This paper presents a new method called adaptive segmentation for correcting and segmenting MRI images. The method uses knowledge of tissue intensity properties and intensity inhomogeneities to improve segmentation accuracy and visualization of MRI data. It employs the Expectation-Maximization (EM) algorithm to estimate the bias field, which accounts for intensity inhomogeneities in MRI data. The method has been shown to be effective in segmenting brain tissue in over 1000 scans, with results comparable to manual segmentation and better than supervised multi-variate classification. The method models intra- and inter-scan intensity inhomogeneities as a spatially-varying gain field that multiplies the intensity data. A logarithmic transformation allows the artifact to be modeled as an additive bias field. The EM algorithm is used to iteratively estimate the bias field and tissue classes. The method is robust to intensity inhomogeneities and does not require manual intervention or supervision for individual scans. The paper describes the implementation of the method for segmenting brain tissue in various types of MRI images, including axial, coronal, and sagittal sections. The method has been tested on a variety of data, including synthetic images and real-world MRI data. Results show that the method provides accurate segmentation of gray and white matter, with performance comparable to manual segmentation and better than supervised classification. The method has been applied to a large longitudinal study of multiple sclerosis patients, demonstrating its effectiveness in segmenting brain tissue across different patients and equipment upgrades. The method has also been shown to be effective in correcting intensity inhomogeneities in images acquired with surface coils, which are known to have severe intensity inhomogeneities. The paper compares the method to manual and supervised segmentation methods, showing that adaptive segmentation provides more accurate and consistent results. The method has been shown to be robust to variations in patient data, scan parameters, and equipment, making it a fully automatic method for segmenting MRI data. The method has been implemented in both single-channel and two-channel configurations, with the two-channel implementation being particularly effective in handling intensity inhomogeneities in multi-channel MRI data.This paper presents a new method called adaptive segmentation for correcting and segmenting MRI images. The method uses knowledge of tissue intensity properties and intensity inhomogeneities to improve segmentation accuracy and visualization of MRI data. It employs the Expectation-Maximization (EM) algorithm to estimate the bias field, which accounts for intensity inhomogeneities in MRI data. The method has been shown to be effective in segmenting brain tissue in over 1000 scans, with results comparable to manual segmentation and better than supervised multi-variate classification. The method models intra- and inter-scan intensity inhomogeneities as a spatially-varying gain field that multiplies the intensity data. A logarithmic transformation allows the artifact to be modeled as an additive bias field. The EM algorithm is used to iteratively estimate the bias field and tissue classes. The method is robust to intensity inhomogeneities and does not require manual intervention or supervision for individual scans. The paper describes the implementation of the method for segmenting brain tissue in various types of MRI images, including axial, coronal, and sagittal sections. The method has been tested on a variety of data, including synthetic images and real-world MRI data. Results show that the method provides accurate segmentation of gray and white matter, with performance comparable to manual segmentation and better than supervised classification. The method has been applied to a large longitudinal study of multiple sclerosis patients, demonstrating its effectiveness in segmenting brain tissue across different patients and equipment upgrades. The method has also been shown to be effective in correcting intensity inhomogeneities in images acquired with surface coils, which are known to have severe intensity inhomogeneities. The paper compares the method to manual and supervised segmentation methods, showing that adaptive segmentation provides more accurate and consistent results. The method has been shown to be robust to variations in patient data, scan parameters, and equipment, making it a fully automatic method for segmenting MRI data. The method has been implemented in both single-channel and two-channel configurations, with the two-channel implementation being particularly effective in handling intensity inhomogeneities in multi-channel MRI data.