Advancing RNA FISH image analysis with 3D deep learning

RNA fluorescence in situ hybridization (RNA FISH) is a primary method by which gene expression is analyzed, gathering information on individual cell expression from classification to localization of individual RNA transcripts. Recent strides in experimental techniques have yielded comprehensive spatial transcriptomic data from entire organs, e.g. the mouse brain. However, analysis of such expansive data sets presents intrinsic challenges, arising from the image size and high RNA density. Current state-of-the-art analysis methods exhibit one or more limitations when used on such large data sets, including: 1) an inability to leverage 3D information, 2) compromised performance at high RNA densities, and 3) prohibitive computational cost scaling with image size.