Sepsis Deterioration Prediction Using Channelled Long Short-Term Memory Networks

摘要: Sepsis is a severe medical condition that results in millions of deaths globally each year. In this paper, we propose Channelled Long-Short Term Memory Network model tasked with predicting 48-hour mortality sepsis against the Sequential Organ Failure Assessment (SOFA) score. We use MIMIC-III critical care database. Our research demonstrates viability deep learning patient outcomes sepsis. When compared published literature for similar tasks, our channelled LSTM models demonstrated comparable AUROC superior precision The showed outperformed SOFA score (0.846–0.896 vs 0.696) and average (0.299–0.485 0.110). Finally, Fully-Channelled outperforms baseline by \(5.4\%\) \(59.9\%\)