Abstract: Ultrasound has been widely used for diagnosis in clinical practice.However,as the contrast of ultrasound image is usually low,the abnormal tissue is hard to be discriminated from the normal anatomic structures.Hence,in this paper,a novel ultrasound simulation method based on CT image is proposed for the training purpose of the ultrasound image recognition.First,a multi-scale method is developed to enhance vascular tree structures to simulate the Doppler phenomenon of ultrasound;Second,reflection parameter of a single transducer element with speckle noise is calculated by the weighted integration of adjacent regions on the ultrasound propagation path in the CT images.Third,a discrete window function is employed to produce integration and fusion effects of multiple transducer elements.Experimental results show that realistic ultrasounds can be generated at any imaging angle from CT image.And the simulation speed reaches 20fps,which can be used for ultrasound training procedures in clinical practices.