Semi-Supervised Learning of the Electronic Health Record for Phenotype Stratification

By Brett K. Beaulieu-Jones, Casey S. Greene. 2016

171 citations recorded by the source corpus.

Patient interactions with health care providers result in entries to electronic health records (EHRs). EHRs were built for clinical and billing purposes but contain many data points about an individual. Mining these records provides opportunities to extract electronic phenotypes, which can be paired with genetic data to identify genes underlying common human diseases. This task remains challenging: high quality phenotyping is costly and requires physician review; many fields in the records are sparsely filled; and our definitions of diseases are continuing to improve over time. Here we develop and evaluate a semi-supervised learning method for EHR phenotype extraction using denoising autoencoders for phenotype stratification. By combining denoising autoencoders with random forests we find classification improvements across multiple simulation models and improved survival prediction in ALS clinical trial data. This is particularly evident in cases where only a small number of patients have high quality phenotypes, a common scenario in EHR-based research. Denoising autoencoders perform dimensionality reduction enabling visualization and clustering for the discovery of new subtypes of disease. This method represents a promising approach to clarify disease subtypes and improve genotype-phenotype association studies that leverage EHRs. GRAPHICAL ABSTRACT HIGHLIGHTS Denoising autoencoders (DAs) can model electronic health records. Semi-supervised learning with DAs improves ALS patient survival predictions. DAs improve patient cluster visualization through dimensionality reduction.

Subfields: Artificial Intelligence, Genetics, Molecular Biology

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 Genetics

In Molecular Biology

Citations in the corpus

Opportunities and obstacles for deep learning in biology and medicine (2017, methodology): “For example, the unsupervised and semi-supervised methods that we have discussed reduce the need for labeled examples [81].”
Natural Language Processing for EHR-Based Computational Phenotyping (2018, background): “Furthermore, semi-supervised methods have also been investigated for computational phenotyping [136], [137], which have the potential to significantly reduce the amount of labeling work and simultaneously retain high accuracy.”
Machine Learning for Synthetic Data Generation: a Review (2023, methodology): “using graphical models as the summarization/representation of a dataset [13], [155], [158], [171]–[173], [174] proposes to use a set of large number of low-degree marginals to represent a dataset.”
Reproducibility of computational workflows is automated using continuous analysis (2017, methodology): “For example, we recently used Jupyter with continuous analysis 22 A n A ly s i s When authors are ready to publish their work, they should archive their repositories, which contain the automatically generated results alongside the analytical source code and scripted commands for data retrieval.”
Artificial Intelligence for Mental Health and Mental Illnesses: An Overview (2019, background): “There are other ML methods like semi-supervised learning (blend of supervised and unsupervised) [39, 40] and reinforcement learning where the algo-rithm acts as an agent in an interactive environment that learns by trial and error using rewards from its own actions and experiences [41].”
Deep EHR: A Survey of Recent Advances in Deep Learning Techniques for Electronic Health Record (EHR) Analysis (2017, unknown): “[47-49], [57], [60], [61] [15], [16], [21],”
Boosting Deep Learning Risk Prediction with Generative Adversarial Networks for Electronic Health Records (2017, methodology): “learning method on structured quantitative EHR data [22].”
PrivLogit: Efficient Privacy-preserving Logistic Regression by Tailoring Numerical Optimizers (2016, methodology): “We consider extending this novel approach to other machine learning models, such as other classiﬁers [Zheng et al., 2017], regressors, and deep learning [Beaulieu-Jones and Greene, 2016] that are increasingly popular in privacy-sensitive domains.”
Artificial intelligence and machine learning in precision and genomic medicine (2021, methodology): “It is essential to maximize the usage of EHRs by incorporating different data-sets and identifying particular patterns of patients' disease progression to deliver high decision support and apply personalized and population health effects, which has a greater possibility to enhance positive clinical ”