SAMson: an automated brain extraction tool for rodents using SAM

Brain extraction, i.e. the precise removal of MRI signal outside the brain boundaries, is a key step in MRI preprocessing pipelines, typically achieved via masks delineating the region of interest (ROI). Existing automated tools often lack accuracy for rodent MRI due to resolution limitations, so large manual editing efforts are required. This work introduces SAMson, a high-precision automated mask generator built on Meta AI’s Segment-Anything Model (SAM). SAM’s adaptability to diverse tasks, akin to other foundation models (Chat-GPT), is harnessed to address the scarcity of training data in this domain. SAMson is a Python-based tool that integrates SAM’s capabilities with the requirements and characteristics of multislice MRI data. SAMson offers two prompt generation approaches: Semi-Auto, enabling manual prompt selection for precise control, and Full-Auto, with automated prompt generation. Evaluation against gold-standard masks extracted by an experienced experimenter demonstrated excellent performance of SAMson, and benchmarking against established methods (BET from FSL) demonstrated SAMson’s superiority. Robustness assessments on datasets obtained from an external MRI facility, confirmed versatility across scanner setups and image resolutions. Our tool’s adaptability and performance underscore its potential for widespread adoption in rodent MRI research, supported by open-source availability for the scientific community.