This paper presents a method to detect the collision between the needle model and voxel-based human tissue model for simulation of needle insertion. Four planes are used in front of the needle to protect penetration into bone in advance. Each plane is defined by using four points to consider thickness of the needle. Hounsfield unit (HU) values which represent the type of human tissue are collected at each point to identify colliding voxels. The size of voxel and needle, and the maximum velocity of the needle is identified and used to determine the relative position, and number of planes and points. The proposed method improves the performance of collision detection while maintaining the update rate at 1 kHz.