Computed Tomography (CT) brain image is an important tool in the area of brain diagnosing. The human eye cannot recognize greater than (32) different gray levels in the image. For this reason there is some probability of error in the diagnosing the type of tumor by radiologists. Automation of CT brain image diagnosing would overcome this problem, and would also assist the radiologist in pcrfoming a more objective and accurate analysis. In this thesis the design and implementation of high level Brain Image Diagnostic System (BIDS) are described. It is implemented in a blackboard environment that holds various types of information and which controls the interpretation process. BIDS consists of four main phases. The first one tries to improve visual quality of the image for both radiologist observation and machine analysis. It uses filters belonging to nonlinear order statistical family and mathematical morphology family. In the second phase, brain image segmentation is performed. The segmentation of brain image into anatomically and pathologically meaningful regions is still lacking. A new method of segmentation based on fuzzy clustering and region growing is implemented. The regions obtained from this phase are meaningful regions according to the anatomy of the brain. A novel approach for region description is used in the third phase of BIDS. The approach is based on combination of several descriptors (shape, topology, texture, histogram, skeletonization, and topography). This combination allows to overcome the drawbacks of one by taking the advantage of the merits of another. In the final phase, knowledge base which Contains the description of brain regions and their relationship is builtIt is used to diagnose different types of normal / abnormal brain regions. The system has been tested with (50) different CT brain images taken From (30) different patients. It demonstrates that it can automatically label (48) different objects from a full set of (5) CT sections. It is likely that the radiologist will benefit from the system in one of two ways. First the radiologist will use the system as a second reader. Second, the radiologist may the system as a consultant on an individual suspect area, the radiologist could point- to the suspected area and could obtain useful information about it from the system.