Small Animal Multivariate Brain Analysis (SAMBA) - a High Throughput Pipeline with a Validation Framework.

Area of Science:

• Neuroscience
• Neuroimaging
• Computational Biology

Background:

• Voxel-based analysis (VBA) is crucial for preclinical neuroimaging but faces challenges due to high computational demands and parameter complexity.
• Current workflows often rely on intuition, with limited validation studies, impacting the robustness of small animal brain research.

Purpose of the Study:

• To develop a high-throughput, publicly shared voxel-based analysis (VBA) pipeline for preclinical neuroimaging.
• To establish a validation framework for quantifying the reliability and error rates of VBA in rodent models.
• To inform VBA workflow parameters and promote standardization in small animal neuroimaging.

Main Methods:

• Developed SAMBA, a high-throughput VBA pipeline utilizing a high-performance computing environment.
• Created a validation framework with morphological phantoms and four metrics to assess VBA reliability.
• Applied the framework to optimize VBA parameters for a mouse epilepsy model and explored standardization with human neuroimaging.

Main Results:

• SAMBA significantly increased computational efficiency, reducing processing time for large arrays from ~1 month to 1-3 days.
• The validation framework successfully quantified variability and reliability, addressing registration and template construction impacts.
• Optimized VBA parameters were determined for a mouse epilepsy model, demonstrating the framework's utility.

Conclusions:

• Verifying the accuracy of preclinical VBA is essential and requires community-wide attention.
• The proposed validation framework enhances quality assurance in preclinical neuroimaging.
• This work facilitates the generation and communication of robust, reproducible results in small animal brain studies.