Li-wei H. Lehman

I am a Research Scientist at the Institute for Medical Engineering & Science (IMES), MIT. I am a Principal Investigator and lead the project on "Learning Optimal Dynamic Treatment Strategies from Temporal ICU Monitoring Data." My research focuses on machine intelligence for health, state space modeling of physiological dynamics, dynamic treatment regimes for sequential treatment decision making, or more generally, the use of machine learning to derive insights from physiological and clinical data for informed treatment decision making. My interests include generative latent variable models, switching state-space models, representation learning, Bayesian non-parametric learning of disease phenotypes, off-policy reinforcement learning, and dynamic treatment regimes. I received my Master’s degree in Computer Science from Georgia Institute of Technology, and my Ph.D. from Massachusetts Institute of Technology in 2005.

Research positions available: I am looking for highly-motivated Research Interns (graduate level), Research Fellows, and Post-Docs to join my research team starting immediately. For more details, see here.

News

• August 2024. Excited to present our paper G-Transformer: Counterfactual Outcome Prediction under Dynamic and Time-varying Treatment Regimes at MLHC 2024, Toronto.
• March 2024. Thrilled to share that our paper "Counterfactual Sepsis Outcome Prediction under Dynamic and Time-Varying Treatment Regimes" received the Clinical Research Informatics Distinguished Paper Award at the AMIA 2024 Informatics Summit! It's truly an honor to receive this recognition for our research. Huge thanks to everyone who contributed to making this possible.
• December 2023. Excited to present our paper VTaC: A Benchmark Dataset of Ventricular Tachycardia Alarms from ICU Monitors at NeurIPS 2023, New Orleans.
• December 2023. Congrats to my student Anna Wong for her first abstract in ML4H Symposium on A Knowledge Distillation Approach for Sepsis Outcome Prediction from Multivariate Clinical Time Series.
• May 2023. Happy to announce that our paper "A Diffusion Model with Contrastive Learning for ICU False Arrhythmia Alarm Reduction" has been accepted to IJCAI 2023.
• November 2022. Happy to share our work "Treatment-RSPN: Recurrent Sum-Product Networks for Sequential Treatment Regimes" at the upcoming NeurIPS Time Series for Health workshop and the Machine Learning for Health Symposium, New Orleans, 2022.
• March 2022. Delighted to read this MIT news article about G-Net, a new deep learning approach developed by my team to simulate patient trajectories and outcomes under time-varying and dynamic treatment strategies.
• Dec. 2021. Happy to announce that our paper Knowledge Distillation via Constrained Variational Inference has been accepted to AAAI 2022! (Acceptance rate 15%).
• Nov. 2021. happy to announce that our work G-Net: a Recurrent Network Approach to G-Computation for Counterfactual Prediction Under a Dynamic Treatment Regime has been published in Machine Learning for Health, 2021.
• Nov. 2020, excited that our work "Is Deep Reinforcement Learning Ready for Practical Applications in Healthcare? A Sensitivity Analysis of Duel-DDQN for Sepsis Treatment," won the Distinguished Paper Award at the AMIA Annual Symposium 2020.
• Feb. 2020, presented our paper "Robust Low-Rank Discovery of Data-Driven Partial Differential Equations" at The Thirty-Fourth AAAI Conference on Artificial Intelligence, 2020 .
• Sep. 2019, presented "Learning Optimal Dynamic Treatment Regimes from Temporal ICU Monitoring Data," at the IBM AI Week, MIT.
• Aug. 2019, I accepted the invitation to serve as a Program Committee (PC) member for the AAAI 2020.
• Aug. 2019, our paper "Retaining Privileged Information for Multi-Task Learning" was published in KDD 2019.
• March 28th, 2019. Honored to give an invited talk "Deriving Insights from Physiological and Clinical Data Using Structured Representation Learning" at the SPIRAL seminar, Northeastern University.
• Aug. 2018, our paper "Representation Learning Approaches to Detect False Arrhythmia Alarms from ECG Dynamics," presented at the Machine Learning for Healthcare Conference (MLHC) 2018.
• Jan. 2018, our paper "A Model-Based Machine Learning Approach to Probing Autonomic Regulation from Nonstationary Vital-Sign Time Series" was published in IEEE JBHI.

Selected Publications

Switching State-Space Approaches for Modeling Physiological Dynamics

• "A Physiological Time Series Dynamics-Based Approach to Patient Monitoring and Outcome Prediction", Li-wei H. Lehman, Ryan P. Adams, Louis Mayaud, George B. Moody, Atul Malhotra, Roger G. Mark, Shamim Nemati, IEEE Journal of Biomedical and Health Informatics, 19(3):1068-1076, May 2015. doi:10.1109/JBHI.2014.2330827. Preprint.
• "Bayesian nonparametric learning of switching dynamics in cohort physiological time series: application in critical care patient monitoring", Li-wei H. Lehman, Matthew J. Johnson, Shamim Nemati, Ryan P. Adams, Roger G. Mark, Chapter 11 in Advanced State Space Methods for Neural and Clinical Data, Cambridge University Press, 2015. Publisher's Version.
• "A Model-Based Machine Learning Approach to Probing Autonomic Regulation from Nonstationary Vital-Sign Time Series", Li-wei H. Lehman, Roger G. Mark, Shamim Nemati, IEEE Journal of Biomedical and Health Informatics, Vol. 22, No. 1, January 2018. doi:10.1109/JBHI.2016.2636808. Preprint .
• “Discovering shared dynamics in physiological signals: Application to patient monitoring in ICU,” Lehman LH, Nemati S, Adams RP, Mark R, Proceedings of the 34th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), San Diego, 2012.
• “Tracking Progression of Patient State of Health in Critical Care Using Inferred Shared Dynamics in Physiological Time Series,” Lehman LH, Nemati S, Adams RP, Moody GB, Malhotra A, Mark RG, Proceedings of the 35th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2013.
• “Hemodynamic Monitoring Using Switching Autoregressive Dynamics of Multivariate Vital Sign Time Series,” Li-wei H. Lehman, Shamim Nemati, and Roger G. Mark, Proceedings of the Computing in Cardiology, France Nice, September 2015.
• “Uncovering Clinical Significance of Vital Sign Dynamics in Critical Care,” Li-wei H. Lehman, Shamim Nemati, George Moody, Thomas Heldt, and Roger G. Mark, Proceedings of the Computing in Cardiology, Boston, September 2014.
• “Patient Prognosis from Vital Sign Time Series: Combining Convolutional Neural Networks with a Dynamical Systems Approach,” Li-wei H. Lehman, Mohammad Ghassemi, Jasper Snoek, and Shamim Nemati, Proceedings of the Computing in Cardiology, France Nice, September 2015.
• “Learning Outcome-Discriminative Dynamics in Multivariate Physiological Cohort Time Series,” Nemati S, Lehman LH, Adams RP, Proceedings of the 35th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2013.

Representation Learning from Physiological and Clinical Data

• Knowledge Distillation via Constrained Variational Inference, Ardavan Saeedi, Yuria Utsumi, Li Sun, Kayhan Batmanghelich, Li-wei H. Lehman, Proceedings of the Thirty-Sixth AAAI Conference on Artificial Intelligence, February 2022. (Acceptance rate 15%.)
• A contrastive learning approach for ICU false arrhythmia alarm reduction, Yuerong Zhou, Guoshuai Zhao, Jun Li, Gan Sun, Xueming Qian, Benjamin Moody, Roger G Mark, Li-wei H. Lehman, Nature Scientific Reports, 2022.
• "Retaining Privileged Information for Multi-Task Learning," Fengyi Tang, Cao Xiao, Fei Wang, Jiayu Zhou, Li-wei H. Lehman, Proceedings of the 25th SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), August 2019, Anchorage, Alaska USA. Research Track.
• "Representation Learning Approaches to Detect False Arrhythmia Alarms from ECG Dynamics," Eric P. Lehman, Rahul G. Krishnan, Xiaopeng Zhao, Roger G. Mark, Li-wei H. Lehman, Proceedings of the 3rd Machine Learning for Healthcare Conference , PMLR 85:571-586, 2018.

Machine Learning and Causal Inference for Informed Clinical Sequential Decision Making

• "Counterfactual Sepsis Outcome Prediction under Dynamic and Time-Varying Treatment Regimes" , Megan Su, Stephanie Hu, Hong Xiong, Elias Baedorf Kassis, Li-wei H Lehman, AMIA Informatics Summit, 2024. Distinguished Paper Award.
• G-Net: a Recurrent Network Approach to G-Computation for Counterfactual Prediction Under a Dynamic Treatment Regime, Rui Li, Stephanie Hu, Mingyu Lu, Yuria Utsumi, Prithwish Chakraborty, Daby M. Sow, Piyush Madan, Mohamed Ghalwash, Zach Shahn, Li-wei H. Lehman , Proceedings of Machine Learning for Health, PMLR 158:280-297, 2021. An earlier ArXiv version of this work is available at arXiv:2003.10551.
• Is Deep Reinforcement Learning Ready for Practical Applications in Healthcare? A Sensitivity Analysis of Duel-DDQN for Sepsis Treatment," MingYu Lu, Zachary Shahn, Daby Sow, Finale Doshi-Velez, Li-wei H. Lehman, arXiv:2005.04301, May 2020. Accepted to AMIA Annual Symposium 2020.
• "Fluid-limiting treatment strategies among sepsis patients in the ICU: a retrospective causal analysis ," Zach Shahn, Nathan I. Shapiro, Patrick D. Tyler, Daniel Talmor and Li-wei H. Lehman, Journal of Critical Care, 24:62, 2020.
• "Should Diuretic Initiation be Delayed in ICU Patients with Recent Vasopressor Use? A Causal Analysis," Somnath Bose, Li-wei H. Lehman, Kechun Huang, Daniel Talmor, Zach Shahn, Critical Care Medicine, Volume 48, Issue 1, p 733, January 2020.
• "Evaluating Reinforcement Learning Algorithms in Observational Health Settings," Omer Gottesman, Fredrik Johansson, Joshua Meier, Jack Dent, Donghun Lee, Srivatsan Srinivasan, Linying Zhang, Yi Ding, David Wihl, Xuefeng Peng, Jiayu Yao, Isaac Lage, Christopher Mosch, Li-wei H. Lehman, Matthieu Komorowski, Aldo Faisal, Leo Anthony Celi, David Sontag, Finale Doshi-Velez, https://arxiv.org/abs/1805.12298v1, 2018.
• "Improving Sepsis Treatment Strategies by Combining Deep and Kernel-Based Reinforcement Learning," Xuefeng Peng, Yi Ding, David Wihl, Omer Gottesman, Matthieu Komorowski, Li-wei H. Lehman, Andrew Ross, Aldo Faisal, Finale Doshi-Velez, Proceedings of the AMIA Annual Symposium, 2018.

Deriving Insights from Clinical Text: Bayesian Non-parametric Learning of Patient Phenotypes

• “Risk stratification of ICU patients using topic models inferred from unstructured progress notes,” Li-wei H. Lehman, Mohammed Saeed, William Long, Joon Lee, Roger G. Mark, Proceedings of the AMIA Annual Symposium, 505-511, Nov. 2012.
• “Latent Topic Discovery of Clinical Concepts from Hospital Discharge Summaries of a Heterogeneous Patient Cohort,” Li-wei H. Lehman, William Long, Mohammmed Saeed, Roger G. Mark, Proceedings of the 36th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Chicago, August 2014.
• “Phenotyping Hypotensive Patients in Critical Care Using Hospital Discharge Summaries,” Yang Dai, Sharukh Lokhandwala, William Long, Roger G. Mark, and Li-wei H. Lehman, Proceedings of the IEEE International Conference on Biomedical and Health Informatics, Orlando FL, Feb 16-17 2017.

Electronic Health Records and Medical Informatics

• “ MIMIC-III, a freely accessible critical care database,” Alistair E. Johnson, Tom J. Pollard, Lu Shen, Li-wei H. Lehman, Mornin Feng, Mohammad Ghassemi, Benjamin Moody, Pete Szolovits, Leo Celi, Roger G. Mark, Scientific Data (Nature publishing) , 3:160035, May 2016. Published online 24 May 2016. (doi:10.1038/sdata.2016.35) (PMID:27219127)
• “Multiparameter intelligent monitoring in intensive care II (MIMIC-II): A public-access intensive care unit database,” Saeed M, Villarroel M, Reisner AT, Clifford G, Lehman LH, Moody G, Heldt T, Kyaw TH, Moody B, Mark RG, Crit Care Med, 39(5):952–960, 2011.