Multimodal Correlative Preclinical Whole Body Imaging and Segmentation

Area of Science:

• Biomedical Imaging
• Machine Learning
• Preclinical Research

Background:

• Accurate segmentation of anatomical structures is crucial for quantitative image analysis in preclinical research.
• Existing methods often struggle with multimodal imaging and atlas-free segmentation of small animals.

Purpose of the Study:

• To develop a novel, automated whole-body segmentation method for small animals using multimodal imaging.
• To improve the accuracy and generalizability of segmentation for preclinical research applications.

Main Methods:

• A machine learning framework integrating MR, CT, and optical imaging data.
• Hierarchical agglomerative clustering for supervoxel generation.
• Support Vector Machine-k-Nearest Neighbors (SVM-kNN) classifiers with heatmap priors for segmentation.

Main Results:

• Successful segmentation of multiple organs (heart, lungs, liver, kidneys, etc.) and skeletal structures in mice.
• Demonstrated superior performance compared to state-of-the-art methods in experimental validation.
• The system achieved automatic, atlas-free segmentation.

Conclusions:

• The proposed method offers a robust and generalizable solution for small animal image segmentation across various modalities.
• This advancement facilitates quantitative analysis and broadens applications in preclinical studies.
• The system's atlas-free nature enhances its utility for diverse research needs.