution Approach 1:
The high-dimensional spatial transcriptomics data is segmented into discrete cellular and subcellular units based on spatial patterns of RNA distribution. This segmentation transforms the continuous high-dimensional data into discrete low-dimensional representations corresponding to individual cells, nuclei, and other biological structures
Solution Approach 2:
The method uses spatial dimension information from the transcriptomics data to reduce dimensionality. By analyzing the spatial coordinates and distribution patterns of RNAs in 2D or 3D space, the system extracts meaningful biological patterns while reducing the complexity of the high-dimensional gene expression data
3Measurement precision
If auxiliary fluorescent staining is used for cell segmentation, then cell boundaries can be identified, but the complexity of the procedure increases and manual labeling is required
Solution Approach 1:
The method extracts cell boundary information directly from the spatial transcriptomics data by removing the need for auxiliary fluorescent staining. The spatial distribution and density patterns of RNAs within cells provide sufficient information to identify cell boundaries without additional staining steps
Solution Approach 2:
The spatial transcriptomics data serves dual purposes: both gene expression analysis and cell segmentation. The RNA spatial patterns inherently encode cell boundary information, allowing the data to perform the segmentation function without requiring separate fluorescent staining procedures
Data Source
Figure 1A~1B
Figure 1C
Figure 1D
AI summary
The present disclosure provides methods for identifying cells in an image. An apparatus for identifying cells in an image is also provided by the present disclosure. Further provided herein is a non-transitory computer-readable storage medium for performing the methods disclosed herein. Methods of diagnosing a disease or disorder and of treating a disease or disorder in a subject using the methods disclosed are also provided herein.