Existing neurosurgical robots are prone to interfering with each other in a narrow operating space due to the overlapping of several arms, which poses a great safety risk during surgery. In this paper, a 7-DOF dual-arm robot is designed to cooperate with the surgeon to complete craniotomy by manipulating the master hand mechanism. Aiming at the pain point that two real-time dynamic slave hands are prone to interference, this paper establishes a simplified model of the robot envelope, and transforms the interference detection problem between the robots into the interference checking of geometric primitives. If the interference between the two arms is detected by the algorithm, the surgical path of the current robotic arm is immediately terminated to ensure the safety of the surgery. Finally, the algorithm is simulated and verified with the robot designed in this paper. The results show that the algorithm is feasible and has high computational efficiency.