Research On Overall Survival Time Prediction Algorithm Of Brain Tumor Patients Based On Deep Hypergraph Membrane System

Membrane computing is an important branch of biological computing.The computational model of membrane computing is called membrane system,which has the characteristics of distributed and maximum parallelism.The spiking neural P system is a membrane computing model inspired by the biological phenomenon of information interaction between neurons using spikes in biological neural networks.It is a hot topic in the field of membrane computing research.Inspired by different biological phenomena,different types of spiking neural P systems have been proposed to further improve the computational power of the model.The spiking neural P system and its variants have broad application prospects,showing strong parallelism and robustness in the fields of medical image processing,power system fault diagnosis,and identification of nuclear output signals.However,the spiking neural P systems and its variants can be regarded as a two-dimensional graph structure,in which neurons can only transmit information to the neurons connected to them,and cannot perform hierarchical calculation and non-result storage,nor can they handle high-order relationships.These problems greatly limit the learning ability of spiking neural P systems in solving various complex practical problems.Hypergraph is a flexible topology that can describe the nonlinear high-order relationship between data samples.Therefore,this paper applies the hypergraph to the hierarchical nested structure of the membrane and combine it with the spiking neural P system to better expand the practical application of membrane computing.Brain tumor is a common central nervous system disease,which seriously endangers human health.It is necessary to accurately predict the preoperative overall survival(OS)time of patients with brain tumors,which will provide information for patients to develop personalized treatment plans and promote common decision-making between doctors and patients.Recently,many deep learning methods based on magnetic resonance imaging(MRI)have been used to predict OS time in patients with brain tumors.There are few studies on OS time prediction based on histopathological images.However,histopathological images reveal disease progression and related molecular processes,and contain rich information about tumor morphological phenotypes.Therefore,it is particularly important to develop a method for accurately predicting preoperative OS time in patients with brain tumors using histopathological images.Based on the spiking neural P system and the hypergraph theory,this paper designs two new spiking neural P systems.On this basis,two deep hypergraph membrane system algorithms are proposed for the OS time prediction of brain tumor patients.The details are as follows:This paper proposes hypergraph-based spiking neural P systems(HSNPS)containing three new classes of neurons to describe higher-order relationships among neurons.Three new kinds of rules among neurons are also designed to expand the model into planar,hierarchical and transmembrane computations.Combining the hypergraph spiking neural P system and deep learning model,this paper develops a deep hypergraph membrane algorithm for predicting OS time of brain tumor patients-deep hypergraph membrane OS time prediction algorithm based on HSNPS(HSNPS-DPOS).The proposed algorithm is evaluated on glioblastoma(GBM)cohorts from The Cancer Genome Atlas(TCGA).The HSNPS-DPOS algorithm achieves good performance compared to the six state-of-the-art methods,thereby verifying the effectiveness of the algorithm in predicting the OS time of brain tumor patients.In addition,ablation experiments prove the effectiveness of the proposed algorithm flow and HSNPS.Then,the attention hypergraph numerical neural membrane system is proposed on the basis of hypergraph-based spiking neural P systems.The system introduces the numerical neural P system to better solve the numerical problem.Attention neurons are also designed to introduce attention mechanisms to focus on more important information.Combining the attention-enabled hypergraph numerical neural P system and deep learning model,this paper develops a deep hypergraph membrane algorithm for predicting the overall survival time of brain tumor patients-deep hypergraph membrane OS time prediction algorithm based on AHNNPS(AHNNPS-DPOS).The algorithm is evaluated on the histopathological images of the GBM cohort in The Cancer Genome Atlas(TCGA).The experimental results demonstrate the accuracy and robustness of AHNNPS-DPOS in predicting OS time in brain tumor patients.In addition,ablation experiments prove the effectiveness of the proposed algorithm flow and attention mechanism.