Performance Comparison of Learning Models to Predict The In-Hospital Mortality
Abstract views: 12
/ 
PDF downloads: 13
Keywords:
Intensive Care Unit, Artificial Neural Networks, Hidden Markov Models, Receiver Operating Characteristic Curve, Multi-Parameter Intelligent Monitoring (MIMIC III)Abstract
The prediction probability of in-hospital mortality who are admitted to the Intensive Care Unit
(ICU) are calculated in this work by applying two common learning models: artificial neural networks
and the hidden Markov model. Both models were applied to the same dataset to ensure a fair comparison.
The clinical data supporting these models were carefully selected from the Multi-Parameter Intelligent
Monitoring (MIMIC III) database, with a particular emphasis on the ICU domain. Thus, accurately the
real-world conditions were reflected. The dataset, comprising 8000 individual records, was divided using
cross-validation techniques. Subsequently, the datasets were utilized as training and test sets for each
learning model. The effectiveness of the models was evaluated using the Area Under the Receiver
Operating Characteristic Curve (AUC-ROC) measure due it aligns well with the fundamental
characteristics of the models. Notably, the single hidden layer neural network model produced a ROC
value of 0.8927, while the multi-hidden layer model generated a significantly lower ROC value of 0.8691.
The hidden Markov model achieved the best result in this study, with a higher ROC value of 0.9038.
Downloads
References
Clifford G.D., Scott D.J., Villarroel M. User guide and documentation for MIMIC III database, Version 1.4, Realised date: 02.09.2016. Available: http://www.physionet.org or http://mimic.physionet.org
Hatamabadi, H. R., Abdalvand, A., Abdollahi, F. Shokoohi, H., Bastani, M., The utility of artificial neural network models in predicting the mortality of patients with traumatic brain injury admitted to intensive care unit. Archives of trauma research 2020; 9(1), 30-35
Chen, Y., Li, L., Du, J. Predicting in-hospital mortality of intensive care unit patients using random forest model. Journal of Medical Systems 2019; 43(8), 239.
Shashikumar, SP., Stanley, MD., Sadiq, I., Li, Q., Holder, A., Clifford, GD., Nemati, S. Early sepsis detection in critical care patients using multiscale blood pressure and heart rate dynamics. Journal of biomedical informatics 2017; 75, 8-15.
Komorowski M, Celi LA, Badawi O, Gordon AC, Faisal AA. The artificial intelligence clinician learns optimal treatment strategies for sepsis in intensive care. Nat Med. 2018;24(6):1716 1720.
M. R. Hassan and B. Nath. Stock market forecasting using hidden markov model: a new approach. In 5th international conference on intelligent systems design and applications (ISDA’05), pages 192–196. IEEE, 2005.
V. De Fonzo, F. Aluffi-Pentini, and V. Parisi. Hidden markov models in bioinformatics. Current Bioinformatics, 2(1):49–61, 2007.
V. Nefian, L. Liang, X. Pi, L. Xiaoxiang, C. Mao, and K. Murphy. A coupled hmm for audio-visual speech recognition. In 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing, volume 2, pages II–2013. IEEE, 2002.
Mirchandani, G., Cao W. On hidden nodes for neural nets. IEEE Transaction on circuits and systems 1989; 36, 5.
