AbstractRecent advances in microscopy techniques, especially in electron microscopy, are transforming biomedical studies by acquiring large quantities of high-precision 3D cell image stacks. However, to study cell morphology and connectivity in organs such as brains, scientists must first perform cell segmentation, which involves extracting individual cell regions of various shapes and sizes from a 3D image. This remains a great challenge because automatic cell segmentation can contain numerous errors, even with advanced deep learning methods. For biomedical research that requires cell segmentation in large 3D image stacks, an efficient semi-automated software solution is still needed. We created Seg2Link, which generates automatic segmentations based on deep learning predictions and allows users to quickly correct errors in the segmentation results. It can perform automatic instance segmentation of 2D cells in each slice, 3D cell linking across slices, and various manual corrections, in order to efficiently transform inaccurate deep learning predictions into accurate segmentation results. Seg2Link’s data structure and algorithms were also optimized to process 3D images with billions of voxels on a personal computer quickly. Thus, Seg2Link offers a simple and effective way for scientists to study cell morphology and connectivity in 3D image stacks.