The sense and access optimization problem in opportunistic spectrum access technology is considered. Based on the belief Markov decision process model, which is equivalent to the original partially observable Markov decision process, the performance differences between two different policies are investigated from a sensitivity-based view with the help of the performance potential. Then the policy iteration algorithm is designed. By analyzing the sample path of the system, the sample-path based policy iteration algorithm is developed. Two examples are provided to illustrate the effectiveness of the algorithm.