A preclinical micro-computed tomography database including 3D whole body organ segmentations

Area of Science:

• Preclinical imaging
• Medical image analysis
• Bioinformatics

Background:

• Manual organ segmentation in preclinical micro-computed tomography (μCT) is the current standard but is inefficient, prone to errors, and lacks reproducibility.
• The absence of robust automated segmentation algorithms hinders large-scale analysis of μCT data.
• Developing automated methods is crucial for advancing preclinical research and reducing variability.

Purpose of the Study:

• To introduce a comprehensive database of 3D whole-body μCT scans from mice.
• To provide manual organ segmentation for key organs and subcutaneous tumors within these scans.
• To facilitate the development and validation of automated organ segmentation algorithms.

Main Methods:

• Compilation of 225 murine 3D whole-body μCT scans.
• Manual delineation of major organs (heart, liver, lung, trachea, spleen, kidneys, stomach, intestine, bladder, thigh muscle, bone) and tumors.
• Inclusion of both native and contrast-enhanced (spleen, liver) μCT data.

Main Results:

• A publicly accessible database containing 225 annotated 3D μCT datasets.
• Detailed manual segmentation masks for multiple organs and tumors across the dataset.
• Data is available on the Figshare online repository for research use.

Conclusions:

• The released database serves as a valuable resource for developing and evaluating automated organ segmentation techniques.
• This initiative aims to improve the efficiency, reproducibility, and scalability of preclinical μCT data analysis.
• By providing a reliable reference dataset, the study promotes reduced manual effort and potentially fewer laboratory animals in research.