A habituation memristor model for lung cancer screening application

Abstract

Abstract —Lung cancer screening is critical to the diagnosis and treatment of patients. Today, computed tomography (CT) scanning technology provides a promising approach for the screening of lung cancer. Nevertheless, the redundant information in CT images often limits the efficiency and accuracy of screening. The biological sensory nervous system has an important mechanism for screening out redundant information, namely habituation. Here, we designed a second-order memristor model with habituation characteristics. Some of the habituation behavior of the memristor model has been demonstrated with LTspice simulation. Furthermore, the habituation memristor model is incorporated in a volatile memristor based leaky integrate and fire (LIF) neuron circuit to construct a simple neural system. The simulation results indicate that the neural system exhibits reliable habituation behaviors. Finally, lung cancer screening tasks have been implemented based on the neural system with habituation behavior. The habituation memristor circuit serves as a data preprocessing layer that filters out relevant information from lung cancer images. The results indicate that the performance and accuracy of lung cancer screening performance are noticeably better than the neural system without habituation behavior. This work provides a new idea for lung cancer screening implementation.