Voxel-Based Morphometry (VBM) is a method for analyzing structural differences in the brain by comparing the local concentration of gray matter between groups of subjects. The process involves spatially normalizing high-resolution images to a common stereotactic space, segmenting gray matter, smoothing the segments, and performing statistical tests to compare the smoothed gray-matter images. Corrections for multiple comparisons are made using Gaussian random field theory. This paper describes the steps involved in VBM, with a focus on segmenting gray matter from MR images with nonuniformity artifacts. It evaluates the assumptions underlying the method, including the accuracy of segmentation and the statistical distribution of data. VBM is used to detect regional differences in gray matter concentration, accounting for global shape differences. Spatial normalization transforms images to a common space, while segmentation and smoothing are used to extract and process gray matter. The logit transform is applied to improve normality of the data before statistical analysis. The general linear model (GLM) is used to identify regions of gray matter concentration significantly related to the effects under study. Statistical analysis involves voxel-wise parametric tests and corrections for multiple comparisons. The paper evaluates the accuracy of the segmentation method, particularly in the presence of nonuniform intensity variations. It compares segmentations with and without nonuniformity correction, showing that correction improves the accuracy of tissue classification. The method is also tested for its robustness to misregistration with prior probability images and its ability to handle nonstationary smoothness in the data. The paper discusses the assumptions of normality in the data and evaluates the validity of statistical tests. It shows that the data are not perfectly normally distributed, but the logit transform and modeling of gray matter as a confound improve normality. The paper also tests the rate of false positives using randomization, showing that the method is robust to mild deviations from normality. The paper concludes that VBM is a valid method for detecting regional differences in gray matter concentration, but it highlights the need for further improvements in segmentation and the importance of accounting for nonstationary smoothness in the data. The method is compared to other approaches such as deformation-based and tensor-based morphometry, and it is suggested that future developments may combine VBM with more advanced techniques to improve the accuracy of structural brain analysis.Voxel-Based Morphometry (VBM) is a method for analyzing structural differences in the brain by comparing the local concentration of gray matter between groups of subjects. The process involves spatially normalizing high-resolution images to a common stereotactic space, segmenting gray matter, smoothing the segments, and performing statistical tests to compare the smoothed gray-matter images. Corrections for multiple comparisons are made using Gaussian random field theory. This paper describes the steps involved in VBM, with a focus on segmenting gray matter from MR images with nonuniformity artifacts. It evaluates the assumptions underlying the method, including the accuracy of segmentation and the statistical distribution of data. VBM is used to detect regional differences in gray matter concentration, accounting for global shape differences. Spatial normalization transforms images to a common space, while segmentation and smoothing are used to extract and process gray matter. The logit transform is applied to improve normality of the data before statistical analysis. The general linear model (GLM) is used to identify regions of gray matter concentration significantly related to the effects under study. Statistical analysis involves voxel-wise parametric tests and corrections for multiple comparisons. The paper evaluates the accuracy of the segmentation method, particularly in the presence of nonuniform intensity variations. It compares segmentations with and without nonuniformity correction, showing that correction improves the accuracy of tissue classification. The method is also tested for its robustness to misregistration with prior probability images and its ability to handle nonstationary smoothness in the data. The paper discusses the assumptions of normality in the data and evaluates the validity of statistical tests. It shows that the data are not perfectly normally distributed, but the logit transform and modeling of gray matter as a confound improve normality. The paper also tests the rate of false positives using randomization, showing that the method is robust to mild deviations from normality. The paper concludes that VBM is a valid method for detecting regional differences in gray matter concentration, but it highlights the need for further improvements in segmentation and the importance of accounting for nonstationary smoothness in the data. The method is compared to other approaches such as deformation-based and tensor-based morphometry, and it is suggested that future developments may combine VBM with more advanced techniques to improve the accuracy of structural brain analysis.