SpatialArtifacts: a computational framework for tissue artifact detection in spatial transcriptomics data

Spatial transcriptomics data are frequently compromised by technical artifacts, such as dry patches, tissue lifting, and uneven reagent coverage, which manifests as regions with low UMI counts, in particular at tissue borders. It can often be challenging to identify these regions using existing quality control methods. Here, we present SpatialArtifacts, a framework that combines median absolute deviation (MAD)-based outlier detection with mathematical morphology operations to identify and classify spatially contiguous tissue artifacts. Focal operations including 3x3 fill, 5x5 outline, and star-pattern connectivity link low-quality spots while preserving true biological domains. We use a hierarchical classification system to distinguish edge versus interior artifacts and large versus small regions, enabling downstream removal or targeted manual review. We demonstrate the performance of our method in human hippocampus, dorsolateral prefrontal cortex, and colorectal cancer tissues using 10x Genomics Visium and VisiumHD platforms. Our SpatialArtifacts package is freely available on Bioconductor at https://bioconductor.org/packages/SpatialArtifacts and on PyPI at https://pypi.org/project/spatial-artifacts/.