Expanding the Diversity of Texts and Applications: Findings from the Section on Clinical Natural Language Processing of the International Medical Informatics Association Yearbook

Spontaneously generated online patient experience data - how and why is it being used in health research: an umbrella scoping review

PURPOSE

Social media has led to fundamental changes in the way that people look for and share health related information. There is increasing interest in using this spontaneously generated patient experience data as a data source for health research. The aim was to summarise the state of the art regarding how and why SGOPE data has been used in health research. We determined the sites and platforms used as data sources, the purposes of the studies, the tools and methods being used, and any identified research gaps.

METHODS

A scoping umbrella review was conducted looking at review papers from 2015 to Jan 2021 that studied the use of SGOPE data for health research. Using keyword searches we identified 1759 papers from which we included 58 relevant studies in our review.

RESULTS

Data was used from many individual general or health specific platforms, although Twitter was the most widely used data source. The most frequent purposes were surveillance based, tracking infectious disease, adverse event identification and mental health triaging. Despite the developments in machine learning the reviews included lots of small qualitative studies. Most NLP used supervised methods for sentiment analysis and classification. Very early days, methods need development. Methods not being explained. Disciplinary differences - accuracy tweaks vs application. There is little evidence of any work that either compares the results in both methods on the same data set or brings the ideas together.

CONCLUSION

Tools, methods, and techniques are still at an early stage of development, but strong consensus exists that this data source will become very important to patient centred health research.

A Question-and-Answer System to Extract Data From Free-Text Oncological Pathology Reports (CancerBERT Network): Development Study

Background

Information in pathology reports is critical for cancer care. Natural language processing (NLP) systems used to extract information from pathology reports are often narrow in scope or require extensive tuning. Consequently, there is growing interest in automated deep learning approaches. A powerful new NLP algorithm, bidirectional encoder representations from transformers (BERT), was published in late 2018. BERT set new performance standards on tasks as diverse as question answering, named entity recognition, speech recognition, and more.

Objective

The aim of this study is to develop a BERT-based system to automatically extract detailed tumor site and histology information from free-text oncological pathology reports.

Methods

We pursued three specific aims: extract accurate tumor site and histology descriptions from free-text pathology reports, accommodate the diverse terminology used to indicate the same pathology, and provide accurate standardized tumor site and histology codes for use by downstream applications. We first trained a base language model to comprehend the technical language in pathology reports. This involved unsupervised learning on a training corpus of 275,605 electronic pathology reports from 164,531 unique patients that included 121 million words. Next, we trained a question-and-answer (Q&A) model that connects a Q&A layer to the base pathology language model to answer pathology questions. Our Q&A system was designed to search for the answers to two predefined questions in each pathology report: What organ contains the tumor? and What is the kind of tumor or carcinoma? This involved supervised training on 8197 pathology reports, each with ground truth answers to these 2 questions determined by certified tumor registrars. The data set included 214 tumor sites and 193 histologies. The tumor site and histology phrases extracted by the Q&A model were used to predict International Classification of Diseases for Oncology, Third Edition (ICD-O-3), site and histology codes. This involved fine-tuning two additional BERT models: one to predict site codes and another to predict histology codes. Our final system includes a network of 3 BERT-based models. We call this CancerBERT network (caBERTnet). We evaluated caBERTnet using a sequestered test data set of 2050 pathology reports with ground truth answers determined by certified tumor registrars.

Results

caBERTnet’s accuracies for predicting group-level site and histology codes were 93.53% (1895/2026) and 97.6% (1993/2042), respectively. The top 5 accuracies for predicting fine-grained ICD-O-3 site and histology codes with 5 or more samples each in the training data set were 92.95% (1794/1930) and 96.01% (1853/1930), respectively.

Conclusions

We have developed an NLP system that outperforms existing algorithms at predicting ICD-O-3 codes across an extensive range of tumor sites and histologies. Our new system could help reduce treatment delays, increase enrollment in clinical trials of new therapies, and improve patient outcomes.

Sepsis prediction, early detection, and identification using clinical text for machine learning: a systematic review

OBJECTIVE

To determine the effects of using unstructured clinical text in machine learning (ML) for prediction, early detection, and identification of sepsis.

MATERIALS AND METHODS

PubMed, Scopus, ACM DL, dblp, and IEEE Xplore databases were searched. Articles utilizing clinical text for ML or natural language processing (NLP) to detect, identify, recognize, diagnose, or predict the onset, development, progress, or prognosis of systemic inflammatory response syndrome, sepsis, severe sepsis, or septic shock were included. Sepsis definition, dataset, types of data, ML models, NLP techniques, and evaluation metrics were extracted.

RESULTS

The clinical text used in models include narrative notes written by nurses, physicians, and specialists in varying situations. This is often combined with common structured data such as demographics, vital signs, laboratory data, and medications. Area under the receiver operating characteristic curve (AUC) comparison of ML methods showed that utilizing both text and structured data predicts sepsis earlier and more accurately than structured data alone. No meta-analysis was performed because of incomparable measurements among the 9 included studies.

DISCUSSION

Studies focused on sepsis identification or early detection before onset; no studies used patient histories beyond the current episode of care to predict sepsis. Sepsis definition affects reporting methods, outcomes, and results. Many methods rely on continuous vital sign measurements in intensive care, making them not easily transferable to general ward units.

CONCLUSIONS

Approaches were heterogeneous, but studies showed that utilizing both unstructured text and structured data in ML can improve identification and early detection of sepsis.

Deep learning in clinical natural language processing: a methodical review

OBJECTIVE

This article methodically reviews the literature on deep learning (DL) for natural language processing (NLP) in the clinical domain, providing quantitative analysis to answer 3 research questions concerning methods, scope, and context of current research.

MATERIALS AND METHODS

We searched MEDLINE, EMBASE, Scopus, the Association for Computing Machinery Digital Library, and the Association for Computational Linguistics Anthology for articles using DL-based approaches to NLP problems in electronic health records. After screening 1,737 articles, we collected data on 25 variables across 212 papers.

RESULTS

DL in clinical NLP publications more than doubled each year, through 2018. Recurrent neural networks (60.8%) and word2vec embeddings (74.1%) were the most popular methods; the information extraction tasks of text classification, named entity recognition, and relation extraction were dominant (89.2%). However, there was a "long tail" of other methods and specific tasks. Most contributions were methodological variants or applications, but 20.8% were new methods of some kind. The earliest adopters were in the NLP community, but the medical informatics community was the most prolific.

DISCUSSION

Our analysis shows growing acceptance of deep learning as a baseline for NLP research, and of DL-based NLP in the medical community. A number of common associations were substantiated (eg, the preference of recurrent neural networks for sequence-labeling named entity recognition), while others were surprisingly nuanced (eg, the scarcity of French language clinical NLP with deep learning).

CONCLUSION

Deep learning has not yet fully penetrated clinical NLP and is growing rapidly. This review highlighted both the popular and unique trends in this active field.

Extracting and classifying diagnosis dates from clinical notes: A case study

Myeloproliferative neoplasms (MPNs) are chronic hematologic malignancies that may progress over long disease courses. The original date of diagnosis is an important piece of information for patient care and research, but is not consistently documented. We describe an attempt to build a pipeline for extracting dates with natural language processing (NLP) tools and techniques and classifying them as relevant diagnoses or not. Inaccurate and incomplete date extraction and interpretation impacted the performance of the overall pipeline. Existing lightweight Python packages tended to have low specificity for identifying and interpreting partial and relative dates in clinical text. A rules-based regular expression (regex) approach achieved recall of 83.0% on dates manually annotated as diagnosis dates, and 77.4% on all annotated dates. With only 3.8% of annotated dates representing initial MPN diagnoses, additional methods of targeting candidate date instances may alleviate noise and class imbalance.

Deep Learning-Based Structure-Activity Relationship Modeling for Multi-Category Toxicity Classification: A Case Study of 10K Tox21 Chemicals With High-Throughput Cell-Based Androgen Receptor Bioassay Data

Deep learning (DL) has attracted the attention of computational toxicologists as it offers a potentially greater power for in silico predictive toxicology than existing shallow learning algorithms. However, contradicting reports have been documented. To further explore the advantages of DL over shallow learning, we conducted this case study using two cell-based androgen receptor (AR) activity datasets with 10K chemicals generated from the Tox21 program. A nested double-loop cross-validation approach was adopted along with a stratified sampling strategy for partitioning chemicals of multiple AR activity classes (i.e., agonist, antagonist, inactive, and inconclusive) at the same distribution rates amongst the training, validation and test subsets. Deep neural networks (DNN) and random forest (RF), representing deep and shallow learning algorithms, respectively, were chosen to carry out structure-activity relationship-based chemical toxicity prediction. Results suggest that DNN significantly outperformed RF (p < 0.001, ANOVA) by 22–27% for four metrics (precision, recall, F-measure, and AUPRC) and by 11% for another (AUROC). Further in-depth analyses of chemical scaffolding shed insights on structural alerts for AR agonists/antagonists and inactive/inconclusive compounds, which may aid in future drug discovery and improvement of toxicity prediction modeling.