A Bioinspired Flexible Memristor-Based Near-Sensor Neuromorphic Computing System for Intelligent Monitoring of Wound Infection Severity

Rapid and accurate assessment of postoperative wound infection is critical for timely adjustment of clinical treatment strategies, which requires real-time tracking of infection progression without interfering with the wound-healing process. To address this need, we developed a flexible threshold memristor based on a carbon nanotube-TiO₂ heterostructure, enabling quantitative differentiation among non-infected, mildly infected, moderately infected, and severely infected wound states through the systematic variation of its threshold voltage. The operating mechanism originates from the efficient adsorption of bacteria by CNTs and the acid-promoted dissolution of the Ag electrode, which together accelerate the directional migration of Ag⁺ ions and the formation of conductive filaments within the TiO₂ switching layer. Furthermore, by integrating the memristor with a leaky integrate-and-fire (LIF) neuron circuit, we constructed a bioinspired “infection-sensing neuron” capable of directly converting infection-induced analog signals into intuitive spike-frequency patterns. This design not only enables on-device processing and compression of biological signals, exhibiting the characteristics of near-sensor computing and significantly reducing data transmission and processing burdens, but also successfully distinguishes different infection levels. Therefore, this study provides a new strategy for developing noninvasive infection-monitoring systems with clinical potential and promotes the application of flexible neuromorphic electronics in wearable intelligent healthcare.