Abstract: Magnetic Resonance (MR) is a relatively safe imaging technique that provides excellent soft tissue contrast and is widely used in neuroimaging research. Intracranial segmentation, commonly referred to as skull-stripping, aims to segment the brain tissue (cerebrum and cerebellum) from the skull and other non-brain tissues in MR images of the brain. A new mathematical algorithm, a model-based level set (MLS), was developed to control the evolution of the level set function in both 2D and 3D domain. The evolution was controlled by a new velocity function that includes two force terms to determine the speed of the level set function. The first force was derived from the mean curvature of the level set function, and the second was designed to model the intensity characteristics of the cortex in MR images. A narrow band around to zero level set was applied to narrow the propagating domain. A reinitialization procedure was applied to rectify the distance mapping of the level set function. Special procedures for preventing leakage in both 2D and 3D MLS algorithms were implemented. Anatomical similarity between adjacent slices was used in the 2D MLS to reduce computation time, while a multi-scale procedure was implemented in the 3D MLS to reduce computational and storage costs. Quantitative evaluation of the 2D and 3D MLS algorithms was performed by comparing the results of the MLS algorithms to those obtained using expert segmentation in 29 sets of pediatric brain MR images and 20 sets of young adult MR images. Another 52 sets of elderly and adult brain MR images were used for evaluating the algorithm. Each of the MLS algorithms was compared to two existing methods, the brain extraction tool (BET) and the brain surface extractor (BSE). The 2D MLS and the 3D MLS algorithms were compared to each other to determine the advantages and disadvantages of each method. Finally, a visualization environment was implemented to integrate the automated algorithms with semi-automated editing functions that are easier-to-use than existing manual editing tools. The automated algorithms, and the integrated segmentation and visualization environment, may accelerate data analysis in large, multi-institutional, population-based neuroimaging studies.