Abstract: The existing medical image fusion methods suller from issues such as energy loss and unclealtextures. To address these problems , this paper proposes a method that separately fuses the high and low.frequeney coefficients in the NSS'T domain using an improved pulse coupled neural network ( PCNN) andnuclear energy. First, multimodal medical images undergo NSST decomposition to obtain the high andlow-frequeney coelficients of the source images. Then, the improved PCNN is employed to fuse the de-composed high-frequeney coelficients, while nuclear energy is used to fuse the low-frequeney coelficients.In the improved PCNN , a Gaussian filter is used to combine surrounding neurons, and the link strength isadaptively adjusted, greatly reducing the need for manually setting parameters. Finally , the fused imagesis obtained through NSST inverse transformation. Six sets of images are randomly selected for fusion andcomparative experiments. The results show that this method not only achieves superior visual performancebut also performs excellently across seven objective evaluation metrics.

Key words: medical image fusion, NSST, improved PCNN, Gaussian filter, nuclear normenergy