Generating Multi-label Discrete Electronic Health Records using Generative Adversarial Networks

By Edward Choi, Siddharth Biswal, Bradley A. Malin, Jon D. Duke, Walter F. Stewart, Jimeng Sun. 2017

78 citations recorded by the source corpus.

Subfields: Generative Models, Evaluation & Benchmarks

How researchers see this paper

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

In Generative Models

In Evaluation & Benchmarks

Citations in the corpus

Privacy-Preserving Generative Deep Neural Networks Support Clinical Data Sharing (2017, background): “A pair of recent preprints have reported participant generation via neural networks (10, 11).”
Data Synthesis based on Generative Adversarial Networks (2018, neutral): “Two recent studies applied GANs based on recurrent neural network (RNN) architectures to synthesize only discrete values in electronic health records (EHR) [15] and patent records [16].”
Opportunities and obstacles for deep learning in biology and medicine (2017, methodology): “propose generative adversarial neural networks as a tool to make sharable EHR data [117]; however, the authors did not take steps to protect the model from such attacks.”
LOGAN: Membership Inference Attacks Against Generative Models (2017, background): “es, discovering that a speciﬁc record was used for training leaks information about the individual’s health. (Note that image synthesis is commonly used to create datasets for healthcare applications [44,13].) Similarly, if images from a database of criminals are used to train a face generation algo”
Real-valued (Medical) Time Series Generation with Recurrent Conditional GANs (2017, background): “Notably, [7] designed a GAN to generate synthetic electronic health record (EHR) datasets.”
LOGAN: Evaluating Privacy Leakage of Generative Models Using Generative Adversarial Networks (2017, methodology): “Note that image synthesis is already a common technique for creating large datasets in the field of healthcare [29, 8, 9].”
NetGAN: Generating Graphs via Random Walks (2018, background): “Focusing on non-sequential discrete data, Choi et al. (2017) generate high-dimensional discrete features (e.g. binary indicators, counts) in patient records.”
Synthea: An approach, method, and software mechanism for generating synthetic patients and the synthetic electronic health care record (2017, methodology): “Choi et al. 27 proposed the medical Generative Adversarial Network (medGAN) to generate realistic synthetic EHRs.”
Protecting Sensory Data against Sensitive Inferences (2018, background): “For example, to protect health records synthetic medical datasets can be published instead of the real ones using generative models training on sensitive real-world medical datasets [3, 6].”
Generating Synthetic but Plausible Healthcare Record Datasets (2018, unknown): “We will compare our proposed approach with that of MedGan [3], which are able to generate the same kind of data relying on a traditional GAN enhanced with an auto-encoder allowing the system to generate binary data. method provided promising results in the original paper but, unlike the method we…”
PATE-GAN: Generating Synthetic Data with Differential Privacy Guarantees (2018, methodology): “Furthermore, [9] also does not provide any differential privacy guarantees and instead uses ad-hoc notions of privacy which are only validated empirically.”
The History Began from AlexNet: A Comprehensive Survey on Deep Learning Approaches (2018, methodology): “GANs are applied for synthetic EHR data which could mitigate risk [219].”
Generative adversarial networks: introduction and outlook (2017, background): “Choi et al. [54] propose medical generative adversarial network (medGAN) to generate realistic synthetic electronic health records (EHRs).”
Generative Adversarial Networks for Electronic Health Records: A Framework for Exploring and Evaluating Methods for Predicting Drug-Induced Laboratory Test Trajectories (2017, methodology): “We used the mini-batch averaging method proposed by Choi et al. [2] with some notable differences: the model was performing best when the autoencoder’s encoder layer had the same size as the input, therefore not compressing but simply identifying meaningful features.”
Correlated discrete data generation using adversarial training (2018, methodology): “[13] optimizing only over input (D(x)), where D(·) is discriminator output, x is a skill vector, and y is a profession vector.”
Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review (2018, methodology): “In 34 , GAN was used to generate static patient records of discrete events such as diagnosis counts.”
Differentially Private Generative Adversarial Network (2018, background): “…Berthelot et al. [4] also use an autoencoder as a discriminator et al. [3] used differentially private version of Auxiliary Classifier Choi et al. [8] proposed medGAN, which is Differential Privacy. include Dwork et al. [11] for sensitivity-based algorithm, which is et al. [6, 7] proposed DP…”