NeuralHMM: A Deep Markov Network for Health Risk Prediction using Electronic Health Records

被引:0
作者
Liu, Yuxi [1 ]
Zhang, Zhenhao [2 ]
Qin, Shaowen [1 ]
机构
[1] Flinders Univ S Australia, Coll Sci & Engn, Adelaide, SA, Australia
[2] Northwest A&F Univ, Coll Life Sci, Yangling, Shaanxi, Peoples R China
来源
2023 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS, IJCNN | 2023年
关键词
electronic health records; health risk prediction; patient representation learning; deep Markov model; model transparency;
D O I
10.1109/IJCNN54540.2023.10191594
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Health risk refers to the probability of the occurrence of a specific health outcome for a specific patient. Interest in health risk prediction has been increasing, especially with the availability of a large amount of electronic health records (EHR). An EHR contains multivariate time series data that records meaningful information associated with a chronological set of clinical events for each patient. Recurrent neural networks (RNN) and hidden Markov models (HMM) have been widely used as generative models of time series data. RNN-based models have strong prediction performance but lack transparency. HMMs have a simple functional form and the ability to provide an intuitive probabilistic interpretation, but their state dynamics are 'memoryless', making it difficult to thoroughly take into account the irregularity in patients' health trajectory. This paper proposes a novel deep Markov network for health risk prediction. The method integrates two modules, a GRU (Gated Recurrent Unit) with attention mechanism and a Neural HMM, into a single network. The GRU generates the inputs required for health risk predictions and uses an attention mechanism to create memorable state dynamics for the Neural HMM. The Neural HMM then provides interpretable structured representations through training. Mixture Density Networks are incorporated in the Neural HMM, which contribute to the modeling of complex patterns found in the transition process. Furthermore, an inference network is designed to embed hidden state representations of GRU and Neural HMM into the same space. The inference network enables the two types of representations to learn from each other during the decoding process of Neural HMM, thereby improving the quality of interpretable structured representations. Experimental results on MIMIC-III and eICU datasets demonstrate that our method can outperform state-of-the-art methods and provide transparency of the model decisions.
引用
收藏
页数:8
相关论文
共 28 条
[1]  
Alaa Ahmed M, 2019, ADV NEURAL INFORM PR, V32
[2]  
[Anonymous], 2015, IEEE I CONF COMP VIS, DOI DOI 10.1109/ICCV.2015.123
[3]  
[Anonymous], 2017, Ph.D. thesis
[4]  
Bishop C. M., 1994, MIXTURE DENSITY NETW
[5]   Recurrent Neural Networks for Multivariate Time Series with Missing Values [J].
Che, Zhengping ;
Purushotham, Sanjay ;
Cho, Kyunghyun ;
Sontag, David ;
Liu, Yan .
SCIENTIFIC REPORTS, 2018, 8
[6]  
Cho K., 2014, P 2014 C EMP METH NA, P1724, DOI 10.3115/v1/d14-1179
[7]  
Dempsey WH, 2017, PR MACH LEARN RES, V70
[8]  
Glorot X., 2010, P 13 INT C ART INT S, P249, DOI DOI 10.1109/LGRS.2016.2565705
[9]  
Graves A, 2012, STUD COMPUT INTELL, V385, P1, DOI [10.1162/neco.1997.9.1.1, 10.1007/978-3-642-24797-2]
[10]   Multitask learning and benchmarking with clinical time series data [J].
Harutyunyan, Hrayr ;
Khachatrian, Hrant ;
Kale, David C. ;
Ver Steeg, Greg ;
Galstyan, Aram .
SCIENTIFIC DATA, 2019, 6 (1)