Training Datasets: A High-Throughput Semi-Automated Bone Segmentation Workflow for Murine Hindpaw Micro-CT Datasets
The provided datasets in "Training Datasets" were used for training in an associated work, "A High-Throughput Semi-Automated Bone Segmentation Workflow for Murine Hindpaw Micro-CT Datasets", where we described a strategy using Amira software (ThermoFisher Scientific) to segment each individual bone in mouse hindpaws for downstream analysis. The training datasets represent 6 in vivo micro-CT image stacks (.dcm) of both hindpaws from 5-month-old male and female C57BL/6 mice, and depict the distribution of accuracy for the automated segmentation workflow before user intervention. In the associated study, a total of 42 in vivo datasets were analyzed (84 hindpaws) and demonstrated an automated segmentation accuracy of 79.2% on average (bones segmented correctly / total bones) with well-described correction processes for the errors. Analysis of ex vivo datasets improved the segmentation accuracy to 91.1% on average (n=6 hindpaws). The Supplementary Videos in the associated work provide 5 training videos (<2 hours total) to teach the segmentation process. After watching the training videos and practicing the segmentation process on the training datasets, new users to the Amira software exhibited accurate and high-throughput performance in the segmentation workflow. An additional dataset from a 2-month-old female C57BL/6 mouse is provided as "Example for 3D Visualization". The dataset is processed with a median filter in 3D, colormap limited to >2,500 Hounsfield units, and converted to 8-bit (.dcm file type). The labels represent the segmented version of the data, where the colormap ranges from 0 - 66 and each value represents a specific bone (.dcm file type). These datasets may be utilized in the associated published article for 3D visualization or available for download by readers. All data provided was acquired using a VivaCT 40 (Scanco Medical) using the following imaging parameters: 55kV, 145μA, 300ms integration time, 2048 x 2048 pixels, 1000 projections over 180°, resolution 17.5μm isotropic voxels. Each scan was completed in 30-45 minutes. All animal experiments were approved by the University Committee for Animal Resources at the University of Rochester.