This paper describes ~m experimented system for automatic segmentation of lesions in crani~d Computed Tomography (CT) hnages. The system performs nonn~dization to reduce cupping ,-u'titact. converts CT dam to gray level images ,and then extracts the region-of-interest. From this hnage, it locates ,and v,'didates the lesion before segmenting ,and delineating is performed. Fuzzy region, symmetric~d measure and minimum directional-ch~mge technique are used to enhance the accuracy of segmentation. The system is applied on severa! CT brain images. The results ,are encouraging with acceptable accuracy ,and consistency.