A Semi-automatic Interactive Registration Software for 3D Biological Images

The complexity and variability inherent in 3D biomedical images present significant challenges for image registration. Despite progress in automatic registration technologies, accurately registering three-dimensional biological images, especially those with complex structures and derived from multiple imaging modalities, still necessitates manual interaction for optimal precision. Current research on interactive registration predominantly focuses on 2D slices, resulting in a loss of inter-slice constraint information for 3D biological images, and requires a lot of manual operation, resulting in low efficiency. This paper presents a pioneering semi-automatic interactive software for 3D biological image registration. Our approach innovatively integrates 3D point cloud extraction, OpenGL-based visualization, effective interaction methodologies, and non-rigid deformation techniques underpinned by both local and global constraints. A standout feature of our system is the introduction of a 3D magnetic lasso tool, which employs Gaussian constraints to significantly enhance the efficiency of interactive registration in 3D spaces. Our software not only boosts registration accuracy but also minimizes manual labor. By melding automated processes with human oversight, our solution strikes a balance between efficiency and accuracy. The validity of the work in this paper is verified by application, which provides a promising semi-automatic biological 3D image registration tool for medical research and diagnosis.