Spatial Omics Data Analysis: From Image Processing to Biological Interpretation

Spatial omics data analysis connects molecular measurements with their physical location within tissue, requiring a workflow that moves from raw image files and count matrices through quality control, segmentation, normalization, and spatial statistics before biological interpretation is possible. This article outlines a practical pipeline for researchers and analysts working with spatial transcriptomics, spatial proteomics, and related data types, with emphasis on tool selection, troubleshooting, and reproducible reporting.

Scope and Reader Context

Spatial omics technologies generate data at scales ranging from subcellular to organismal, integrating imaging with molecular profiling across sequencing and mass spectrometry platforms. For bioinformaticians, the analytical challenge is twofold: processing image-derived information accurately and integrating that spatial context with molecular measurements. The workflow described here applies to imaging-based platforms such as Xenium In Situ, sequencing-based approaches like Stereo-seq, and mass spectrometry-based spatial proteomics. Analysts working with plant tissues face additional considerations because rigid cell walls and variable cell sizes require adaptation of methods developed for mammalian samples. The pipeline steps, quality checks, and troubleshooting guidance below apply across platforms, with platform-specific adjustments noted where relevant.

At a Glance: Spatial Omics Analysis Workflow

Core Principles of Spatial Omics Data Analysis

Spatial Resolution Determines Analytical Strategy

The resolution of a spatial omics dataset dictates which analytical methods are appropriate. Imaging-based platforms can achieve subcellular resolution, allowing detection of transcripts or proteins at single-molecule level within individual cells. Sequencing-based approaches vary in resolution, with some capturing spots that contain multiple cells. The Xenium In Situ platform, for example, maps hundreds of genes at subcellular resolution, while other commercial platforms capture spatial information at different granularity. Analysts must match their segmentation strategy and downstream interpretation to the actual resolution of their platform instead of assuming single-cell resolution universally.

Image Analysis and Molecular Data Are Interdependent

Spatial omics analysis requires treating image-derived information and molecular measurements as a single integrated dataset. Cell segmentation quality directly affects the accuracy of cell-by-gene expression matrices. Poor segmentation can merge adjacent cells, creating artificial doublets, or split single cells, diluting signal across multiple units. Conversely, molecular information can inform image analysis. The DISSECT approach demonstrates this principle by integrating cytological images with spatial transcriptomic profiles to improve segmentation accuracy, using image features and transcriptome-derived information to refine cell boundaries.

Spatial Context Adds Analytical Dimensions Beyond Single-Cell Data

Standard single-cell RNA sequencing analysis addresses cellular heterogeneity but loses tissue architecture. Spatial omics preserves the physical arrangement of cells, enabling analysis of neighborhood relationships, tissue domains, and cell-cell communication. Spatial domain identification, which segments tissue into functionally distinct regions, is a fundamental task that requires integrating molecular similarity with spatial proximity. Methods that ignore spatial context or rely on single omics layers face substantial limitations when applied to spatial multi-omics data.

Practical Workflow for Spatial Omics Analysis

Step 1: Data Acquisition and Format Familiarization

Before analysis begins, document the platform, tissue type, species, and data format. Spatial omics datasets typically include raw or processed images, transcript or protein coordinate files, and count matrices. Familiarize yourself with the output structure of your specific platform. For imaging-based platforms, understand the relationship between image pixels and physical distance. For sequencing-based approaches, note the spot size and spacing, as these determine whether each spot represents one cell or multiple cells.

Store raw data according to FAIR principles, ensuring findability, accessibility, interoperability, and reusability. The FAIR Guiding Principles provide a framework for data management that supports reproducibility and sharing. Document file naming conventions, metadata standards, and storage locations before analysis begins.

Step 2: Image Preprocessing

Image preprocessing prepares raw images for segmentation or spot detection. Key tasks include:

• Background subtraction to remove autofluorescence and imaging noise
• Channel alignment to correct for chromatic aberration across fluorescence channels
• Tissue region identification to exclude background areas from analysis
• Artifact masking to remove debris, folds, or damaged tissue regions

For mass spectrometry-based spatial proteomics, preprocessing includes peak detection, mass calibration, and pixel alignment. The analytical workflow must accommodate the specific data structure of each omics type.

Record all preprocessing parameters, including background subtraction methods, alignment algorithms, and artifact detection thresholds. These parameters affect downstream results and must be documented for reproducibility.

Step 3: Cell Segmentation and Spot Detection

Cell segmentation identifies individual cell boundaries within tissue images. This step is critical because segmentation errors propagate through all downstream analyses. Several approaches exist:

• Nuclear segmentation followed by cell boundary expansion
• Whole-cell segmentation using membrane or cytoplasmic markers
• Deep learning-based segmentation trained on annotated images
• Integration of image features with transcript or protein spatial distributions

The DISSECT model exemplifies the integration approach, using a pretrained deep generative model to denoise multiscale image features and predicting cell instances with an instance-aware detection module. Benchmarking across multiple datasets showed that this integrated approach achieved higher mean average precision than several existing segmentation tools.

For sequencing-based platforms without images, spot detection assigns transcripts or proteins to spatial coordinates. The resolution of spot detection determines whether downstream analysis operates at subcellular, cellular, or multicellular level.

Segmentation quality assessment should include visual inspection of cell boundaries overlaid on tissue images, comparison of cell counts with expected values from histology, and examination of marker gene distributions. If segmentation tools produce inconsistent results across samples, consider whether tissue processing or staining variation is the cause.

Step 4: Quality Control

Quality control for spatial omics data addresses both molecular and spatial dimensions. Molecular quality metrics include:

• Total transcripts or proteins detected per cell or spot
• Number of genes detected per cell or spot
• Mitochondrial or housekeeping gene proportions
• Duplicate or dropout rates

Spatial quality metrics include:

• Coverage of tissue area by detected cells or spots
• Uniformity of detection across tissue regions
• Presence of spatial artifacts or batch effects

Filtering thresholds should be established based on the distribution of these metrics in your dataset instead of applying arbitrary cutoffs. Examine distributions visually and set thresholds that remove clear outliers while retaining biologically meaningful variation.

The Xenium In Situ platform has been evaluated across multiple tissues and species, with benchmarking of open-source computational tools for preprocessing, cell segmentation, selection of spatially variable features, and domain identification. These evaluations provide guidance on expected data quality and tool performance, but each dataset requires individual assessment.

Step 5: Normalization

Normalization adjusts for technical variation in detection efficiency across cells or spots. Common approaches include:

• Library size normalization, scaling counts to a common total
• Log transformation with pseudocount addition
• Variance stabilizing transformations
• Platform-specific normalization methods

For spatial data, normalization must account for spatial variation in detection efficiency. Some regions of tissue may have systematically lower detection due to tissue thickness, permeabilization differences, or imaging artifacts. Spatial normalization methods adjust for these gradients.

Record normalization parameters and compare normalized distributions across samples to identify batch effects. If samples were processed in multiple batches, consider batch correction methods appropriate for spatial data.

Step 6: Dimension Reduction and Feature Selection

Dimension reduction projects high-dimensional expression data into a lower-dimensional space for visualization and clustering. Principal component analysis is commonly used as an initial step, followed by nonlinear methods such as UMAP or t-SNE for visualization.

Spatially variable feature selection identifies genes or proteins whose expression varies meaningfully across tissue space. These features are particularly informative for spatial domain identification and tissue architecture analysis. Methods for spatially variable feature detection use spatial coordinates in addition to expression values to identify patterns that would be missed by nonspatial approaches.

Step 7: Spatial Clustering and Domain Identification

Spatial clustering groups cells or spots into domains based on combined molecular similarity and spatial proximity. This step differs from standard single-cell clustering because spatial information constrains which cells can be grouped together.

Several algorithmic approaches exist:

• Graph-based methods that construct spatial neighbor graphs and apply community detection
• Methods that integrate molecular profiles with spatial coordinates through graph neural networks
• Multi-omics integration methods that combine information from multiple molecular layers

The SIMID framework exemplifies multi-omics spatial domain identification, using graph mutual information to capture cellular spatial proximity and molecular profile similarity, then employing deep subspace learning to construct cell networks for each omics layer. This approach decomposes the cell multi-layer network into consistent and complementary structures, enabling domain identification from spatial multi-omics data.

Similarly, information-aware multi-view contrastive learning methods optimize spatial and feature neighborhood graphs for each omics, learning fused representations that integrate shared and unique information across molecular layers.

For spatial domain identification, validate results by examining whether identified domains align with known tissue anatomy, whether marker genes show expected spatial patterns, and whether domain boundaries are biologically plausible.

Step 8: Cell Type Annotation

Cell type annotation assigns biological identities to clusters or domains. Approaches include:

• Marker gene-based annotation using known cell type markers
• Reference-based annotation transferring labels from single-cell RNA sequencing data
• Manual annotation guided by literature and histology

For spatial data, annotation should incorporate spatial context. Cell types may show characteristic spatial distributions, such as immune cells infiltrating specific tissue regions or epithelial cells organized into glandular structures. Spatial validation of annotations strengthens confidence in cell type assignments.

Step 9: Downstream Spatial Analysis

Once cells are annotated and domains identified, downstream analyses address biological questions:

• Differential expression between spatial domains or conditions
• Cell-cell communication analysis using ligand-receptor pairs with spatial constraints
• Neighborhood analysis examining cell type composition around each cell
• Trajectory or pseudotime analysis incorporating spatial information
• Integration with histopathology for disease profiling

Spatial statistics quantify patterns such as colocalization, segregation, or gradient distributions. These analyses require careful definition of spatial neighborhoods and appropriate statistical controls for spatial autocorrelation.

Step 10: Integration with Single-Cell and Other Omics Data

Integrating spatial omics data with single-cell RNA sequencing data can enhance cell type annotation and provide deeper molecular characterization. Single-cell data offers higher gene coverage per cell, while spatial data provides tissue context. Integration methods map cells between datasets based on shared molecular features.

Multi-omics integration combines spatial transcriptomics with spatial proteomics, epigenomics, or metabolomics. Each omics layer provides complementary information, and integration methods must account for different data structures, resolutions, and noise profiles. The single-cell analyst platform supports integrated analysis of multiple single-cell omics types including spatial transcriptomics, enabling researchers to perform comprehensive analyses without requiring programming skills.

Tool Comparison and Selection

Seurat

Seurat is an R package widely used for single-cell and spatial transcriptomics analysis. It provides functions for quality control, normalization, dimension reduction, clustering, and visualization. Spatial analysis in Seurat includes integration with image data, spatial feature visualization, and spatial clustering methods. Seurat is appropriate for researchers comfortable with R and those working primarily with sequencing-based spatial platforms.

Scanpy

Scanpy is a Python package for single-cell and spatial omics analysis. It offers scalable analysis of large datasets and integrates with the broader Python scientific ecosystem. Scanpy provides preprocessing, clustering, trajectory inference, and differential expression tools. Its command-line and scriptable interface suits researchers who prefer Python or need to process very large datasets.

Squidpy

Squidpy is a Python framework specifically designed for spatial omics analysis. It brings together tools from omics and image analysis to enable scalable description of spatial molecular data, such as transcriptomes or multivariate proteins. Squidpy provides efficient infrastructure and numerous analysis methods for storing, manipulating, and interactively visualizing spatial omics data. It is extensible and can interface with existing libraries for scalable analysis. Squidpy is particularly well suited for researchers working with imaging-based spatial platforms and those needing image analysis capabilities integrated with molecular data analysis.

Platform-Specific Tools

Many spatial omics platforms provide proprietary analysis software. These tools offer streamlined workflows but may limit flexibility and reproducibility. Independent benchmarking of open-source tools applied to Xenium datasets has provided recommendations for preprocessing, cell segmentation, spatially variable feature selection, and domain identification. Consider using open-source tools for critical analysis steps to enable reproducibility and method transparency.

Web-Based Platforms

For researchers without programming expertise, web-based platforms offer accessible alternatives. The single-cell analyst platform supports multiple single-cell omics types including spatial transcriptomics, automating quality control, data processing, and phenotype-specific analyses while providing interactive visualizations. SeekSoul Online similarly provides a comprehensive platform for single-cell multi-omics analysis with automated workflows and report generation. These platforms reduce the learning curve for spatial omics analysis but may limit customization compared with code-based approaches.

Tool Selection Criteria

Choose tools based on:

• Data type and platform compatibility
• Required analysis features
• Programming language preference
• Computational resources available
• Reproducibility and documentation requirements
• Community support and maintenance status

Document tool versions and parameters for all analysis steps. Version changes can alter results, so record the specific versions used.

Records and Measurements

Analysis Documentation

Maintain a complete record of the analysis workflow, including:

• Raw data file locations and checksums
• Software versions and installation details
• Parameter values for each analysis step
• Quality control metrics and filtering decisions
• Normalization methods and settings
• Clustering parameters and seed values
• Annotation sources and evidence

This documentation supports reproducibility and enables troubleshooting when results are unexpected.

Quality Metrics to Track

Track these measurements throughout the analysis:

• Cell or spot counts before and after filtering
• Median transcripts or proteins per cell or spot
• Genes detected per cell or spot
• Segmentation quality metrics such as overlap with nuclear stains
• Spatial coverage statistics
• Batch effect metrics when multiple samples are analyzed

Compare these metrics across samples and batches to identify technical variation that could confound biological interpretation.

Version Control

Use version control for analysis scripts and configuration files. Record the exact code used to generate each result. For interactive analyses, document the session history or use notebooks that capture code and output together.

Common Failure Patterns and Troubleshooting

Segmentation Failures

Segmentation errors manifest as abnormal cell counts, irregular cell size distributions, or poor marker gene localization. If segmentation produces too few cells, check whether the tissue was properly permeabilized or whether the segmentation algorithm parameters are appropriate for your tissue type. If segmentation produces too many small objects, examine whether debris or background signal is being called as cells. The DISSECT approach demonstrates that integrating image features with transcriptomic profiles can improve segmentation accuracy, suggesting that segmentation failures may be addressed by incorporating molecular information into the segmentation process.

Batch Effects Between Samples

When multiple samples show systematic differences unrelated to biology, check whether they were processed in different batches. Batch effects can arise from differences in tissue processing, reagent lots, imaging sessions, or sequencing runs. Document batch information and consider batch correction methods appropriate for spatial data.

Spatial Artifacts

Tissue folds, tears, or uneven permeabilization create spatial artifacts that appear as regions with abnormal detection. Examine spatial maps of total counts to identify artifact regions. Consider masking these regions before downstream analysis or applying spatial normalization methods that account for detection gradients.

Low Gene Detection

Some spatial platforms detect relatively few genes per cell compared with single-cell RNA sequencing. This limitation affects cell type annotation and differential expression power. If gene detection is low, consider whether the platform gene panel is appropriate for your biological question or whether analysis methods need adjustment for sparse data.

Domain Boundary Ambiguity

Spatial domains may not have sharp boundaries, particularly in tissues with gradual transitions. If domain identification produces unstable results, examine whether the spatial neighborhood definition is appropriate. Methods that integrate multiple omics layers may provide more stable domain identification than single-omics approaches.

Computational Resource Limitations

Large spatial datasets require substantial memory and processing time. If analyses fail due to resource constraints, consider downsampling, using more efficient algorithms, or moving to a computing cluster. Squidpy was designed for scalable analysis of spatial omics data, and tools like it may handle large datasets more efficiently than general-purpose single-cell tools.

Limitations and Interpretation Boundaries

Resolution Constraints

The spatial resolution of the platform limits the biological questions that can be addressed. Subcellular resolution enables analysis of transcript localization within cells, while multicellular spot resolution only supports tissue-level analysis. Do not interpret multicellular spots as single cells.

Gene Panel Coverage

Targeted spatial platforms measure a predefined gene panel instead of the full transcriptome. This limitation affects discovery potential and may bias cell type annotation toward well-characterized markers. Results are constrained by the genes included in the panel.

Sensitivity Tradeoffs

Spatial platforms face an inherent tension between spatial resolution and molecular sensitivity. Higher resolution often comes with lower detection efficiency. This tradeoff affects the ability to detect lowly expressed genes and must be considered when interpreting negative results.

Sample Quality Dependence

Spatial omics results depend heavily on tissue quality. Human postmortem tissue, for example, may have RNA degradation that affects detection. The scarcity and insufficient standardization of human samples presents challenges for spatial analysis, particularly in neurological research. Document sample quality metrics and consider their impact on interpretation.

Cross-Species Differences

Spatial analysis methods developed for mammalian tissues may not transfer directly to other organisms. Plant tissues, with rigid cell walls and variable cell sizes, require adaptation of analytical methods. Similarly, differences between human and mouse models must be considered when interpreting spatial data.

Statistical Considerations

Spatial data violate the independence assumption of many standard statistical tests because nearby measurements are correlated. Spatial statistics must account for this autocorrelation. Methods that ignore spatial dependence may produce inflated significance values.

Safety and Regulatory Context

Data Sharing and Privacy

Spatial omics data derived from human samples may contain sensitive information. The NIH Genomic Data Sharing Policy provides requirements for sharing genomic data while protecting participant privacy. Review applicable policies before depositing or sharing spatial omics data. De-identification of spatial images requires careful consideration because tissue morphology may be recognizable.

Data Management Standards

Follow FAIR Guiding Principles for data management to ensure that spatial omics data are findable, accessible, interoperable, and reusable. These principles support scientific reproducibility and enable data sharing across research groups. Document metadata thoroughly, including tissue source, processing protocols, platform parameters, and analysis methods.

Reproducibility Requirements

Funding agencies and journals increasingly require reproducible analysis workflows. Maintain complete records of analysis methods, including software versions and parameters. Consider depositing analysis code and processed data in public repositories to support verification and reuse.

Professional Escalation Criteria

Seek expert consultation when:

• Segmentation quality is poor across multiple parameter settings
• Batch effects cannot be resolved with standard correction methods
• Spatial domain identification produces biologically implausible results
• Integration across omics layers fails to converge or produces conflicting results
• Results have clinical implications that require validation in additional cohorts

For clinical translation, spatial omics findings require validation in independent cohorts and appropriate regulatory oversight. The transition of spatial omics from basic research to clinical practice, including biomarker discovery and diagnostics implementation, involves challenges that extend beyond computational analysis.

Frequently Asked Questions

What is the difference between spatial transcriptomics and single-cell RNA sequencing analysis?

Single-cell RNA sequencing measures gene expression in individual cells but loses the spatial context of where those cells were located in tissue. Spatial transcriptomics preserves the physical location of measurements, enabling analysis of tissue architecture, cell neighborhoods, and spatial gradients. Analysis of spatial data requires additional steps for image processing, spatial clustering, and spatial statistics that are not needed for standard single-cell data.

How do I choose between imaging-based and sequencing-based spatial platforms?

The choice depends on your biological question, required resolution, gene panel size, and available resources. Imaging-based platforms can achieve subcellular resolution and directly visualize transcripts or proteins in tissue context. Sequencing-based approaches may offer higher gene throughput but often at lower spatial resolution. Consider whether your question requires single-cell resolution, how many genes you need to measure, and whether your tissue type is compatible with each platform.

What is cell segmentation and why does it matter for spatial analysis?

Cell segmentation identifies the boundaries of individual cells in tissue images. This step determines which detected transcripts or proteins are assigned to which cell, directly affecting the accuracy of cell-by-gene expression matrices. Poor segmentation can merge adjacent cells or split single cells, creating artifacts that propagate through all downstream analyses. Segmentation quality should be visually inspected and quantitatively assessed before proceeding.

How should I normalize spatial omics data?

Normalization adjusts for technical variation in detection efficiency across cells or spots. Library size normalization scales counts to a common total, and log transformation stabilizes variance. Spatial data may require additional normalization to account for spatial gradients in detection efficiency caused by tissue thickness or permeabilization differences. Choose normalization methods appropriate for your platform and validate that normalized data show no systematic spatial artifacts.

What are spatial domains and how are they identified?

Spatial domains are functionally distinct regions within tissue that share molecular and structural characteristics. Domain identification groups cells or spots based on combined molecular similarity and spatial proximity. Methods range from graph-based community detection to deep learning approaches that integrate multiple omics layers. Validating identified domains against known tissue anatomy and marker gene patterns is essential.

Can I integrate spatial data with single-cell RNA sequencing data?

Yes, integration of spatial and single-cell data is a common strategy. Single-cell RNA sequencing provides deeper gene coverage per cell, while spatial data provides tissue context. Integration methods map cells between datasets based on shared molecular features, enabling cell type annotation in spatial data using reference information from single-cell data. This approach combines the strengths of both technologies.

What tools are available for researchers who do not program?

Web-based platforms such as single-cell analyst and SeekSoul Online provide graphical interfaces for spatial omics analysis without requiring programming skills. These platforms automate quality control, data processing, and visualization while supporting multiple omics data types. They reduce the learning curve for spatial analysis but may offer less flexibility than code-based approaches like Seurat, Scanpy, or Squidpy.

How do I troubleshoot poor spatial domain identification results?

First, examine whether segmentation quality is adequate, since errors here propagate downstream. Check whether the spatial neighborhood definition matches the biological scale of tissue organization. Consider whether the gene panel includes sufficient informative markers for your tissue type. If single-omics domain identification is unstable, methods that integrate multiple omics layers may provide more consistent results. Validate domains against known anatomy and marker gene distributions.

Related Bioinformatics Guides

• Cryo-EM Image Processing and 3D Reconstruction
• Alternative Splicing Analysis from RNA-Seq Data
• Hidden Markov Models in Biological Sequence Analysis
• Spatial Transcriptomics Alignment and Cellular Neighborhood Analysis
• MicroRNA Target Prediction Tools

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Single-cell omics: experimental workflow, data analyses and applications.. Science China. Life sciences, 2025.
• Mass-spectrometry-based proteomics: from single cells to clinical applications.. Nature, 2025.
• Spatial proteomics in neurons at single-protein resolution.. Cell, 2024.
• Optimizing Xenium In Situ data utility by quality assessment and best-practice analysis workflows.. Nature methods, 2025.
• Guidelines for bioinformatics of single-cell sequencing data analysis in Alzheimer's disease: review, recommendation, implementation and application.. Molecular neurodegeneration, 2022.
• Advances in plant spatial multi-omics data analysis.. Trends in plant science, 2026.
• Comprehensive analysis of multi-omics single-cell data using the single-cell analyst.. iMeta, 2025.
• Squidpy: a scalable framework for spatial omics analysis.. Nature methods, 2022.
• Opportunities and challenges for the application of single-cell RNA sequencing and spatial transcriptomics in ischemic stroke.. 2026.
• Integrative single-cell and spatial transcriptomics combined with machine learning to discover complement-associated biomarkers in temporal lobe epilepsy with hippocampal sclerosis.. 2026.
• Integrating cytological images and spatial transcriptomics for cell segmentation with DISSECT.. 2026.
• Enablers and challenges of spatial omics, a melting pot of technologies. Molecular Systems Biology, 2023.
• Identifying Spatial Domains From Spatial Multi-Omics Data With Graph Mutual Information and Deep Subspace Learning. IEEE Transactions on Computational Biology and Bioinformatics, 2025.
• SeekSoul Online: A user-friendly bioinformatics platform focused on single-cell multi-omics analysis. The Innovation Life, 2025.
• Spatial Multi-Omics Integration Via Information-Aware Multi-View Contrastive Learning. IEEE International Conference on Bioinformatics and Biomedicine, 2025.
• Visual cohort comparison for spatial single-cell omics-data. IEEE Transactions on Visualization and Computer Graphics, 2021.
• Next-Generation Disease Profiling by Integrating Histopathology with Spatial Multi-Omics Data. American Journal of Pathology, 2026.
• Galaxy single-cell & spatial omics community update: Navigating new frontiers in 2025. Cell Genomics, 2025.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.