A numerical model was developed which can use digitized layer interfaces to calculate ultrasound wave absorption, diffraction, reflection, and refraction. This model was used to evaluate the feasibility of ultrasound therapy and surgery through a human skull. A digitized human skull profile was obtained from magnetic resonance (MR) images and used to calculate the ultrasound field in the brain of a volunteer from a spherically curved phased array. With no phase correction, the focus of the array was shifted and defocused. The phased array technique was used to correct focal shift, reduce side lobes, and enhance focal amplitude. The optimum source element width was estimated for each frequency to obtain a near optimium focus, and an appropriate frequency range for transskull ultrasound therapy and surgery was determined. Acoustic pressure amplitude on the skull surfaces was examined, and it was shown that the skull heating problem could be overcome. Despite high attenuation, complex interface shape, and nonuniform thickness of a human skull, a sharply focused transskull ultrasound field can be generated for noninvasive ultrasound therapy and surgery in the brain.