MEGA-FISH: multi-omics extensible GPU-accelerated FISH processing framework for huge-scale spatial omics

ABSTRACT Spatial omics enables comprehensive mapping of cell types and states in their spatial context, providing profound insights into cellular communication and tissue organization. However, analyzing large tissue sections, especially crucial for clinical applications, remains a significant challenge due to the computational demands of current image processing methods. To overcome these limitations, we developed MEGA-FISH, a flexible, GPU-accelerated Python framework optimized for large-scale spatial omics image analysis. Benchmarking on simulated and tissue images demonstrated that MEGA-FISH achieved high accuracy in spot detection while significantly reducing processing times compared with established tools. The framework’s adaptable computational capabilities optimize resource allocation (e.g., GPU or multi-core CPU) for diverse tasks, and its scalable architecture enables integration with advanced imaging and segmentation techniques. By bridging cutting-edge imaging methods and single-cell analysis, MEGA-FISH provides an efficient platform for multi-modal analysis and advances research and clinical applications of spatial omics at organ and organism scales.