Enhancing sepsis detection with nature-inspired optimization integrated neural network

Abstract

Sepsis, a disorder that can be fatal that is brought on by a dysregulated immunological reaction to an infection, keeps presenting complex problems for early detection and treatment. In this article, they suggest a novel (MBSO-LSTM) method for improving sepsis diagnosis that combines long short-term memory (LSTM) networks and modified bird swarm optimization (MBSO). The biases and weights of the network known as LSTM are optimized using the MBSO algorithm, which was developed in response to research on the social behavior of bird flocks. MBSO enables faster and more successful learning of the LSTM model, resulting in increased sepsis detection performance by emulating the self-organizing behavior of bird swarms. Our work uses a sizable collection of medical records from individuals who have probable sepsis, including demographic information, laboratory findings, and vital signs. The LSTM network in the suggested framework recognizes the time-dependent relationships and patterns present in the data by using these characteristics as inputs. They contrast our method's effectiveness with other cutting-edge sepsis detection strategies, such as conventional approaches. The outcomes show that concerning sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC-ROC), our proposed system performs better than alternative methods.