BrainSuite is an automated tool for generating cortical surface representations from T1-weighted MR images of the human brain. It performs a sequence of low-level operations, including skull and scalp removal, image nonuniformity compensation, voxel-based tissue classification, topological correction, rendering, and editing. The tool requires minimal user interaction and produces accurate brain segmentations in clinical time. The paper describes the theory behind each stage of the cortical surface identification process and presents validation results using real and phantom data. The tool addresses challenges in extracting cortical surfaces from MR images, such as measurement noise, partial volume effects, and image nonuniformity. It uses a parametric model to estimate local variations in the gain field of the image by comparing the intensity properties of the whole MR volume with local neighborhoods. This model incorporates spatially variant bias terms and uses a mixture model to account for partial volume effects. The tool also uses a Gibbs' prior to incorporate a spatial model for brain tissues into the classification scheme. The tool includes a Topological Constraint Algorithm (TCA) that decomposes a volumetric object into a graph representation to determine topological equivalence to a sphere. The algorithm iteratively edits regions where small topological handles or holes exist, ensuring the final surface is topologically equivalent to a sphere. The tool has been validated on both phantom and real data, showing improved performance compared to existing methods. BrainSuite has been implemented as a stand-alone interactive tool for use on Windows NT/2000 platforms. It can produce cortical volumes within minutes on common desktop hardware. The tool includes visualization, smoothing, and inflation functions for cortical surfaces and works with the MEG/EEG Matlab toolbox produced by Baillet et al. Validation results show that BrainSuite outperforms existing methods in tissue classification and topological constraint. The tool has been tested on 20 normal MR brain data sets from the Internet Brain Segmentation Repository, showing improved performance in Jaccard similarity metrics. The tool also processes 18 of the 20 IBSR brain volumes to produce inner-cerebral masks, with the topological constraint algorithm successfully producing objects with spherical surface topology. The tool has been shown to achieve high accuracy with minimal changes to the initial set membership. The processing time for the tool is relatively short, with cortical volumes identified from MR images in less than 30 minutes of total operator time.BrainSuite is an automated tool for generating cortical surface representations from T1-weighted MR images of the human brain. It performs a sequence of low-level operations, including skull and scalp removal, image nonuniformity compensation, voxel-based tissue classification, topological correction, rendering, and editing. The tool requires minimal user interaction and produces accurate brain segmentations in clinical time. The paper describes the theory behind each stage of the cortical surface identification process and presents validation results using real and phantom data. The tool addresses challenges in extracting cortical surfaces from MR images, such as measurement noise, partial volume effects, and image nonuniformity. It uses a parametric model to estimate local variations in the gain field of the image by comparing the intensity properties of the whole MR volume with local neighborhoods. This model incorporates spatially variant bias terms and uses a mixture model to account for partial volume effects. The tool also uses a Gibbs' prior to incorporate a spatial model for brain tissues into the classification scheme. The tool includes a Topological Constraint Algorithm (TCA) that decomposes a volumetric object into a graph representation to determine topological equivalence to a sphere. The algorithm iteratively edits regions where small topological handles or holes exist, ensuring the final surface is topologically equivalent to a sphere. The tool has been validated on both phantom and real data, showing improved performance compared to existing methods. BrainSuite has been implemented as a stand-alone interactive tool for use on Windows NT/2000 platforms. It can produce cortical volumes within minutes on common desktop hardware. The tool includes visualization, smoothing, and inflation functions for cortical surfaces and works with the MEG/EEG Matlab toolbox produced by Baillet et al. Validation results show that BrainSuite outperforms existing methods in tissue classification and topological constraint. The tool has been tested on 20 normal MR brain data sets from the Internet Brain Segmentation Repository, showing improved performance in Jaccard similarity metrics. The tool also processes 18 of the 20 IBSR brain volumes to produce inner-cerebral masks, with the topological constraint algorithm successfully producing objects with spherical surface topology. The tool has been shown to achieve high accuracy with minimal changes to the initial set membership. The processing time for the tool is relatively short, with cortical volumes identified from MR images in less than 30 minutes of total operator time.