SpaCut: Transcript-Aware Morphology Fusion for Robust Cell Segmentation in Subcellular Spatial Transcriptomics

Subcellular Spatial Transcriptomics (SST) offers unprecedented opportunities to decipher tissue architecture at subcellular resolution. Consequently, accurate cell segmentation is a fundamental prerequisite for analyzing SST data, as precise transcript assignment is essential for downstream single-cell interpretation. Existing methods often overlook the spatial organization of transcripts, rely on simplified geometric priors, or struggle to coordinate transcript-derived features with morphology-derived spatial guidance. This limitation restricts their adaptability to irregular cell shapes, often leading to transcript misassignment. To address this, we propose SpaCut, a unified deep learning framework designed to effectively integrate transcriptomic feature learning with dynamic morphological guidance. SpaCut comprises a Gene-Wise Attention (GWA) module and a Nuclei-Guided Attentive Fusion (NGAF) module. GWA encodes the high-dimensional transcript map and aggregates features across multiple depths. NGAF uses the encoded transcript features to modulate distance-based nuclear softmasks and refine nucleus-anchored cell boundaries through progressive training. Extensive evaluations across seven datasets from diverse platforms (including Xenium, CosMx, MERSCOPE, and Stereo-seq) demonstrate that SpaCut significantly outperforms state-of-the-art methods in both segmentation accuracy and consistency with single-cell reference data. The source code is available at https://github.com/wenwenmin/SpaCut.