In this study, the 2D pulse coupled neural network (PCNN) model is extended to the 3D space, and a new rule for optimal image segmentation, named product mutual information (PMI), is proposed. Based on the 3D PCNN and PMI, an automatic segmentation algorithm is developed for 3D medical image segmentation. Three-dimensional CT lung images are segmented with the proposed method, showing reduced execution time and improved computation efficiency with high segmentation accuracy. The method is potentially useful for medical image segmentation.