Spiking neural P system with synaptic vesicles and applications in multiple brain metastasis segmentation

Spiking neural (SN) P systems, performing operations by a series of spikes and rules, are a kind of membrane computing models and have been applied in various fields. However, in the model, only one set of spikes is computed in a neuron, resulting in too many spikes and neurons being consumed in solving a problem. To solve this issue, we propose SN P systems with synaptic vesicles (SN-SV P system), where each neuron contains multiple synaptic vesicles to deal with different sets of spikes simultaneously. Three kinds of rules are also designed for communications in the new P system. Based on the SN-SV P system, several multifusion attention mechanism-based single-shot deep learning models with different initializations are implemented simultaneously to perform ensemble learning on brain metastasis (BMs) segmentation. The experimental results demonstrate that the SN-SV P system outperforms state-of-the-art methods and performs well on BMs with various phenotypes.