SegRef3D Lite: A browser-based workflow for artificial intelligence–assisted segmentation and quantitative three-dimensional morphological analysis

Abstract Segmentation is a major bottleneck in three-dimensional morphological analysis. Although artificial intelligence–assisted segmentation workflows can be used to address this issue, their implementations are constrained by dedicated software environments or local computing resources. To address these challenges, we developed SegRef3D Lite, a browser-based platform that integrates two complementary artificial intelligence–assisted segmentation approaches (namely, SegAnything and SegCT/MRI) within a common human-in-the-loop workflow. SegAnything provides general-purpose, prompt-based segmentation by using Segment Anything Model 2, and SegCT/MRI provides anatomy-specific segmentations of computed tomography and magnetic resonance imaging datasets by using TotalSegmentator. Both approaches generate segmentation masks that can be imported into the same browser-based workspace for visual review, manual correction, quantitative volume measurement, three-dimensional visualization, and surface-model export. In this workspace, routine image handling and refinement are performed locally in the browser, whereas artificial intelligence inference is executed through user-initiated Google Colaboratory workflows, thus eliminating the need for a compatible graphics processing unit on a local computer. We applied our platform to serial block-face photographs, volumetric computed tomography, serial light microscopy, serial electron microscopy, and human abdominal computed tomography. Across these diverse imaging modalities and spatial scales, our platform is able to process outputs from different segmentation approaches by using the same downstream workflow for review, refinement, quantitative analysis, and three-dimensional reconstruction. Therefore, SegRef3D Lite provides an accessible framework for integrating complementary artificial intelligence–assisted segmentation strategies with researcher-guided verification and quantitative morphological analysis. Mini-abstract : SegRef3D Lite integrates general-purpose and anatomy-specific artificial intelligence–assisted segmentation with browser-based human refinement, volume measurement, and three-dimensional reconstruction across diverse morphological imaging modalities.