Dataset Information

Learning Adaptive Forecasting Models from Irregularly Sampled Multivariate Clinical Data.

ABSTRACT: Building accurate predictive models of clinical multivariate time series is crucial for understanding of the patient condition, the dynamics of a disease, and clinical decision making. A challenging aspect of this process is that the model should be flexible and adaptive to reflect well patient-specific temporal behaviors and this also in the case when the available patient-specific data are sparse and short span. To address this problem we propose and develop an adaptive two-stage forecasting approach for modeling multivariate, irregularly sampled clinical time series of varying lengths. The proposed model (1) learns the population trend from a collection of time series for past patients; (2) captures individual-specific short-term multivariate variability; and (3) adapts by automatically adjusting its predictions based on new observations. The proposed forecasting model is evaluated on a real-world clinical time series dataset. The results demonstrate the benefits of our approach on the prediction tasks for multivariate, irregularly sampled clinical time series, and show that it can outperform both the population based and patient-specific time series prediction models in terms of prediction accuracy.

SUBMITTER: Liu Z

PROVIDER: S-EPMC4980099 | biostudies-literature | 2016 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Learning Adaptive Forecasting Models from Irregularly Sampled Multivariate Clinical Data.

Liu Zitao Z Hauskrecht Milos M

Proceedings of the ... AAAI Conference on Artificial Intelligence. AAAI Conference on Artificial Intelligence 20160201

Building accurate predictive models of clinical multivariate time series is crucial for understanding of the patient condition, the dynamics of a disease, and clinical decision making. A challenging aspect of this process is that the model should be flexible and adaptive to reflect well patient-specific temporal behaviors and this also in the case when the available patient-specific data are sparse and short span. To address this problem we propose and develop an adaptive two-stage forecasting a ...[more]

PMID: 27525189

Similar Datasets

Project description:BackgroundDengue fever (DF) represents a significant health burden in Vietnam, which is forecast to worsen under climate change. The development of an early-warning system for DF has been selected as a prioritised health adaptation measure to climate change in Vietnam.ObjectiveThis study aimed to develop an accurate DF prediction model in Vietnam using a wide range of meteorological factors as inputs to inform public health responses for outbreak prevention in the context of future climate change.MethodsConvolutional neural network (CNN), Transformer, long short-term memory (LSTM), and attention-enhanced LSTM (LSTM-ATT) models were compared with traditional machine learning models on weather-based DF forecasting. Models were developed using lagged DF incidence and meteorological variables (measures of temperature, humidity, rainfall, evaporation, and sunshine hours) as inputs for 20 provinces throughout Vietnam. Data from 1997-2013 were used to train models, which were then evaluated using data from 2014-2016 by Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).Results and discussionLSTM-ATT displayed the highest performance, scoring average places of 1.60 for RMSE-based ranking and 1.95 for MAE-based ranking. Notably, it was able to forecast DF incidence better than LSTM in 13 or 14 out of 20 provinces for MAE or RMSE, respectively. Moreover, LSTM-ATT was able to accurately predict DF incidence and outbreak months up to 3 months ahead, though performance dropped slightly compared to short-term forecasts. To the best of our knowledge, this is the first time deep learning methods have been employed for the prediction of both long- and short-term DF incidence and outbreaks in Vietnam using unique, rich meteorological features.ConclusionThis study demonstrates the usefulness of deep learning models for meteorological factor-based DF forecasting. LSTM-ATT should be further explored for mitigation strategies against DF and other climate-sensitive diseases in the coming years.

Project description:MotivationIdentification of system-wide causal relationships can contribute to our understanding of long-distance, intercellular signalling in biological organisms. Dynamic transcriptome analysis holds great potential to uncover coordinated biological processes between organs. However, many existing dynamic transcriptome studies are characterized by sparse and often unevenly spaced time points that make the identification of causal relationships across organs analytically challenging. Application of existing statistical models, designed for regular time series with abundant time points, to sparse data may fail to reveal biologically significant, causal relationships. With increasing research interest in biological time series data, there is a need for new statistical methods that are able to determine causality within and between time series data sets. Here, a statistical framework was developed to identify (Granger) causal gene-gene relationships of unevenly spaced, multivariate time series data from two different tissues of Arabidopsis thaliana in response to a nitrogen signal.ResultsThis work delivers a statistical approach for modelling irregularly sampled bivariate signals which embeds functions from the domain of engineering that allow to adapt the model's dependence structure to the specific sampling time. Using maximum-likelihood to estimate the parameters of this model for each bivariate time series, it is then possible to use bootstrap procedures for small samples (or asymptotics for large samples) in order to test for Granger-Causality. When applied to the A.thaliana data, the proposed approach produced 3078 significant interactions, in which 2012 interactions have root causal genes and 1066 interactions have shoot causal genes. Many of the predicted causal and target genes are known players in local and long-distance nitrogen signalling, including genes encoding transcription factors, hormones and signalling peptides. Of the 1007 total causal genes (either organ), 384 are either known or predicted mobile transcripts, suggesting that the identified causal genes may be directly involved in long-distance nitrogen signalling through intercellular interactions. The model predictions and subsequent network analysis identified nitrogen-responsive genes that can be further tested for their specific roles in long-distance nitrogen signalling.Availability and implementationThe method was developed with the R statistical software and is made available through the R package 'irg' hosted on the GitHub repository https://github.com/SMAC-Group/irg where also a running example vignette can be found (https://smac-group.github.io/irg/articles/vignette.html). A few signals from the original data set are made available in the package as an example to apply the method and the complete A.thaliana data can be found at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE97500.Supplementary informationSupplementary data are available at Bioinformatics online.

Dataset Information

Learning Adaptive Forecasting Models from Irregularly Sampled Multivariate Clinical Data.

Publications

Learning Adaptive Forecasting Models from Irregularly Sampled Multivariate Clinical Data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets