SlicerMorph photogrammetry: an open-source photogrammetry workflow for reconstructing 3D models

Area of Science:

• Paleontology
• Computational Biology
• 3D Imaging

Background:

• High-fidelity three-dimensional (3D) models of skeletal specimens are crucial for ecological and evolutionary research.
• Existing workflows for 3D model reconstruction can be complex and time-consuming.

Purpose of the Study:

• To present a fully open-source pipeline within the 3D Slicer platform for generating accurate 3D skeletal models.
• To improve the accuracy and efficiency of 3D model reconstruction for scientific analyses.

Main Methods:

• Developed a 3D Slicer extension integrating automatic image masking using the Segment Anything Model (SAM) and surface reconstruction via the OpenDroneMap (NodeODM) engine.
• Photographed 14 mountain-beaver skulls and reconstructed 3D models using the new pipeline and a previous workflow.
• Compared the accuracy of the reconstructed models against micro-CT references using mean surface distance, root mean square error (RMSE), Hausdorff, and Chamfer metrics.

Main Results:

• The improved pipeline, incorporating SAM masking, reduced mean surface distance and RMSE by 10-15% compared to the previous workflow.
• Visible artifacts around delicate skeletal elements, such as zygomatic arches, were reduced.
• Hausdorff distance showed minimal change, indicating global improvements rather than outlier reduction.

Conclusions:

• The new open-source pipeline offers a user-friendly and accurate method for 3D skeletal model reconstruction.
• This integrated workflow enhances the quality of 3D models for ecological and evolutionary studies.
• The pipeline streamlines the process of masking, scaling, and reconstruction within a single interface.