Unsupervised pattern discovery in electronic health care data using probabilistic clustering models

By Benjamin M. Marlin, David C. Kale, Robinder G. Khemani, Randall C. Wetzel. 2012

186 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

In Signal Processing

Citations in the corpus

Natural Language Processing for EHR-Based Computational Phenotyping (2018, methodology): “[44], a diagonal covariance Gaussian mixture”
Learning to Diagnose with LSTM Recurrent Neural Networks (2015, methodology): “They previously have been shown to be effective for these data (Marlin et al., 2012; Che et al., 2015).”
Mining Electronic Health Records: A Survey (2017, methodology): “Marlin et al. [Marlin et al. 2012] developed a probabilistic clustering method to mitigate the effects of unevenly spaced data, which is inherent in EHRs.”
Why Is My Classifier Discriminatory? (2018, methodology): “Following prior work on enrichment of topics in clinical notes (Marlin et al., 2012; Ghassemi et al., 2014), we estimate the probability of patient mortality Y given a topic c as p̂(Y |C = c) := ( ∑n i=1 yiqic)/( ∑n i=1 qic) where yi is the hospital mortality of patient i.”
Patient Subtyping via Time-Aware LSTM Networks (2017, methodology): “is challenge motivates the development of computational methods for biomedical informatics research [5, 17, 24, 28, 31].”
Unfolding Physiological State: Mortality Modelling in Intensive Care Units (2014, methodology): “We used an enrichment measure defined by Marlin et al [12], where the probability of mortality for each topic is calculated as θk = ∑N n=1 qn,k∗yk∑N n=1 qn,k , where yn is the noted mortality out- come (0 for a patient that lives, and 1 for a patient that dies).”
Sparse Multi-Output Gaussian Processes for Medical Time Series Prediction (2017, methodology): “In an alternative approach, noisy and sparse time series data were smoothed temporally by putting Gaussian priors on the mean parameters of the Gaussian mixture model, which is related to a Gaussian process prior, although the distribution is over a finite-dimensional vector (Marlin et al., 2012).”
Deep Computational Phenotyping (2015, background): “Although [22] do not use deep learning, they share similar goals: discovering meaningful physiologic patterns (or physiomes) in multivariate clinical time series.”
Multitask learning and benchmarking with clinical time series data (2017, methodology): “To evaluate prediction accuracy for this problem formulation, we use Cohen’s linear weighted kappa [7, 15], which measures correlation between ordered items.”
Machine Learning and Decision Support in Critical Care (2016, background): “Additionally, generative approaches to unsupervised feature learning [169], [170] may be hamstrung by the shortcomings of approximate inference, or the underlying models may be underspecified with respect to the nuanced features associated with the outcomes of interest.”
Directly Modeling Missing Data in Sequences with RNNs: Improved Classification of Clinical Time Series (2016, methodology): “…strategy for dealing with sparse, variable sampling is to re-sample the data to discrete time series, but the resulting sequences often contain missing values (Marlin et al., 2012) Most often, researchers ﬁll missing values using a heuristic or unsupervised imputation strategy (Lasko et al., 2013).”
Phenotyping of Clinical Time Series with LSTM Recurrent Neural Networks (2015, methodology): “Clinical time series data, as recorded in the pediatric intensive care unit (PICU), exhibit these properties and others, including irregular sampling and non-random missing values [11].”
Modeling Missing Data in Clinical Time Series with RNNs (2016, methodology): “Problematically, the resulting sequences often contain missing values (Marlin et al., 2012).”
Data Mining for Wearable Sensors in Health Monitoring Systems: A Review of Recent Trends and Challenges (2013, methodology): “SpO2 [31,33,40,48, 93] [44,53,60,61] [31,40,48,60, 61] [33,44,53] [31,33,40,48,53, 61] [44,60,86,93]”
Interpretable Deep Models for ICU Outcome Prediction (2016, background): “This has resulted in development of novel and sophisticated machine learning solutions 3,4,5,6,7,8 .”