Magnetic resonance (MR) imaging is a medical technique which permits the visualization of a variety of tumors, lesions, and abnormalities present within the soft biological tissues of the body. Segmentation of medical image data is the process of assigning anatomically-meaningful labels to each component of the image. This thesis describes the development of a tool for the segmentation of MR images of the head. In particular, the tool is designed for the detection of multiple sclerosis lesions of the brain. The design was based on two objectives: (i) to evaluate the effectiveness of incorporating a priori knowledge of brain anatomy in the classification process, and (ii) to compare the statistical and symbolic approaches to machine learning.  Knowledge of neuroanatomy is represented in the form of a tissue probability model. The model was constructed to provide a priori probabilities of brain tissue distribution per unit voxel in a standardized 3D 'brain space'. Use of the model to detect multiple sclerosis lesions reduces the number of false positive lesions by 50%.  The performance of the statistical minimum distance and Bayesian classifiers was compared to that of a symbolic decision tree learning algorithm. A version of this algorithm for the handling of noisy data was included in the comparative study. Each classifier performed at about the same level of accuracy. The statistical classifiers were the fastest in training, yet were the slowest in recall.