Review
Computer Science, Interdisciplinary Applications
Feng Xie, Han Yuan, Yilin Ning, Marcus Eng Hock Ong, Mengling Feng, Wynne Hsu, Bibhas Chakraborty, Nan Liu
Summary: This study systematically examines deep learning solutions for temporal data representation in electronic health records (EHRs). The study identifies challenges in representing temporal data, such as irregularity, heterogeneity, sparsity, and model opacity. It explores how deep learning techniques address these challenges and discusses open challenges in the field. The study concludes that deep learning solutions can partially address the challenges of temporal EHR data, but future research should focus on designing comprehensive and integrated solutions and incorporating clinical domain knowledge and model interpretability.
JOURNAL OF BIOMEDICAL INFORMATICS
(2022)
Article
Engineering, Biomedical
Ye Liang, Chonghui Guo
Summary: Accurate prediction of heart failure and identification of sub-phenotypes are crucial for timely interventions and treatments, as well as better understanding of disease pathophysiology. This study proposes a novel Patient Representation model called tBNA-PR, which effectively models heterogeneous and temporal Electronic Health Records (EHRs) data and achieves accurate heart failure prediction and reasonable patient stratification. The study also identifies significant features of sub-phenotypes and provides clinical decision support.
BIOCYBERNETICS AND BIOMEDICAL ENGINEERING
(2023)
Review
Computer Science, Information Systems
Irene Li, Jessica Pan, Jeremy Goldwasser, Neha Verma, Wai Pan Wong, Muhammed Yavuz Nuzumlali, Benjamin Rosand, Yixin Li, Matthew Zhang, David Chang, R. Andrew Taylor, Harlan M. Krumholz, Dragomir Radev
Summary: Electronic health records (EHRs) are digital collections of patient healthcare events and observations that play a critical role in healthcare delivery, operations, and research. However, a significant portion of the information stored in EHRs is unstructured text, making it challenging to process automatically. Recent advances in neural network and deep learning methods for Natural Language Processing have shown promise in unlocking the potential of this unstructured text in EHRs.
COMPUTER SCIENCE REVIEW
(2022)
Review
Computer Science, Interdisciplinary Applications
Yuqi Si, Jingcheng Du, Zhao Li, Xiaoqian Jiang, Timothy Miller, Fei Wang, W. Jim Zheng, Kirk Roberts
Summary: Patient representation learning involves developing dense mathematical representations of patients from Electronic Health Records (EHRs) using advanced deep learning methods. Studies from 2015 to 2019 saw a doubling in publications on this topic, with structured EHR data, recurrent neural networks, and supervised learning being commonly used approaches. Disease prediction was the most common application, while privacy concerns and lack of benchmark datasets were challenges faced by researchers in this field.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Review
Computer Science, Information Systems
Lucia A. Carrasco-Ribelles, Jose Llanes-Jurado, Carlos Gallego-Moll, Margarita Cabrera-Bean, Monica Monteagudo-Zaragoza, Concepcion Violan, Edurne Zabaleta-del-Olmo
Summary: The objective of this study is to describe and evaluate the use of artificial intelligence techniques in handling longitudinal data from electronic health records to predict health-related outcomes. The review included 81 studies and found heterogeneity in reporting methodology and results, as well as a lack of public EHR datasets and code sharing, making replication of the research complex.
JOURNAL OF THE AMERICAN MEDICAL INFORMATICS ASSOCIATION
(2023)
Article
Chemistry, Analytical
Atieh Khodadadi, Nima Ghanbari Bousejin, Soheila Molaei, Vinod Kumar Chauhan, Tingting Zhu, David A. Clifton
Summary: An electronic health record (EHR) is a crucial part of medical concepts, and discovering implicit correlations within this data can improve treatment and management. This paper introduces Patient Forest, an innovative approach for learning patient representations from tree-structured data, which outperforms existing models in predicting readmission and mortality. Experiments on MIMIC-III and eICU datasets demonstrate the accuracy and reliability of Patient Forest, especially when training data is limited. The qualitative evaluation using t-SNE visualization further confirms the effectiveness of this model in learning EHR representations.
Article
Computer Science, Interdisciplinary Applications
Meikun Ma, Xiaoyan Hao, Jumin Zhao, Shijie Luo, Yi Liu, Dengao Li
Summary: We propose a deep fusion learning model (DFL-IMP) that utilizes time series and category data from electronic health records to predict in-hospital mortality in patients with heart failure. By considering 41 time series features and 17 category features as predictors, our model achieved the best performance with an AUC of 0.914 when the observation window was 5 days and the prediction window was 30 days. Compared to other baseline models, the DFL-IMP model outperformed them significantly. This tool allows for predicting the expected pathway of heart failure patients and intervening early in the treatment process, thus improving their life expectancy significantly.
MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING
(2023)
Article
Computer Science, Information Systems
Tianhao Li, Zhishun Wang, Wei Lu, Qian Zhang, Dengfeng Li
Summary: The study proposed an EHRs-based reinforcement learning algorithm to optimize sequential treatment strategies for diseases, achieving good results in experiments. The research includes modeling process and reinforcement learning process, utilizing deep Q network to explore optimal insulin dose for patients and extending the algorithm to cooperative learning environment.
INFORMATION SYSTEMS
(2022)
Article
Public, Environmental & Occupational Health
Eric Appiah Mantey, Conghua Zhou, S. R. Srividhya, Sanjiv Kumar Jain, B. Sundaravadivazhagan
Summary: This article introduces the application of blockchain and deep learning, and proposes an integrated environment for analyzing electronic health records. The environment uses deep learning algorithms to analyze data stored in the blockchain and sends reminders to patients.
FRONTIERS IN PUBLIC HEALTH
(2022)
Article
Multidisciplinary Sciences
Yi-Cheng Shen, Te-Chun Hsia, Ching-Hsien Hsu
Summary: This paper presents an adaptive hybridized deep neural network for electronic health records and discusses the importance and applications of electronic health records in the healthcare industry. The research findings demonstrate that deep learning architectures can better handle the temporal structure of electronic health records.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Interdisciplinary Applications
Xinyu Dong, Jianyuan Deng, Wei Hou, Sina Rashidian, Richard N. Rosenthal, Mary Saltz, Joel H. Saltz, Fusheng Wang
Summary: This study aims to predict patients at high risk for opioid overdose using a deep learning model, with LSTM-based models outperforming other methods in predicting overdose risk. The LSTM model with an attention mechanism achieved the highest F-1 score, indicating its effectiveness in identifying predictive features such as medications and vital signs. The study demonstrates the potential of using deep learning models for early detection and intervention to reduce opioid overdose.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Article
Genetics & Heredity
Zeyu Yang, Amy Shikany, Yizhao Ni, Ge Zhang, K. Nicole Weaver, Jing Chen
Summary: This study investigated the utility of using electronic health records (EHRs) to identify patients at high risk of Noonan syndrome (NS) and developed a deep learning model to analyze EHR data. The results showed that the text-based deep learning method performed better than previous methods and could potentially be used as a tool to identify patients with features of rare diseases.
GENETICS IN MEDICINE
(2022)
Article
Medicine, General & Internal
Hossein Estiri, Alaleh Azhir, Deborah L. Blacker, Christine S. Ritchie, Chirag J. Patel, Shawn N. Murphy
Summary: This study developed computational models for identifying Alzheimer's Disease (AD) cohorts and compared the utility of AD diagnosis codes and temporal representations from electronic health records (EHRs) for characterizing AD cohorts. The models with sequential features improved AD classification by 3-16% over the use of diagnosis codes alone. These findings have important implications for accelerating AD research and precision drug development.
Article
Multidisciplinary Sciences
Aixia Guo, Sakima Smith, Yosef M. Khan, James R. Langabeer, Randi E. Foraker
Summary: The study used a deep learning technique-long short-term memory (LSTM) model to predict cardiac dysrhythmias and demonstrated that LSTM model outperformed traditional machine learning models, with blood pressure being the most influential feature. These findings may be used to prevent cardiac dysrhythmias in the outpatient setting.
Article
Computer Science, Artificial Intelligence
Shishir Rao, Mohammad Mamouei, Gholamreza Salimi-Khorshidi, Yikuan Li, Rema Ramakrishnan, Abdelaali Hassaine, Dexter Canoy, Kazem Rahimi
Summary: Observational causal inference plays an important role in decision-making when randomized clinical trials are not feasible. This study explores the use of a transformer-based model coupled with doubly robust estimation for causal modeling in electronic health records. The model provides accurate estimates of risk ratio and shows consistency with results derived from randomized clinical trials.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Information Systems
Sungrim Moon, Sijia Liu, Christopher G. Scott, Sujith Samudrala, Mohamed M. Abidian, Jeffrey B. Geske, Peter A. Noseworthy, Jane L. Shellum, Rajeev Chaudhry, Steve R. Ommen, Rick A. Nishimura, Hongfang Liu, Adelaide M. Arruda-Olson
INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS
(2019)
Article
Computer Science, Interdisciplinary Applications
Feichen Shen, Suyuan Peng, Yadan Fan, Andrew Wen, Sijia Liu, Yanshan Wang, Liwei Wang, Hongfang Liu
JOURNAL OF BIOMEDICAL INFORMATICS
(2019)
Review
Computer Science, Interdisciplinary Applications
Sunyang Fu, David Chen, Huan He, Sijia Liu, Sungrim Moon, Kevin J. Peterson, Feichen Shen, Liwei Wang, Yanshan Wang, Andrew Wen, Yiqing Zhao, Sunghwan Sohn, Hongfang Liu
JOURNAL OF BIOMEDICAL INFORMATICS
(2020)
Article
Medical Informatics
Sijia Liu, Yanshan Wang, Andrew Wen, Liwei Wang, Na Hong, Feichen Shen, Steven Bedrick, William Hersh, Hongfang Liu
JMIR MEDICAL INFORMATICS
(2020)
Article
Computer Science, Interdisciplinary Applications
Andrew Wen, Liwei Wang, Huan He, Sijia Liu, Sunyang Fu, Sunghwan Sohn, Jacob A. Kugel, Vinod C. Kaggal, Ming Huang, Yanshan Wang, Feichen Shen, Jungwei Fan, Hongfang Liu
Summary: After the outbreak of the pandemic, early detection and intervention are key to managing the situation. Syndromic surveillance could offer a timelier screening option, but existing solutions often struggle to distinguish outbreaks of diseases sharing similar symptoms, posing a challenge for monitoring COVID-19.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Article
Medical Informatics
Feichen Shen, Sijia Liu, Sunyang Fu, Yanshan Wang, Sam Henry, Ozlem Uzuner, Hongfang Liu
Summary: The n2c2/OHNLP FH extraction task aimed to standardize evaluation and system development on FH extraction, with 17 teams participating and top performance by Harbin Institute of Technology. Results indicate that relation extraction from FH is more challenging than entity identification task.
JMIR MEDICAL INFORMATICS
(2021)
Article
Biochemical Research Methods
Huan He, Sunyang Fu, Liwei Wang, Sijia Liu, Andrew Wen, Hongfang Liu
Summary: Building a high-quality annotation corpus is time-consuming and requires expertise, but existing annotation tools often have difficulties with installation, integration, and usability. This paper presents MedTator, a new serverless annotation tool with an intuitive and interactive user interface, focusing on the core steps of corpus annotation.
Article
Computer Science, Interdisciplinary Applications
Yue Yu, Nansu Zong, Andrew Wen, Sijia Liu, Daniel J. Stone, David Knaack, Alanna M. Chamberlain, Emily Pfaff, Davera Gabriel, Christopher G. Chute, Nilay Shah, Guoqian Jiang
Summary: This study designed, developed, and evaluated an ETL tool that transforms data from the PCORnet CDM format to the OMOP CDM format. The results showed that the tool successfully converted the data, with minimal information loss and high mapping accuracy. The tool was also able to be used for COVID-19 surveillance and met the data collection criteria for the MN EHR Consortium COVID-19 project.
JOURNAL OF BIOMEDICAL INFORMATICS
(2022)
Review
Oncology
Liwei Wang, Sunyang Fu, Andrew Wen, Xiaoyang Ruan, Huan He, Sijia Liu, Sungrim Moon, Michelle Mai, Irbaz B. Riaz, Nan Wang, Ping Yang, Hua Xu, Jeremy L. Warner, Hongfang Liu
Summary: This review assesses the use of natural language processing (NLP) in electronic health records (EHRs) for cancer research and patient care. The findings highlight the need for additional data elements beyond the Minimal Common Oncology Data Elements (mCODE) for comprehensive analysis and evaluation. The review also identifies challenges and barriers in the adoption of NLP methods for cancer research and patient care.
JCO CLINICAL CANCER INFORMATICS
(2022)
Article
Medical Informatics
Liwei Wang, Huan He, Andrew Wen, Sungrim Moon, Sunyang Fu, Kevin J. Peterson, Xuguang Ai, Sijia Liu, Ramakanth Kavuluru, Hongfang Liu
Summary: Without a standardized method to capture family history (FH) information, FH information in electronic health records is difficult to use in data analytics or clinical decision support applications. This study aimed to construct an FH lexical resource for information extraction and normalization. Using a transformer-based method, a lexicon was developed and demonstrated through the development of rule-based and deep learning-based FH systems. The evaluation showed that the rule-based FH system performed well, and combining rule-based and deep learning-based systems improved FH information recall.
JMIR MEDICAL INFORMATICS
(2023)
Article
Computer Science, Information Systems
Sijia Liu, Andrew Wen, Liwei Wang, Huan He, Sunyang Fu, Robert Miller, Andrew Williams, Daniel Harris, Ramakanth Kavuluru, Mei Liu, Noor Abu-el-Rub, Dalton Schutte, Rui Zhang, Masoud Rouhizadeh, John D. Osborne, Yongqun He, Umit Topaloglu, Stephanie S. Hong, Joel H. Saltz, Thomas Schaffter, Emily Pfaff, Christopher G. Chute, Tim Duong, Melissa A. Haendel, Rafael Fuentes, Peter Szolovits, Hua Xu, Hongfang Liu
Summary: Despite recent advancements in clinical natural language processing (NLP), the adoption of clinical NLP models in translational research is hindered by process heterogeneity and human factor variations. Developing NLP models in multi-site settings is challenging, but essential for algorithm robustness and generalizability. This study reports on the development of an NLP solution for COVID-19 signs and symptom extraction using an open NLP framework, highlighting the benefits of multi-site data and the need for federated annotation and evaluation to overcome challenges.
JOURNAL OF THE AMERICAN MEDICAL INFORMATICS ASSOCIATION
(2023)
Article
Health Care Sciences & Services
Ming Huang, Aditya Khurana, George Mastorakos, Andrew Wen, Huan He, Liwei Wang, Sijia Liu, Yanshan Wang, Nansu Zong, Julie Prigge, Brian Costello, Nilay Shah, Henry Ting, Jungwei Fan, Christi Patten, Hongfang Liu
Summary: This study analyzed patient portal messages during the COVID-19 pandemic to understand patient responses to the crisis. Most messages were related to COVID-19 symptom assessment and testing results. Trends in message usage correlated with national data on new cases and hospitalizations.
JMIR HUMAN FACTORS
(2022)
Article
Health Care Sciences & Services
Himanshu S. Sahoo, Greg M. Silverman, Nicholas E. Ingraham, Monica Lupei, Michael A. Puskarich, Raymond L. Finzel, John Sartori, Rui Zhang, Benjamin C. Knoll, Sijia Liu, Hongfang Liu, Genevieve B. Melton, Christopher J. Tignanelli, Serguei V. S. Pakhomov
Summary: The rule-based gazetteer developed in this study showed superior speed, resource utilization, and performance, providing an effective solution for real-time symptom identification and integration of unstructured data elements into clinical decision support systems. Fine-tuning lexical rules and running on multiple compute nodes were identified as opportunities to further enhance its performance.
Article
Health Care Sciences & Services
Andrew Wen, Sunyang Fu, Sungrim Moon, Mohamed El Wazir, Andrew Rosenbaum, Vinod C. Kaggal, Sijia Liu, Sunghwan Sohn, Hongfang Liu, Jungwei Fan
NPJ DIGITAL MEDICINE
(2019)
Article
Health Care Sciences & Services
David Chen, Sijia Liu, Paul Kingsbury, Sunghwan Sohn, Curtis B. Storlie, Elizabeth B. Habermann, James M. Naessens, David W. Larson, Hongfang Liu
NPJ DIGITAL MEDICINE
(2019)