PRMix: Primary Region Mix Augmentation and Benchmark Dataset for Precise Whole Mouse Brain Anatomical Delineation.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• The mouse brain's structural similarity to the human brain makes it a vital model for neurological studies.
• Accurate delineation of brain regions in molecular images is crucial for atlas registration but remains challenging.
• Existing automated segmentation methods struggle with boundary ambiguity and limited data.

Purpose of the Study:

• To curate a high-resolution dual-fluorescence mouse brain microscopy (DMBM) dataset with expert annotations.
• To develop a novel data augmentation method (primary region mix - PRMix) to address segmentation challenges.
• To improve automated segmentation performance for detailed whole-brain analysis.

Main Methods:

• Curated a DMBM dataset of high-resolution, dual-fluorescence microscopy images.
• Annotated 118 subregions in parasagittal mouse brain sections.
• Developed and applied the PRMix data augmentation technique.

Main Results:

• The DMBM dataset offers detailed molecular and structural insights into the mouse brain.
• PRMix enhances synthetic data realism and minimizes region overlap.
• The combined approach achieved superior mouse brain segmentation performance compared to existing methods.

Conclusions:

• The DMBM dataset and PRMix method establish a new benchmark for mouse brain image segmentation.
• This work facilitates more accurate whole-brain analysis in neuroscience research.
• Improved segmentation supports the study of brain pathologies and regional specialization.