Bidirectional RNN for Medical Event Detection in Electronic Health Records

The Role and Applications of Artificial Intelligence in the Treatment of Chronic Pain

PURPOSE OF REVIEW

This review aims to explore the interface between artificial intelligence (AI) and chronic pain, seeking to identify areas of focus for enhancing current treatments and yielding novel therapies.

RECENT FINDINGS

In the United States, the prevalence of chronic pain is estimated to be upwards of 40%. Its impact extends to increased healthcare costs, reduced economic productivity, and strain on healthcare resources. Addressing this condition is particularly challenging due to its complexity and the significant variability in how patients respond to treatment. Current options often struggle to provide long-term relief, with their benefits rarely outweighing the risks, such as dependency or other side effects. Currently, AI has impacted four key areas of chronic pain treatment and research: (1) predicting outcomes based on clinical information; (2) extracting features from text, specifically clinical notes; (3) modeling 'omic data to identify meaningful patient subgroups with potential for personalized treatments and improved understanding of disease processes; and (4) disentangling complex neuronal signals responsible for pain, which current therapies attempt to modulate. As AI advances, leveraging state-of-the-art architectures will be essential for improving chronic pain treatment. Current efforts aim to extract meaningful representations from complex data, paving the way for personalized medicine. The identification of unique patient subgroups should reveal targets for tailored chronic pain treatments. Moreover, enhancing current treatment approaches is achievable by gaining a more profound understanding of patient physiology and responses. This can be realized by leveraging AI on the increasing volume of data linked to chronic pain.

Predicting Drugs Suspected of Causing Adverse Drug Reactions Using Graph Features and Attention Mechanisms

BACKGROUND

Adverse drug reactions (ADRs) refer to an unintended harmful reaction that occurs after the administration of a medication for therapeutic purposes, which is unrelated to the intended pharmacological action of the drug. In the United States, ADRs account for 6% of all hospital admissions annually. The cost of ADR-related illnesses in 2016 was estimated at USD 528.4 billion. Increasing the awareness of ADRs is an effective measure to prevent them. Assessing suspected drugs in adverse events helps to enhance the awareness of ADRs.

METHODS

In this study, a suspect drug assisted judgment model (SDAJM) is designed to identify suspected drugs in adverse events. This framework utilizes the graph isomorphism network (GIN) and an attention mechanism to extract features based on patients' demographic information, drug information, and ADR information.

RESULTS

By comparing it with other models, the results of various tests show that this model performs well in predicting the suspected drugs in adverse reaction events. ADR signal detection was conducted on a group of cardiovascular system drugs, and case analyses were performed on two classic drugs, Mexiletine and Captopril, as well as on two classic antithyroid drugs. The results indicate that the model can accomplish the task of predicting drug ADRs. Validation using benchmark datasets from ten drug discovery domains shows that the model is applicable to classification tasks on the Tox21 and SIDER datasets.

CONCLUSIONS

This study applies deep learning methods to construct the SDAJM model for three purposes: (1) identifying drugs suspected to cause adverse drug events (ADEs), (2) predicting the ADRs of drugs, and (3) other drug discovery tasks. The results indicate that this method can offer new directions for research in the field of ADRs.

Machine learning models to detect and predict patient safety events using electronic health records: A systematic review

INTRODUCTION

Identifying patient safety events using electronic health records (EHRs) and automated machine learning-based detection methods can help improve the efficiency and quality of healthcare service provision.

OBJECTIVE

This study aimed to systematically review machine learning-based methods and techniques, as well as their results for patient safety event management using EHRs.

METHODS

We reviewed the studies that focused on machine learning techniques, including automatic prediction and detection of patient safety events and medical errors through EHR analysis to manage patient safety events. The data were collected by searching Scopus, PubMed (Medline), Web of Science, EMBASE, and IEEE Xplore databases.

RESULTS

After screening, 41 papers were reviewed. Support vector machine (SVM), random forest, conditional random field (CRF), and bidirectional long short-term memory with conditional random field (BiLSTM-CRF) algorithms were mostly applied to predict, identify, and classify patient safety events using EHRs; however, they had different performances. BiLSTM-CRF was employed in most of the studies to extract and identify concepts, e.g., adverse drug events (ADEs) and adverse drug reactions (ADRs), as well as relationships between drug and severity, drug and ADEs, drug and ADRs. Recurrent neural networks (RNN) and BiLSTM-CRF had the best results in detecting ADEs compared to other patient safety events. Linear classifiers and Naive Bayes (NB) had the highest performance for ADR detection. Logistic regression had the best results in detecting surgical site infections. According to the findings, the quality of articles has non-significantly improved in recent years, but they had low average scores.

CONCLUSIONS

Machine learning can be useful in automatic detection and prediction of patient safety events. However, most of these algorithms have not yet been externally validated or prospectively tested. Therefore, further studies are required to improve the performance of these automated systems.

BIR: Biomedical Information Retrieval System for Cancer Treatment in Electronic Health Record Using Transformers

The rapid growth of electronic health records (EHRs) has led to unprecedented biomedical data. Clinician access to the latest patient information can improve the quality of healthcare. However, clinicians have difficulty finding information quickly and easily due to the sheer data mining volume. Biomedical information retrieval (BIR) systems can help clinicians find the information required by automatically searching EHRs and returning relevant results. However, traditional BIR systems cannot understand the complex relationships between EHR entities. Transformers are a new type of neural network that is very effective for natural language processing (NLP) tasks. As a result, transformers are well suited for tasks such as machine translation and text summarization. In this paper, we propose a new BIR system for EHRs that uses transformers for predicting cancer treatment from EHR. Our system can understand the complex relationships between the different entities in an EHR, which allows it to return more relevant results to clinicians. We evaluated our system on a dataset of EHRs and found that it outperformed state-of-the-art BIR systems on various tasks, including medical question answering and information extraction. Our results show that Transformers are a promising approach for BIR in EHRs, reaching an accuracy and an F1-score of 86.46%, and 0.8157, respectively. We believe that our system can help clinicians find the information they need more quickly and easily, leading to improved patient care.

Adverse drug event detection using natural language processing: A scoping review of supervised learning methods

To reduce adverse drug events (ADEs), hospitals need a system to support them in monitoring ADE occurrence routinely, rapidly, and at scale. Natural language processing (NLP), a computerized approach to analyze text data, has shown promising results for the purpose of ADE detection in the context of pharmacovigilance. However, a detailed qualitative assessment and critical appraisal of NLP methods for ADE detection in the context of ADE monitoring in hospitals is lacking. Therefore, we have conducted a scoping review to close this knowledge gap, and to provide directions for future research and practice. We included articles where NLP was applied to detect ADEs in clinical narratives within electronic health records of inpatients. Quantitative and qualitative data items relating to NLP methods were extracted and critically appraised. Out of 1,065 articles screened for eligibility, 29 articles met the inclusion criteria. Most frequent tasks included named entity recognition (n = 17; 58.6%) and relation extraction/classification (n = 15; 51.7%). Clinical involvement was reported in nine studies (31%). Multiple NLP modelling approaches seem suitable, with Long Short Term Memory and Conditional Random Field methods most commonly used. Although reported overall performance of the systems was high, it provides an inflated impression given a steep drop in performance when predicting the ADE entity or ADE relation class. When annotating corpora, treating an ADE as a relation between a drug and non-drug entity seems the best practice. Future research should focus on semi-automated methods to reduce the manual annotation effort, and examine implementation of the NLP methods in practice.

iADRGSE: A Graph-Embedding and Self-Attention Encoding for Identifying Adverse Drug Reaction in the Earlier Phase of Drug Development

Adverse drug reactions (ADRs) are a major issue to be addressed by the pharmaceutical industry. Early and accurate detection of potential ADRs contributes to enhancing drug safety and reducing financial expenses. The majority of the approaches that have been employed to identify ADRs are limited to determining whether a drug exhibits an ADR, rather than identifying the exact type of ADR. By introducing the "multi-level feature-fusion deep-learning model", a new predictor, called iADRGSE, has been developed, which can be used to identify adverse drug reactions at the early stage of drug discovery. iADRGSE integrates a self-attentive module and a graph-network module that can extract one-dimensional sub-structure sequence information and two-dimensional chemical-structure graph information of drug molecules. As a demonstration, cross-validation and independent testing were performed with iADRGSE on a dataset of ADRs classified into 27 categories, based on SOC (system organ classification). In addition, experiments comparing iADRGSE with approaches such as NPF were conducted on the OMOP dataset, using the jackknife test method. Experiments show that iADRGSE was superior to existing state-of-the-art predictors.

Extraction and Quantification of Words Representing Degrees of Diseases: Combining the Fuzzy C-Means Method and Gaussian Membership

Background

Due to the development of medical data, a large amount of clinical data has been generated. These unstructured data contain substantial information. Extracting useful knowledge from this data and making scientific decisions for diagnosing and treating diseases have become increasingly necessary. Unstructured data, such as in the Marketplace for Medical Information in Intensive Care III (MIMIC-III) data set, contain several ambiguous words that demonstrate the subjectivity of doctors, such as descriptions of patient symptoms. These data could be used to further improve the accuracy of medical diagnostic system assessments. To the best of our knowledge, there is currently no method for extracting subjective words that express the extent of these symptoms (hereinafter, “degree words”).

Objective

Therefore, we propose using the fuzzy c-means (FCM) method and Gaussian membership to quantify the degree words in the clinical medical data set MIMIC-III.

Methods

First, we preprocessed the 381,091 radiology reports collected in MIMIC-III, and then we used the FCM method to extract degree words from unstructured text. Thereafter, we used the Gaussian membership method to quantify the extracted degree words, which transform the fuzzy words extracted from the medical text into computer-recognizable numbers.

Results

The results showed that the digitization of ambiguous words in medical texts is feasible. The words representing each degree of each disease had a range of corresponding values. Examples of membership medians were 2.971 (atelectasis), 3.121 (pneumonia), 2.899 (pneumothorax), 3.051 (pulmonary edema), and 2.435 (pulmonary embolus). Additionally, all extracted words contained the same subjective words (low, high, etc), which allows for an objective evaluation method. Furthermore, we will verify the specific impact of the quantification results of ambiguous words such as symptom words and degree words on the use of medical texts in subsequent studies. These same ambiguous words may be used as a new set of feature values to represent the disorders.

Conclusions

This study proposes an innovative method for handling subjective words. We used the FCM method to extract the subjective degree words in the English-interpreted report of the MIMIC-III and then used the Gaussian functions to quantify the subjective degree words. In this method, words containing subjectivity in unstructured texts can be automatically processed and transformed into numerical ranges by digital processing. It was concluded that the digitization of ambiguous words in medical texts is feasible.

The Use of Artificial Intelligence in Pharmacovigilance: A Systematic Review of the Literature

INTRODUCTION

Artificial intelligence through machine learning uses algorithms and prior learnings to make predictions. Recently, there has been interest to include more artificial intelligence in pharmacovigilance of products already in the market and pharmaceuticals in development.

OBJECTIVE

The aim of this study was to identify and describe the uses of artificial intelligence in pharmacovigilance through a systematic literature review.

METHODS

Embase and MEDLINE database searches were conducted for articles published from January 1, 2015 to July 9, 2021 using search terms such as 'pharmacovigilance,' 'patient safety,' 'artificial intelligence,' and 'machine learning' in the title or abstract. Scientific articles that contained information on the use of artificial intelligence in all modalities of patient safety or pharmacovigilance were reviewed and synthesized using a pre-specified data extraction template. Articles with incomplete information and letters to editor, notes, and commentaries were excluded.

RESULTS

Sixty-six articles were identified for evaluation. Most relevant articles on artificial intelligence focused on machine learning, and it was used in patient safety in the identification of adverse drug events (ADEs) and adverse drug reactions (ADRs) (57.6%), processing safety reports (21.2%), extraction of drug-drug interactions (7.6%), identification of populations at high risk for drug toxicity or guidance for personalized care (7.6%), prediction of side effects (3.0%), simulation of clinical trials (1.5%), and integration of prediction uncertainties into diagnostic classifiers to increase patient safety (1.5%). Artificial intelligence has been used to identify safety signals through automated processes and training with machine learning models; however, the findings may not be generalizable given that there were different types of data included in each source.

CONCLUSION

Artificial intelligence allows for the processing and analysis of large amounts of data and can be applied to various disease states. The automation and machine learning models can optimize pharmacovigilance processes and provide a more efficient way to analyze information relevant to safety, although more research is needed to identify if this optimization has an impact on the quality of safety analyses. It is expected that its use will increase in the near future, particularly with its role in the prediction of side effects and ADRs.

A Novel Encoder-Decoder Model for Multivariate Time Series Forecasting

The time series is a kind of complex structure data, which contains some special characteristics such as high dimension, dynamic, and high noise. Moreover, multivariate time series (MTS) has become a crucial study in data mining. The MTS utilizes the historical data to forecast its variation trend and has turned into one of the hotspots. In the era of rapid information development and big data, accurate prediction of MTS has attracted much attention. In this paper, a novel deep learning architecture based on the encoder-decoder framework is proposed for MTS forecasting. In this architecture, firstly, the gated recurrent unit (GRU) is taken as the main unit structure of both the procedures in encoding and decoding to extract the useful successive feature information. Then, different from the existing models, the attention mechanism (AM) is introduced to exploit the importance of different historical data for reconstruction at the decoding stage. Meanwhile, feature reuse is realized by skip connections based on the residual network for alleviating the influence of previous features on data reconstruction. Finally, in order to enhance the performance and the discriminative ability of the new MTS, the convolutional structure and fully connected module are established. Furthermore, to better validate the effectiveness of MTS forecasting, extensive experiments are executed on two different types of MTS such as stock data and shared bicycle data, respectively. The experimental results adequately demonstrate the effectiveness and the feasibility of the proposed method.

Automatic extraction of 12 cardiovascular concepts from German discharge letters using pre-trained language models

Objective

A vast amount of medical data is still stored in unstructured text documents. We present an automated method of information extraction from German unstructured clinical routine data from the cardiology domain enabling their usage in state-of-the-art data-driven deep learning projects.

Methods

We evaluated pre-trained language models to extract a set of 12 cardiovascular concepts in German discharge letters. We compared three bidirectional encoder representations from transformers pre-trained on different corpora and fine-tuned them on the task of cardiovascular concept extraction using 204 discharge letters manually annotated by cardiologists at the University Hospital Heidelberg. We compared our results with traditional machine learning methods based on a long short-term memory network and a conditional random field.

Results

Our best performing model, based on publicly available German pre-trained bidirectional encoder representations from the transformer model, achieved a token-wise micro-average F1-score of 86% and outperformed the baseline by at least 6%. Moreover, this approach achieved the best trade-off between precision (positive predictive value) and recall (sensitivity).

Conclusion

Our results show the applicability of state-of-the-art deep learning methods using pre-trained language models for the task of cardiovascular concept extraction using limited training data. This minimizes annotation efforts, which are currently the bottleneck of any application of data-driven deep learning projects in the clinical domain for German and many other European languages.

Transformer + GAN based Traditional Chinese Medicine inpatient prescription recommendation (Preprint)

Background

Traditional Chinese medicine (TCM) practitioners usually follow a 4-step evaluation process during patient diagnosis: observation, auscultation, olfaction, inquiry, pulse feeling, and palpation. The information gathered in this process, along with laboratory test results and other measurements such as vital signs, is recorded in the patient’s electronic health record (EHR). In fact, all the information needed to make a treatment plan is contained in the EHR; however, only a seasoned TCM physician could use this information well to make a good treatment plan as the reasoning process is very complicated, and it takes years of practice for a medical graduate to master the reasoning skill. In this digital medicine era, with a deluge of medical data, ever-increasing computing power, and more advanced artificial neural network models, it is not only desirable but also readily possible for a computerized system to mimic the decision-making process of a TCM physician.

Objective

This study aims to develop an assistive tool that can predict prescriptions for inpatients in a hospital based on patients’ clinical EHRs.

Methods

Clinical health records containing medical histories, as well as current symptoms and diagnosis information, were used to train a transformer-based neural network model using the corresponding physician’s prescriptions as the target. This was accomplished by extracting relevant information, such as the patient’s current illness, medicines taken, nursing care given, vital signs, examinations, and laboratory results from the patient’s EHRs. The obtained information was then sorted chronologically to produce a sequence of data for the patient. These time sequence data were then used as input to a modified transformer network, which was chosen as a prescription prediction model. The output of the model was the prescription for the patient. The ultimate goal is for this tool to generate a prescription that matches what an expert TCM physician would prescribe. To alleviate the issue of overfitting, a generative adversarial network was used to augment the training sample data set by generating noise-added samples from the original training samples.

Results

In total, 21,295 copies of inpatient electronic medical records from Guang’anmen Hospital were used in this study. These records were generated between January 2017 and December 2018, covering 6352 types of medicines. These medicines were sorted into 819 types of first-category medicines based on their class relationships. As shown by the test results, the performance of a fully trained transformer model can have an average precision rate of 80.58% and an average recall rate of 68.49%.

Conclusions

As shown by the preliminary test results, the transformer-based TCM prescription recommendation model outperformed the existing conventional methods. The extra training samples generated by the generative adversarial network help to overcome the overfitting issue, leading to further improved recall and precision rates.

A Context-Enhanced De-identification System

Many modern entity recognition systems, including the current state-of-the-art de-identification systems, are based on bidirectional long short-term memory (biLSTM) units augmented by a conditional random field (CRF) sequence optimizer. These systems process the input sentence by sentence. This approach prevents the systems from capturing dependencies over sentence boundaries and makes accurate sentence boundary detection a prerequisite. Since sentence boundary detection can be problematic especially in clinical reports, where dependencies and co-references across sentence boundaries are abundant, these systems have clear limitations. In this study, we built a new system on the framework of one of the current state-of-the-art de-identification systems, NeuroNER, to overcome these limitations. This new system incorporates context embeddings through forward and backward n -grams without using sentence boundaries. Our context-enhanced de-identification (CEDI) system captures dependencies over sentence boundaries and bypasses the sentence boundary detection problem altogether. We enhanced this system with deep affix features and an attention mechanism to capture the pertinent parts of the input. The CEDI system outperforms NeuroNER on the 2006 i2b2 de-identification challenge dataset, the 2014 i2b2 shared task de-identification dataset, and the 2016 CEGS N-GRID de-identification dataset (p < 0.01). All datasets comprise narrative clinical reports in English but contain different note types varying from discharge summaries to psychiatric notes. Enhancing CEDI with deep affix features and the attention mechanism further increased performance.

Medical Text Classification Using Hybrid Deep Learning Models with Multihead Attention

To unlock information present in clinical description, automatic medical text classification is highly useful in the arena of natural language processing (NLP). For medical text classification tasks, machine learning techniques seem to be quite effective; however, it requires extensive effort from human side, so that the labeled training data can be created. For clinical and translational research, a huge quantity of detailed patient information, such as disease status, lab tests, medication history, side effects, and treatment outcomes, has been collected in an electronic format, and it serves as a valuable data source for further analysis. Therefore, a huge quantity of detailed patient information is present in the medical text, and it is quite a huge challenge to process it efficiently. In this work, a medical text classification paradigm, using two novel deep learning architectures, is proposed to mitigate the human efforts. The first approach is that a quad channel hybrid long short-term memory (QC-LSTM) deep learning model is implemented utilizing four channels, and the second approach is that a hybrid bidirectional gated recurrent unit (BiGRU) deep learning model with multihead attention is developed and implemented successfully. The proposed methodology is validated on two medical text datasets, and a comprehensive analysis is conducted. The best results in terms of classification accuracy of 96.72% is obtained with the proposed QC-LSTM deep learning model, and a classification accuracy of 95.76% is obtained with the proposed hybrid BiGRU deep learning model.

High-throughput phenotyping with temporal sequences

OBJECTIVE

High-throughput electronic phenotyping algorithms can accelerate translational research using data from electronic health record (EHR) systems. The temporal information buried in EHRs is often underutilized in developing computational phenotypic definitions. This study aims to develop a high-throughput phenotyping method, leveraging temporal sequential patterns from EHRs.

MATERIALS AND METHODS

We develop a representation mining algorithm to extract 5 classes of representations from EHR diagnosis and medication records: the aggregated vector of the records (aggregated vector representation), the standard sequential patterns (sequential pattern mining), the transitive sequential patterns (transitive sequential pattern mining), and 2 hybrid classes. Using EHR data on 10 phenotypes from the Mass General Brigham Biobank, we train and validate phenotyping algorithms.

RESULTS

Phenotyping with temporal sequences resulted in a superior classification performance across all 10 phenotypes compared with the standard representations in electronic phenotyping. The high-throughput algorithm's classification performance was superior or similar to the performance of previously published electronic phenotyping algorithms. We characterize and evaluate the top transitive sequences of diagnosis records paired with the records of risk factors, symptoms, complications, medications, or vaccinations.

DISCUSSION

The proposed high-throughput phenotyping approach enables seamless discovery of sequential record combinations that may be difficult to assume from raw EHR data. Transitive sequences offer more accurate characterization of the phenotype, compared with its individual components, and reflect the actual lived experiences of the patients with that particular disease.

CONCLUSION

Sequential data representations provide a precise mechanism for incorporating raw EHR records into downstream machine learning. Our approach starts with user interpretability and works backward to the technology.

Automatic Prediction of Recurrence of Major Cardiovascular Events: A Text Mining Study Using Chest X-Ray Reports

Methods

We used EHR data of patients included in the Second Manifestations of ARTerial disease (SMART) study. We propose a deep learning-based multimodal architecture for our text mining pipeline that integrates neural text representation with preprocessed clinical predictors for the prediction of recurrence of major cardiovascular events in cardiovascular patients. Text preprocessing, including cleaning and stemming, was first applied to filter out the unwanted texts from X-ray radiology reports. Thereafter, text representation methods were used to numerically represent unstructured radiology reports with vectors. Subsequently, these text representation methods were added to prediction models to assess their clinical relevance. In this step, we applied logistic regression, support vector machine (SVM), multilayer perceptron neural network, convolutional neural network, long short-term memory (LSTM), and bidirectional LSTM deep neural network (BiLSTM).

Results

We performed various experiments to evaluate the added value of the text in the prediction of major cardiovascular events. The two main scenarios were the integration of radiology reports (1) with classical clinical predictors and (2) with only age and sex in the case of unavailable clinical predictors. In total, data of 5603 patients were used with 5-fold cross-validation to train the models. In the first scenario, the multimodal BiLSTM (MI-BiLSTM) model achieved an area under the curve (AUC) of 84.7%, misclassification rate of 14.3%, and F1 score of 83.8%. In this scenario, the SVM model, trained on clinical variables and bag-of-words representation, achieved the lowest misclassification rate of 12.2%. In the case of unavailable clinical predictors, the MI-BiLSTM model trained on radiology reports and demographic (age and sex) variables reached an AUC, F1 score, and misclassification rate of 74.5%, 70.8%, and 20.4%, respectively.

Conclusions

Using the case study of routine care chest X-ray radiology reports, we demonstrated the clinical relevance of integrating text features and classical predictors in our text mining pipeline for cardiovascular risk prediction. The MI-BiLSTM model with word embedding representation appeared to have a desirable performance when trained on text data integrated with the clinical variables from the SMART study. Our results mined from chest X-ray reports showed that models using text data in addition to laboratory values outperform those using only known clinical predictors.

Improved machine learning performances with transfer learning to predicting need for hospitalization in arboviral infections against the small dataset

The prediction of hospital patients and outpatients with suspected arboviral infection individuals in research-limited settings of the urban areas is defined as a challenging process for clinicians. Dengue, Chikungunya, and Zika arboviruses have gained attention in recent years because of the high prevalence in the society and financial burden of major global health systems. In this study, we proposed a machine learning algorithm based prediction model over retrospective medical records, which are named as SISA (the Severity Index for Suspected Arbovirus) and SISAL (the Severity Index for Suspected Arbovirus with Laboratory) datasets. Therefore, we aim to inform the clinicians about the use of machine learning with transfer learning success for diagnosis and comprehensive comparison of the classification performances over the SISA/SISAL datasets in the resource-limited settings that may cause to the small datasets of arboviral infection. In this study, Convolutional Neural Network and Long Short-Term Memory have achieved 100% accuracy and 1 of area under the curve (AUC) score, Fully Connected Deep Network has provided 92.86% accuracy and 0.969 AUC score in the SISAL dataset with transfer learning. Moreover, 98.73% accuracy and 0.988 AUC score were obtained by Convolutional Neural Network and Long Short-Term Memory for the SISA dataset. Furthermore, Linear Discriminant Analysis (shallow algorithm) has provided reaching up to 96.43% accuracy. Notably, deep learning based models have achieved improved performances compared to the previously reported study.

Medical Specialty Recommendations by an Artificial Intelligence Chatbot on a Smartphone: Development and Deployment

Background

The COVID-19 pandemic has limited daily activities and even contact between patients and primary care providers. This makes it more difficult to provide adequate primary care services, which include connecting patients to an appropriate medical specialist. A smartphone-compatible artificial intelligence (AI) chatbot that classifies patients’ symptoms and recommends the appropriate medical specialty could provide a valuable solution.

Objective

In order to establish a contactless method of recommending the appropriate medical specialty, this study aimed to construct a deep learning–based natural language processing (NLP) pipeline and to develop an AI chatbot that can be used on a smartphone.

Methods

We collected 118,008 sentences containing information on symptoms with labels (medical specialty), conducted data cleansing, and finally constructed a pipeline of 51,134 sentences for this study. Several deep learning models, including 4 different long short-term memory (LSTM) models with or without attention and with or without a pretrained FastText embedding layer, as well as bidirectional encoder representations from transformers for NLP, were trained and validated using a randomly selected test data set. The performance of the models was evaluated on the basis of the precision, recall, F₁-score, and area under the receiver operating characteristic curve (AUC). An AI chatbot was also designed to make it easy for patients to use this specialty recommendation system. We used an open-source framework called “Alpha” to develop our AI chatbot. This takes the form of a web-based app with a frontend chat interface capable of conversing in text and a backend cloud-based server application to handle data collection, process the data with a deep learning model, and offer the medical specialty recommendation in a responsive web that is compatible with both desktops and smartphones.

Results

The bidirectional encoder representations from transformers model yielded the best performance, with an AUC of 0.964 and F₁-score of 0.768, followed by LSTM model with embedding vectors, with an AUC of 0.965 and F₁-score of 0.739. Considering the limitations of computing resources and the wide availability of smartphones, the LSTM model with embedding vectors trained on our data set was adopted for our AI chatbot service. We also deployed an Alpha version of the AI chatbot to be executed on both desktops and smartphones.

Conclusions

With the increasing need for telemedicine during the current COVID-19 pandemic, an AI chatbot with a deep learning–based NLP model that can recommend a medical specialty to patients through their smartphones would be exceedingly useful. This chatbot allows patients to identify the proper medical specialist in a rapid and contactless manner, based on their symptoms, thus potentially supporting both patients and primary care providers.

Predicting mortality and hospitalization in heart failure using machine learning: A systematic literature review

Objective

The partnership between humans and machines can enhance clinical decisions accuracy, leading to improved patient outcomes. Despite this, the application of machine learning techniques in the healthcare sector, particularly in guiding heart failure patient management, remains unpopular. This systematic review aims to identify factors restricting the integration of machine learning derived risk scores into clinical practice when treating adults with acute and chronic heart failure.

Methods

Four academic research databases and Google Scholar were searched to identify original research studies where heart failure patient data was used to build models predicting all-cause mortality, cardiac death, all-cause and heart failure-related hospitalization.

Results

Thirty studies met the inclusion criteria. The selected studies' sample size ranged between 71 and 716 790 patients, and the median age was 72.1 (interquartile range: 61.1–76.8) years. The minimum and maximum area under the receiver operating characteristic curve (AUC) for models predicting mortality were 0.48 and 0.92, respectively. Models predicting hospitalization had an AUC of 0.47 to 0.84. Nineteen studies (63%) used logistic regression, 53% random forests, and 37% of studies used decision trees to build predictive models. None of the models were built or externally validated using data originating from Africa or the Middle-East.

Conclusions

The variation in the aetiologies of heart failure, limited access to structured health data, distrust in machine learning techniques among clinicians and the modest accuracy of existing predictive models are some of the factors precluding the widespread use of machine learning derived risk calculators.

Health-adjusted life expectancy (HALE) in Chongqing, China, 2017: An artificial intelligence and big data method estimating the burden of disease at city level

BACKGROUND

A universally applicable approach that provides standard HALE measurements for different regions has yet to be developed because of the difficulties of health information collection. In this study, we developed a natural language processing (NLP) based HALE estimation approach by using individual-level electronic medical records (EMRs), which made it possible to calculate HALE timely in different temporal or spatial granularities.

METHODS

We performed diagnostic concept extraction and normalisation on 13•99 million EMRs with NLP to estimate the prevalence of 254 diseases in WHO Global Burden of Disease Study (GBD). Then, we calculated HALE in Chongqing, 2017, by using the life table technique and Sullivan's method, and analysed the contribution of diseases to the expected years "lost" due to disability (DLE).

FINDINGS

Our method identified a life expectancy at birth (LE0) of 77•9 years and health-adjusted life expectancy at birth (HALE0) of 71•7 years for the general Chongqing population of 2017. In particular, the male LE0 and HALE0 were 76•3 years and 68•9 years, respectively, while the female LE0 and HALE0 were 80•0 years and 74•4 years, respectively. Cerebrovascular diseases, cancers, and injuries were the top three deterioration factors, which reduced HALE by 2•67, 2•15, and 1•19 years, respectively.

INTERPRETATION

The results demonstrated the feasibility and effectiveness of EMRs-based HALE estimation. Moreover, the method allowed for a potentially transferable framework that facilitated a more convenient comparison of cross-sectional and longitudinal studies on HALE between regions. In summary, this study provided insightful solutions to the global ageing and health problems that the world is facing.

FUNDING

National Key R and D Program of China (2018YFC2000400).

Extracting medications and associated adverse drug events using a natural language processing system combining knowledge base and deep learning

OBJECTIVE

Detecting adverse drug events (ADEs) and medications related information in clinical notes is important for both hospital medical care and medical research. We describe our clinical natural language processing (NLP) system to automatically extract medical concepts and relations related to ADEs and medications from clinical narratives. This work was part of the 2018 National NLP Clinical Challenges Shared Task and Workshop on Adverse Drug Events and Medication Extraction.

MATERIALS AND METHODS

The authors developed a hybrid clinical NLP system that employs a knowledge-based general clinical NLP system for medical concepts extraction, and a task-specific deep learning system for relations identification using attention-based bidirectional long short-term memory networks.

RESULTS

The systems were evaluated as part of the 2018 National NLP Clinical Challenges challenge, and our attention-based bidirectional long short-term memory networks based system obtained an F-measure of 0.9442 for relations identification task, ranking fifth at the challenge, and had <2% difference from the best system. Error analysis was also conducted targeting at figuring out the root causes and possible approaches for improvement.

CONCLUSIONS

We demonstrate the generic approaches and the practice of connecting general purposed clinical NLP system to task-specific requirements with deep learning methods. Our results indicate that a well-designed hybrid NLP system is capable of ADE and medication-related information extraction, which can be used in real-world applications to support ADE-related researches and medical decisions.

Identifying relations of medications with adverse drug events using recurrent convolutional neural networks and gradient boosting

OBJECTIVE

To develop a natural language processing system that identifies relations of medications with adverse drug events from clinical narratives. This project is part of the 2018 n2c2 challenge.

MATERIALS AND METHODS

We developed a novel clinical named entity recognition method based on an recurrent convolutional neural network and compared it to a recurrent neural network implemented using the long-short term memory architecture, explored methods to integrate medical knowledge as embedding layers in neural networks, and investigated 3 machine learning models, including support vector machines, random forests and gradient boosting for relation classification. The performance of our system was evaluated using annotated data and scripts provided by the 2018 n2c2 organizers.

RESULTS

Our system was among the top ranked. Our best model submitted during this challenge (based on recurrent neural networks and support vector machines) achieved lenient F1 scores of 0.9287 for concept extraction (ranked third), 0.9459 for relation classification (ranked fourth), and 0.8778 for the end-to-end relation extraction (ranked second). We developed a novel named entity recognition model based on a recurrent convolutional neural network and further investigated gradient boosting for relation classification. The new methods improved the lenient F1 scores of the 3 subtasks to 0.9292, 0.9633, and 0.8880, respectively, which are comparable to the best performance reported in this challenge.

CONCLUSION

This study demonstrated the feasibility of using machine learning methods to extract the relations of medications with adverse drug events from clinical narratives.

Extracting entities with attributes in clinical text via joint deep learning

OBJECTIVE

Extracting clinical entities and their attributes is a fundamental task of natural language processing (NLP) in the medical domain. This task is typically recognized as 2 sequential subtasks in a pipeline, clinical entity or attribute recognition followed by entity-attribute relation extraction. One problem of pipeline methods is that errors from entity recognition are unavoidably passed to relation extraction. We propose a novel joint deep learning method to recognize clinical entities or attributes and extract entity-attribute relations simultaneously.

MATERIALS AND METHODS

The proposed method integrates 2 state-of-the-art methods for named entity recognition and relation extraction, namely bidirectional long short-term memory with conditional random field and bidirectional long short-term memory, into a unified framework. In this method, relation constraints between clinical entities and attributes and weights of the 2 subtasks are also considered simultaneously. We compare the method with other related methods (ie, pipeline methods and other joint deep learning methods) on an existing English corpus from SemEval-2015 and a newly developed Chinese corpus.

RESULTS

Our proposed method achieves the best F1 of 74.46% on entity recognition and the best F1 of 50.21% on relation extraction on the English corpus, and 89.32% and 88.13% on the Chinese corpora, respectively, which outperform the other methods on both tasks.

CONCLUSIONS

The joint deep learning-based method could improve both entity recognition and relation extraction from clinical text in both English and Chinese, indicating that the approach is promising.

GHS-NET a generic hybridized shallow neural network for multi-label biomedical text classification

Exponential growth of biomedical literature and clinical data demands more robust yet precise computational methodologies to extract useful insights from biomedical literature and to perform accurate assignment of disease-specific codes. Such approaches can largely enhance the effectiveness of diverse biomedicine and bioinformatics applications. State-of-the-art computational biomedical text classification methodologies either solely leverage discrimintaive features extracted through convolution operations performed by deep convolutional neural network or contextual information extracted by recurrent neural network. However, none of the methodology takes advantage of both convolutional and recurrent neural networks. Further, existing methodologies lack to produce decent performance for the classification of different genre biomedical text such as biomedical literature or clinical notes. We, for the very first time, present a generic deep learning based hybrid multi-label classification methodology namely GHS-NET which can be utilized to accurately classify biomedical text of diverse genre. GHS-NET makes use of convolutional neural network to extract most discriminative features and bi-directional Long Short-Term Memory to acquire contextual information. GHS-NET effectiveness is evaluated for extreme multi-label biomedical literature classification and assignment of ICD-9 codes to clinical notes. For the task of extreme multi-label biomedical literature classification, performance comparison of GHS-Net and state-of-the-art deep learning based methodology reveals that GHS-Net marks the increment of 1%, 6%, and 1% for hallmarks of cancer dataset, 10%, 16%, and 11% for chemical exposure dataset in terms of precision, recall, and F1-score. For the task of clinical notes classification, GHS-Net outperforms previous best deep learning based methodology over Medical Information Mart for Intensive Care dataset (MIMIC-III) by the significant margin of 6%, 8% in terms of recall and F1-score. GHS-NET is available as a web service at¹ and potentially can be used to accurately classify multi-variate disease and chemical exposure specific text.

Evaluating shallow and deep learning strategies for the 2018 n2c2 shared task on clinical text classification

Objective

Automated clinical phenotyping is challenging because word-based features quickly turn it into a high-dimensional problem, in which the small, privacy-restricted, training datasets might lead to overfitting. Pretrained embeddings might solve this issue by reusing input representation schemes trained on a larger dataset. We sought to evaluate shallow and deep learning text classifiers and the impact of pretrained embeddings in a small clinical dataset.

Materials and Methods

We participated in the 2018 National NLP Clinical Challenges (n2c2) Shared Task on cohort selection and received an annotated dataset with medical narratives of 202 patients for multilabel binary text classification. We set our baseline to a majority classifier, to which we compared a rule-based classifier and orthogonal machine learning strategies: support vector machines, logistic regression, and long short-term memory neural networks. We evaluated logistic regression and long short-term memory using both self-trained and pretrained BioWordVec word embeddings as input representation schemes.

Results

Rule-based classifier showed the highest overall micro F₁ score (0.9100), with which we finished first in the challenge. Shallow machine learning strategies showed lower overall micro F₁ scores, but still higher than deep learning strategies and the baseline. We could not show a difference in classification efficiency between self-trained and pretrained embeddings.

Discussion

Clinical context, negation, and value-based criteria hindered shallow machine learning approaches, while deep learning strategies could not capture the term diversity due to the small training dataset.

Conclusion

Shallow methods for clinical phenotyping can still outperform deep learning methods in small imbalanced data, even when supported by pretrained embeddings.

Bleeding Entity Recognition in Electronic Health Records: A Comprehensive Analysis of End-to-End Systems

A bleeding event is a common adverse drug reaction amongst patients on anticoagulation and factors critically into a clinician's decision to prescribe or continue anticoagulation for atrial fibrillation. However, bleeding events are not uniformly captured in the administrative data of electronic health records (EHR). As manual review is prohibitively expensive, we investigate the effectiveness of various natural language processing (NLP) methods for automatic extraction of bleeding events. Using our expert-annotated 1,079 de-identified EHR notes, we evaluated state-of-the-art NLP models such as biLSTM-CRF with language modeling, and different BERT variants for six entity types. On our dataset, the biLSTM-CRF surpassed other models resulting in a macro F1-score of 0.75 whereas the performance difference is negligible for sentence and document-level predictions with the best macro F1-scores of 0.84 and 0.96, respectively. Our error analyses suggest that the models' incorrect predictions can be attributed to variability in entity spans, memorization, and missing negation signals.

Inferring ADR causality by predicting the Naranjo Score from Clinical Notes

Clinical judgment studies are an integral part of drug safety surveillance and pharmacovigilance frameworks. They help quantify the causal relationship between medication and its adverse drug reactions (ADRs). To conduct such studies, physicians need to review patients' charts manually to answer Naranjo questionnaire¹. In this paper, we propose a methodology to automatically infer causal relations from patients' discharge summaries by combining the capabilities of deep learning and statistical learning models. We use Bidirectional Encoder Representations from Transformers (BERT)² to extract relevant paragraphs for each Naranjo question and then use a statistical learning model such as logistic regression to predict the Naranjo score and the causal relation between the medication and an ADR. Our methodology achieves a macro-averaged f1-score of 0.50 and weighted f1-score of 0.63.

Neural Multi-Task Learning for Adverse Drug Reaction Extraction

A reliable and searchable knowledge database of adverse drug reactions (ADRs) is highly important and valuable for improving patient safety at the point of care. In this paper, we proposed a neural multi-task learning system, NeuroADR, to extract ADRs as well as relevant modifiers from free-text drug labels. Specifically, the NeuroADR system exploited a hierarchical multi-task learning (HMTL) framework to perform named entity recognition (NER) and relation extraction (RE) jointly, where interactions among the learned deep encoder representations from different subtasks are explored. Different from the conventional HMTL approach, NeuroADR adopted a novel task decomposition strategy to generate auxiliary subtasks for more inter-task interactions and integrated a new label encoding schema for better handling discontinuous entities. Experimental results demonstrate the effectiveness of the proposed system.

A Novel Protein Mapping Method for Predicting the Protein Interactions in COVID-19 Disease by Deep Learning

The new type of corona virus (SARS-COV-2) emerging in Wuhan, China has spread rapidly to the world and has become a pandemic. In addition to having a significant impact on daily life, it also shows its effect in different areas, including public health and economy. Currently, there is no vaccine or antiviral drug available to prevent the COVID-19 disease. Therefore, determination of protein interactions of new types of corona virus is vital in clinical studies, drug therapy, identification of preclinical compounds and protein functions. Protein–protein interactions are important to examine protein functions and pathways involved in various biological processes and to determine the cause and progression of diseases. Various high-throughput experimental methods have been used to identify protein–protein interactions in organisms, yet, there is still a huge gap in specifying all possible protein interactions in an organism. In addition, since the experimental methods used include cloning, labeling, affinity purification mass spectrometry, the processes take a long time. Determining these interactions with artificial intelligence-based methods rather than experimental approaches may help to identify protein functions faster. Thus, protein–protein interaction prediction using deep-learning algorithms has been employed in conjunction with experimental method to explore new protein interactions. However, to predict protein interactions with artificial intelligence techniques, protein sequences need to be mapped. There are various types and numbers of protein-mapping methods in the literature. In this study, we wanted to contribute to the literature by proposing a novel protein-mapping method based on the AVL tree. The proposed method was inspired by the fast search performance on the dictionary structure of AVL tree and was used to verify the protein interactions between SARS-COV-2 virus and human. First, protein sequences were mapped by both the proposed method and various protein-mapping methods. Then, the mapped protein sequences were normalized and classified by bidirectional recurrent neural networks. The performance of the proposed method was evaluated with accuracy, f1-score, precision, recall, and AUC scores. Our results indicated that our mapping method predicts the protein interactions between SARS-COV-2 virus proteins and human proteins at an accuracy of 97.76%, precision of 97.60%, recall of 98.33%, f1-score of 79.42%, and with AUC 89% in average.

Domain specific word embeddings for natural language processing in radiology

BACKGROUND

There has been increasing interest in machine learning based natural language processing (NLP) methods in radiology; however, models have often used word embeddings trained on general web corpora due to lack of a radiology-specific corpus.

PURPOSE

We examined the potential of Radiopaedia to serve as a general radiology corpus to produce radiology specific word embeddings that could be used to enhance performance on a NLP task on radiological text.

MATERIALS AND METHODS

Embeddings of dimension 50, 100, 200, and 300 were trained on articles collected from Radiopaedia using a GloVe algorithm and evaluated on analogy completion. A shallow neural network using input from either our trained embeddings or pre-trained Wikipedia 2014 + Gigaword 5 (WG) embeddings was used to label the Radiopaedia articles. Labeling performance was evaluated based on exact match accuracy and Hamming loss. The McNemar's test with continuity and the Benjamini-Hochberg correction and a 5×2 cross validation paired two-tailed t-test were used to assess statistical significance.

RESULTS

For accuracy in the analogy task, 50-dimensional (50-D) Radiopaedia embeddings outperformed WG embeddings on tumor origin analogies (p < 0.05) and organ adjectives (p < 0.01) whereas WG embeddings tended to outperform on inflammation location and bone vs. muscle analogies (p < 0.01). The two embeddings had comparable performance on other subcategories. In the labeling task, the Radiopaedia-based model outperformed the WG based model at 50, 100, 200, and 300-D for exact match accuracy (p < 0.001, p < 0.001, p < 0.01, and p < 0.05, respectively) and Hamming loss (p < 0.001, p < 0.001, p < 0.01, and p < 0.05, respectively).

CONCLUSION

We have developed a set of word embeddings from Radiopaedia and shown that they can preserve relevant medical semantics and augment performance on a radiology NLP task. Our results suggest that the cultivation of a radiology-specific corpus can benefit radiology NLP models in the future.

Detecting Miscoded Diabetes Diagnosis Codes in Electronic Health Records for Quality Improvement: Temporal Deep Learning Approach

BACKGROUND

Diabetes affects more than 30 million patients across the United States. With such a large disease burden, even a small error in classification can be significant. Currently billing codes, assigned at the time of a medical encounter, are the "gold standard" reflecting the actual diseases present in an individual, and thus in aggregate reflect disease prevalence in the population. These codes are generated by highly trained coders and by health care providers but are not always accurate.

OBJECTIVE

This work provides a scalable deep learning methodology to more accurately classify individuals with diabetes across multiple health care systems.

METHODS

We leveraged a long short-term memory-dense neural network (LSTM-DNN) model to identify patients with or without diabetes using data from 5 acute care facilities with 187,187 patients and 275,407 encounters, incorporating data elements including laboratory test results, diagnostic/procedure codes, medications, demographic data, and admission information. Furthermore, a blinded physician panel reviewed discordant cases, providing an estimate of the total impact on the population.

RESULTS

When predicting the documented diagnosis of diabetes, our model achieved an 84% F1 score, 96% area under the curve-receiver operating characteristic curve, and 91% average precision on a heterogeneous data set from 5 distinct health facilities. However, in 81% of cases where the model disagreed with the documented phenotype, a blinded physician panel agreed with the model. Taken together, this suggests that 4.3% of our studied population have either missing or improper diabetes diagnosis.

CONCLUSIONS

This study demonstrates that deep learning methods can improve clinical phenotyping even when patient data are noisy, sparse, and heterogeneous.

Application of Augmented Intelligence for Pharmacovigilance Case Seriousness Determination

INTRODUCTION

Identification of adverse events and determination of their seriousness ensures timely detection of potential patient safety concerns. Adverse event seriousness is a key factor in defining reporting timelines and is often performed manually by pharmacovigilance experts. The dramatic increase in the volume of safety reports necessitates exploration of scalable solutions that also meet reporting timeline requirements.

OBJECTIVE

The aim of this study was to develop an augmented intelligence methodology for automatically identifying adverse event seriousness in spontaneous, solicited, and medical literature safety reports. Deep learning models were evaluated for accuracy and/or the F1 score against a ground truth labeled by pharmacovigilance experts.

METHODS

Using a stratified random sample of safety reports received by Celgene, we developed three neural networks for addressing identification of adverse event seriousness: (1) a binary adverse-event level seriousness classifier; (2) a classifier for determining seriousness categorization at the adverse-event level; and (3) an annotator for identifying seriousness criteria terms to provide supporting evidence at the document level.

RESULTS

The seriousness classifier achieved an accuracy of 83.0% in post-marketing reports, 92.9% in solicited reports, and 86.3% in medical literature reports. F1 scores for seriousness categorization were 77.7 for death, 78.9 for hospitalization, and 75.5 for important medical events. The seriousness annotator achieved an F1 score of 89.9 in solicited reports, and 75.2 in medical literature reports.

CONCLUSIONS

The results of this study indicate that a neural network approach can provide an accurate and scalable solution for potentially augmenting pharmacovigilance practitioner determination of adverse event seriousness in spontaneous, solicited, and medical literature reports.

Medical Information Extraction in the Age of Deep Learning

OBJECTIVES

We survey recent developments in medical Information Extraction (IE) as reported in the literature from the past three years. Our focus is on the fundamental methodological paradigm shift from standard Machine Learning (ML) techniques to Deep Neural Networks (DNNs). We describe applications of this new paradigm concentrating on two basic IE tasks, named entity recognition and relation extraction, for two selected semantic classes-diseases and drugs (or medications)-and relations between them.

METHODS

For the time period from 2017 to early 2020, we searched for relevant publications from three major scientific communities: medicine and medical informatics, natural language processing, as well as neural networks and artificial intelligence.

RESULTS

In the past decade, the field of Natural Language Processing (NLP) has undergone a profound methodological shift from symbolic to distributed representations based on the paradigm of Deep Learning (DL). Meanwhile, this trend is, although with some delay, also reflected in the medical NLP community. In the reporting period, overwhelming experimental evidence has been gathered, as illustrated in this survey for medical IE, that DL-based approaches outperform non-DL ones by often large margins. Still, small-sized and access-limited corpora create intrinsic problems for data-greedy DL as do special linguistic phenomena of medical sublanguages that have to be overcome by adaptive learning strategies.

CONCLUSIONS

The paradigm shift from (feature-engineered) ML to DNNs changes the fundamental methodological rules of the game for medical NLP. This change is by no means restricted to medical IE but should also deeply influence other areas of medical informatics, either NLP- or non-NLP-based.

Machine learning mortality classification in clinical documentation with increased accuracy in visual-based analyses

AIM

The role of machine learning on clinical documentation for predictive outcomes remains undefined. We aimed to compare three neural networks on inpatient providers' notes to predict mortality in neonatal hypoxic-ischaemic encephalopathy (HIE).

METHODS

Using Children's Hospitals Neonatal Database, non-anomalous neonates with HIE treated with therapeutic hypothermia were identified at a single-centre. Data were linked with the initial seven days of documentation. Exposures were derived using the databases and applying convolutional and two recurrent neural networks. The primary outcome was mortality. The predictive accuracy and performance measures for models were determined.

RESULTS

The cohort included 52 eligible infants. Most infants survived (n = 36, 69%) and 23 had severe HIE (44%). Neural networks performed above baseline and differed in their median accuracy for predicting mortality (P = .0001): recurrent models with long short-term memory 69% (25^th , 75^th percentile 65, 73%) and gated-recurrent model units 65% (62, 69%) and convolutional 72% (64, 96%). Convolutional networks' median specificity was 81% (72, 97%).

CONCLUSION

The neural network models demonstrated fundamental validity in predicting mortality using inpatient provider documentation. Convolutional models had high specificity for (excluding) mortality in neonatal HIE. These findings provide a platform for future model training and ultimately tool development to assist clinicians in patient assessments and risk stratifications.

A hybrid method of recurrent neural network and graph neural network for next-period prescription prediction

Electronic health records (EHRs) have been widely used to help physicians to make decisions by predicting medical events such as diseases, prescriptions, outcomes, and so on. How to represent patient longitudinal medical data is the key to making these predictions. Recurrent neural network (RNN) is a popular model for patient longitudinal medical data representation from the view of patient status sequences, but it cannot represent complex interactions among different types of medical information, i.e., temporal medical event graphs, which can be represented by graph neural network (GNN). In this paper, we propose a hybrid method of RNN and GNN, called RGNN, for next-period prescription prediction from two views, where RNN is used to represent patient status sequences, and GNN is used to represent temporal medical event graphs. Experiments conducted on the public MIMIC-III ICU data show that the proposed method is effective for next-period prescription prediction, and RNN and GNN are mutually complementary.

Using case-level context to classify cancer pathology reports

Individual electronic health records (EHRs) and clinical reports are often part of a larger sequence-for example, a single patient may generate multiple reports over the trajectory of a disease. In applications such as cancer pathology reports, it is necessary not only to extract information from individual reports, but also to capture aggregate information regarding the entire cancer case based off case-level context from all reports in the sequence. In this paper, we introduce a simple modular add-on for capturing case-level context that is designed to be compatible with most existing deep learning architectures for text classification on individual reports. We test our approach on a corpus of 431,433 cancer pathology reports, and we show that incorporating case-level context significantly boosts classification accuracy across six classification tasks-site, subsite, laterality, histology, behavior, and grade. We expect that with minimal modifications, our add-on can be applied towards a wide range of other clinical text-based tasks.

Clinical Text Data in Machine Learning: Systematic Review

Background

Clinical narratives represent the main form of communication within health care, providing a personalized account of patient history and assessments, and offering rich information for clinical decision making. Natural language processing (NLP) has repeatedly demonstrated its feasibility to unlock evidence buried in clinical narratives. Machine learning can facilitate rapid development of NLP tools by leveraging large amounts of text data.

Objective

The main aim of this study was to provide systematic evidence on the properties of text data used to train machine learning approaches to clinical NLP. We also investigated the types of NLP tasks that have been supported by machine learning and how they can be applied in clinical practice.

Methods

Our methodology was based on the guidelines for performing systematic reviews. In August 2018, we used PubMed, a multifaceted interface, to perform a literature search against MEDLINE. We identified 110 relevant studies and extracted information about text data used to support machine learning, NLP tasks supported, and their clinical applications. The data properties considered included their size, provenance, collection methods, annotation, and any relevant statistics.

Results

The majority of datasets used to train machine learning models included only hundreds or thousands of documents. Only 10 studies used tens of thousands of documents, with a handful of studies utilizing more. Relatively small datasets were utilized for training even when much larger datasets were available. The main reason for such poor data utilization is the annotation bottleneck faced by supervised machine learning algorithms. Active learning was explored to iteratively sample a subset of data for manual annotation as a strategy for minimizing the annotation effort while maximizing the predictive performance of the model. Supervised learning was successfully used where clinical codes integrated with free-text notes into electronic health records were utilized as class labels. Similarly, distant supervision was used to utilize an existing knowledge base to automatically annotate raw text. Where manual annotation was unavoidable, crowdsourcing was explored, but it remains unsuitable because of the sensitive nature of data considered. Besides the small volume, training data were typically sourced from a small number of institutions, thus offering no hard evidence about the transferability of machine learning models. The majority of studies focused on text classification. Most commonly, the classification results were used to support phenotyping, prognosis, care improvement, resource management, and surveillance.

Conclusions

We identified the data annotation bottleneck as one of the key obstacles to machine learning approaches in clinical NLP. Active learning and distant supervision were explored as a way of saving the annotation efforts. Future research in this field would benefit from alternatives such as data augmentation and transfer learning, or unsupervised learning, which do not require data annotation.

Annotating and detecting phenotypic information for chronic obstructive pulmonary disease

Objectives

Chronic obstructive pulmonary disease (COPD) phenotypes cover a range of lung abnormalities. To allow text mining methods to identify pertinent and potentially complex information about these phenotypes from textual data, we have developed a novel annotated corpus, which we use to train a neural network-based named entity recognizer to detect fine-grained COPD phenotypic information.

Materials and methods

Since COPD phenotype descriptions often mention other concepts within them (proteins, treatments, etc.), our corpus annotations include both outermost phenotype descriptions and concepts nested within them. Our neural layered bidirectional long short-term memory conditional random field (BiLSTM-CRF) network firstly recognizes nested mentions, which are fed into subsequent BiLSTM-CRF layers, to help to recognize enclosing phenotype mentions.

Results

Our corpus of 30 full papers (available at: http://www.nactem.ac.uk/COPD) is annotated by experts with 27 030 phenotype-related concept mentions, most of which are automatically linked to UMLS Metathesaurus concepts. When trained using the corpus, our BiLSTM-CRF network outperforms other popular approaches in recognizing detailed phenotypic information.

Discussion

Information extracted by our method can facilitate efficient location and exploration of detailed information about phenotypes, for example, those specifically concerning reactions to treatments.

Conclusion

The importance of our corpus for developing methods to extract fine-grained information about COPD phenotypes is demonstrated through its successful use to train a layered BiLSTM-CRF network to extract phenotypic information at various levels of granularity. The minimal human intervention needed for training should permit ready adaption to extracting phenotypic information about other diseases.

Recent advances in Swedish and Spanish medical entity recognition in clinical texts using deep neural approaches

BACKGROUND

Text mining and natural language processing of clinical text, such as notes from electronic health records, requires specific consideration of the specialized characteristics of these texts. Deep learning methods could potentially mitigate domain specific challenges such as limited access to in-domain tools and data sets.

METHODS

A bi-directional Long Short-Term Memory network is applied to clinical notes in Spanish and Swedish for the task of medical named entity recognition. Several types of embeddings, both generated from in-domain and out-of-domain text corpora, and a number of generation and combination strategies for embeddings have been evaluated in order to investigate different input representations and the influence of domain on the final results.

RESULTS

For Spanish, a micro averaged F1-score of 75.25 was obtained and for Swedish, the corresponding score was 76.04. The best results for both languages were achieved using embeddings generated from in-domain corpora extracted from electronic health records, but embeddings generated from related domains were also found to be beneficial.

CONCLUSIONS

A recurrent neural network with in-domain embeddings improved the medical named entity recognition compared to shallow learning methods, showing this combination to be suitable for entity recognition in clinical text for both languages.

Classifying relations in clinical narratives using segment graph convolutional and recurrent neural networks (Seg-GCRNs)

We propose to use segment graph convolutional and recurrent neural networks (Seg-GCRNs), which use only word embedding and sentence syntactic dependencies, to classify relations from clinical notes without manual feature engineering. In this study, the relations between 2 medical concepts are classified by simultaneously learning representations of text segments in the context of sentence syntactic dependency: preceding, concept1, middle, concept2, and succeeding segments. Seg-GCRN was systematically evaluated on the i2b2/VA relation classification challenge datasets. Experiments show that Seg-GCRN attains state-of-the-art micro-averaged F-measure for all 3 relation categories: 0.692 for classifying medical treatment-problem relations, 0.827 for medical test-problem relations, and 0.741 for medical problem-medical problem relations. Comparison with the previous state-of-the-art segment convolutional neural network (Seg-CNN) suggests that adding syntactic dependency information helps refine medical word embedding and improves concept relation classification without manual feature engineering. Seg-GCRN can be trained efficiently for the i2b2/VA dataset on a GPU platform.

Naranjo Question Answering using End-to-End Multi-task Learning Model

In the clinical domain, it is important to understand whether an adverse drug reaction (ADR) is caused by a particular medication. Clinical judgement studies help judge the causal relation between a medication and its ADRs. In this study, we present the first attempt to automatically infer the causality between a drug and an ADR from electronic health records (EHRs) by answering the Naranjo questionnaire, the validated clinical question answering set used by domain experts for ADR causality assessment. Using physicians' annotation as the gold standard, our proposed joint model, which uses multi-task learning to predict the answers of a subset of the Naranjo questionnaire, significantly outperforms the baseline pipeline model with a good margin, achieving a macro-weighted f-score between 0.3652 - 0.5271 and micro-weighted f-score between 0.9523 - 0.9918.

Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma

BACKGROUND

Predictive modeling with longitudinal electronic health record (EHR) data offers great promise for accelerating personalized medicine and better informs clinical decision-making. Recently, deep learning models have achieved state-of-the-art performance for many healthcare prediction tasks. However, deep models lack interpretability, which is integral to successful decision-making and can lead to better patient care. In this paper, we build upon the contextual decomposition (CD) method, an algorithm for producing importance scores from long short-term memory networks (LSTMs). We extend the method to bidirectional LSTMs (BiLSTMs) and use it in the context of predicting future clinical outcomes using patients' EHR historical visits.

METHODS

We use a real EHR dataset comprising 11071 patients, to evaluate and compare CD interpretations from LSTM and BiLSTM models. First, we train LSTM and BiLSTM models for the task of predicting which pre-school children with respiratory system-related complications will have asthma at school-age. After that, we conduct quantitative and qualitative analysis to evaluate the CD interpretations produced by the contextual decomposition of the trained models. In addition, we develop an interactive visualization to demonstrate the utility of CD scores in explaining predicted outcomes.

RESULTS

Our experimental evaluation demonstrate that whenever a clear visit-level pattern exists, the models learn that pattern and the contextual decomposition can appropriately attribute the prediction to the correct pattern. In addition, the results confirm that the CD scores agree to a large extent with the importance scores generated using logistic regression coefficients. Our main insight was that rather than interpreting the attribution of individual visits to the predicted outcome, we could instead attribute a model's prediction to a group of visits.

CONCLUSION

We presented a quantitative and qualitative evidence that CD interpretations can explain patient-specific predictions using CD attributions of individual visits or a group of visits.

Automatic Detection of Hypoglycemic Events From the Electronic Health Record Notes of Diabetes Patients: Empirical Study

BACKGROUND

Hypoglycemic events are common and potentially dangerous conditions among patients being treated for diabetes. Automatic detection of such events could improve patient care and is valuable in population studies. Electronic health records (EHRs) are valuable resources for the detection of such events.

OBJECTIVE

In this study, we aim to develop a deep-learning-based natural language processing (NLP) system to automatically detect hypoglycemic events from EHR notes. Our model is called the High-Performing System for Automatically Detecting Hypoglycemic Events (HYPE).

METHODS

Domain experts reviewed 500 EHR notes of diabetes patients to determine whether each sentence contained a hypoglycemic event or not. We used this annotated corpus to train and evaluate HYPE, the high-performance NLP system for hypoglycemia detection. We built and evaluated both a classical machine learning model (ie, support vector machines [SVMs]) and state-of-the-art neural network models.

RESULTS

We found that neural network models outperformed the SVM model. The convolutional neural network (CNN) model yielded the highest performance in a 10-fold cross-validation setting: mean precision=0.96 (SD 0.03), mean recall=0.86 (SD 0.03), and mean F1=0.91 (SD 0.03).

CONCLUSIONS

Despite the challenges posed by small and highly imbalanced data, our CNN-based HYPE system still achieved a high performance for hypoglycemia detection. HYPE can be used for EHR-based hypoglycemia surveillance and population studies in diabetes patients.

Use of Natural Language Processing to Extract Clinical Cancer Phenotypes from Electronic Medical Records

Current models for correlating electronic medical records with -omics data largely ignore clinical text, which is an important source of phenotype information for patients with cancer. This data convergence has the potential to reveal new insights about cancer initiation, progression, metastasis, and response to treatment. Insights from this real-world data will catalyze clinical care, research, and regulatory activities. Natural language processing (NLP) methods are needed to extract these rich cancer phenotypes from clinical text. Here, we review the advances of NLP and information extraction methods relevant to oncology based on publications from PubMed as well as NLP and machine learning conference proceedings in the last 3 years. Given the interdisciplinary nature of the fields of oncology and information extraction, this analysis serves as a critical trail marker on the path to higher fidelity oncology phenotypes from real-world data.