Atlas-based liver segmentation for nonhuman primate research.

Area of Science:

• Medical imaging
• Preclinical research
• Infectious disease

Background:

• Viral infections can cause systemic damage, particularly affecting the liver.
• Liver damage from viral infections is often heterogeneous, necessitating whole-organ analysis.
• Manual liver segmentation in preclinical imaging is labor-intensive and challenging.

Purpose of the Study:

• To develop an automated liver segmentation method for nonhuman primates.
• To support quantitative image analysis in preclinical infectious disease research.
• To overcome the limitations of manual liver delineation.

Main Methods:

• An atlas-based liver segmentation technique was developed using 82 nonhuman primate CT scans.
• The method employs rigid and deformable registrations, a majority vote algorithm, and post-processing.
• Evaluation used Dice similarity, Hausdorff distance, and Bland-Altman plots.

Main Results:

• The automated segmentation achieved a median Dice score of 0.91, comparable to manual contouring.
• The method can detect liver changes as small as 3 Hounsfield units (CT) and 0.4 SUVmean (PET).
• These detectable changes are significantly smaller than those typically observed in viral disease responses.

Conclusions:

• The atlas-based segmentation is generalizable across nonhuman primate sizes and species.
• This technique facilitates quantitative analysis in preclinical infectious disease studies.
• The methodology can be implemented using open-source frameworks.