Sepsis deterioration prediction using channelled long short-term memory networksSvenson, P., Haralabopoulos, G. and Torres Torres, M. (2020) Sepsis deterioration prediction using channelled long short-term memory networks. In: 18th International Conference on Artificial Intelligence in Medicine, AIME 2020, August 25–28, 2020, Minneapolis, MN, USA, pp. 359-370, https://doi.org/10.1007/978-3-030-59137-3_32. Full text not archived in this repository. It is advisable to refer to the publisher's version if you intend to cite from this work. See Guidance on citing. To link to this item DOI: 10.1007/978-3-030-59137-3_32 Abstract/SummarySepsis is a severe medical condition that results in millions of deaths globally each year. In this paper, we propose a Channelled Long-Short Term Memory Network model tasked with predicting 48-hour mortality in sepsis against the Sequential Organ Failure Assessment (SOFA) score. We use the MIMIC-III critical care database. Our research demonstrates the viability of deep learning in predicting patient outcomes in sepsis. When compared with published literature for similar tasks, our channelled LSTM models demonstrated a comparable AUROC with superior precision score. The results showed that deep learning models outperformed the SOFA score in predicting 48-hour mortality in sepsis in AUROC (0.846–0.896 vs 0.696) and average precision score (0.299–0.485 vs 0.110). Finally, our Fully-Channelled LSTM outperforms a baseline LSTM by 5.4% in AUROC and 59.9% in average precision score.
Altmetric Deposit Details University Staff: Request a correction | Centaur Editors: Update this record |
Lists
Lists
